Use Blackwell FlashInfer MXFP4 MoE by default if available (#23008)

Signed-off-by: mgoin <mgoin64@gmail.com>

.buildkite/lm-eval-harness/run-lm-eval-gsm-vllm-baseline.sh

@@ -46,6 +46,6 @@ while getopts "m:b:l:f:t:" OPT; do
 done
 
 lm_eval --model vllm \
-  --model_args "pretrained=$MODEL,tensor_parallel_size=$TP_SIZE,distributed_executor_backend=ray,trust_remote_code=true,max_model_len=4096" \
+  --model_args "pretrained=$MODEL,tensor_parallel_size=$TP_SIZE,add_bos_token=true,trust_remote_code=true,max_model_len=4096" \
   --tasks gsm8k --num_fewshot "$FEWSHOT" --limit "$LIMIT" \
   --batch_size "$BATCH_SIZE"

.buildkite/lm-eval-harness/test_lm_eval_correctness.py

@@ -18,12 +18,14 @@ RTOL = 0.08
 
 def launch_lm_eval(eval_config, tp_size):
     trust_remote_code = eval_config.get("trust_remote_code", False)
+    max_model_len = eval_config.get("max_model_len", 4096)
     model_args = (
         f"pretrained={eval_config['model_name']},"
         f"tensor_parallel_size={tp_size},"
         f"enforce_eager=true,"
         f"add_bos_token=true,"
-        f"trust_remote_code={trust_remote_code}"
+        f"trust_remote_code={trust_remote_code},"
+        f"max_model_len={max_model_len}"
     )
     results = lm_eval.simple_evaluate(
         model="vllm",

.buildkite/nightly-benchmarks/README.md

@@ -7,11 +7,11 @@ This directory contains two sets of benchmark for vllm.
 - Performance benchmark: benchmark vllm's performance under various workload, for **developers** to gain clarity on whether their PR improves/degrades vllm's performance
 - Nightly benchmark: compare vllm's performance against alternatives (tgi, trt-llm and lmdeploy), for **the public** to know when to choose vllm.
 
-See [vLLM performance dashboard](https://perf.vllm.ai) for the latest performance benchmark results and [vLLM GitHub README](https://github.com/vllm-project/vllm/blob/main/README.md) for latest nightly benchmark results.
+See [vLLM performance dashboard](https://hud.pytorch.org/benchmark/llms?repoName=vllm-project%2Fvllm) for the latest performance benchmark results and [vLLM GitHub README](https://github.com/vllm-project/vllm/blob/main/README.md) for latest nightly benchmark results.
 
 ## Performance benchmark quick overview
 
-**Benchmarking Coverage**: latency, throughput and fix-qps serving on A100 (the support for FP8 benchmark on H100 is coming!), with different models.
+**Benchmarking Coverage**: latency, throughput and fix-qps serving on A100 (the support for FP8 benchmark on H100 is coming!) and Intel® Xeon® Processors, with different models.
 
 **Benchmarking Duration**: about 1hr.
 
@@ -28,16 +28,34 @@ See [vLLM performance dashboard](https://perf.vllm.ai) for the latest performanc
 ## Trigger the benchmark
 
 Performance benchmark will be triggered when:
+
 - A PR being merged into vllm.
 - Every commit for those PRs with `perf-benchmarks` label AND `ready` label.
 
+Manually Trigger the benchmark
+
+```bash
+bash .buildkite/nightly-benchmarks/scripts/run-performance-benchmarks.sh
+```
+
+Runtime environment variables:
+
+- `ON_CPU`: set the value to '1' on Intel® Xeon® Processors. Default value is 0.
+- `SERVING_JSON`: JSON file to use for the serving tests. Default value is empty string (use default file).
+- `LATENCY_JSON`: JSON file to use for the latency tests. Default value is empty string (use default file).
+- `THROUGHPUT_JSON`: JSON file to use for the throughout tests. Default value is empty string (use default file).
+- `REMOTE_HOST`: IP for the remote vLLM service to benchmark. Default value is empty string.
+- `REMOTE_PORT`: Port for the remote vLLM service to benchmark. Default value is empty string.
+
 Nightly benchmark will be triggered when:
+
 - Every commit for those PRs with `perf-benchmarks` label and `nightly-benchmarks` label.
 
 ## Performance benchmark details
 
 See [performance-benchmarks-descriptions.md](performance-benchmarks-descriptions.md) for detailed descriptions, and use `tests/latency-tests.json`, `tests/throughput-tests.json`, `tests/serving-tests.json` to configure the test cases.
-
+> NOTE: For Intel® Xeon® Processors, use `tests/latency-tests-cpu.json`, `tests/throughput-tests-cpu.json`, `tests/serving-tests-cpu.json` instead.
+>
 ### Latency test
 
 Here is an example of one test inside `latency-tests.json`:
@@ -60,7 +78,7 @@ Here is an example of one test inside `latency-tests.json`:
 In this example:
 
 - The `test_name` attributes is a unique identifier for the test. In `latency-tests.json`, it must start with `latency_`.
-- The `parameters` attribute control the command line arguments to be used for `benchmark_latency.py`. Note that please use underline `_` instead of the dash `-` when specifying the command line arguments, and `run-performance-benchmarks.sh` will convert the underline to dash when feeding the arguments to `benchmark_latency.py`. For example, the corresponding command line arguments for `benchmark_latency.py` will be `--model meta-llama/Meta-Llama-3-8B --tensor-parallel-size 1 --load-format dummy --num-iters-warmup 5 --num-iters 15`
+- The `parameters` attribute control the command line arguments to be used for `vllm bench latency`. Note that please use underline `_` instead of the dash `-` when specifying the command line arguments, and `run-performance-benchmarks.sh` will convert the underline to dash when feeding the arguments to `vllm bench latency`. For example, the corresponding command line arguments for `vllm bench latency` will be `--model meta-llama/Meta-Llama-3-8B --tensor-parallel-size 1 --load-format dummy --num-iters-warmup 5 --num-iters 15`
 
 Note that the performance numbers are highly sensitive to the value of the parameters. Please make sure the parameters are set correctly.
 
@@ -68,13 +86,13 @@ WARNING: The benchmarking script will save json results by itself, so please do
 
 ### Throughput test
 
-The tests are specified in `throughput-tests.json`. The syntax is similar to `latency-tests.json`, except for that the parameters will be fed forward to `benchmark_throughput.py`.
+The tests are specified in `throughput-tests.json`. The syntax is similar to `latency-tests.json`, except for that the parameters will be fed forward to `vllm bench throughput`.
 
 The number of this test is also stable -- a slight change on the value of this number might vary the performance numbers by a lot.
 
 ### Serving test
 
-We test the throughput by using `benchmark_serving.py` with request rate = inf to cover the online serving overhead. The corresponding parameters are in `serving-tests.json`, and here is an example:
+We test the throughput by using `vllm bench serve` with request rate = inf to cover the online serving overhead. The corresponding parameters are in `serving-tests.json`, and here is an example:
 
 ```json
 [
@@ -86,7 +104,6 @@ We test the throughput by using `benchmark_serving.py` with request rate = inf t
             "tensor_parallel_size": 1,
             "swap_space": 16,
             "disable_log_stats": "",
-            "disable_log_requests": "",
             "load_format": "dummy"
         },
         "client_parameters": {
@@ -104,8 +121,8 @@ Inside this example:
 
 - The `test_name` attribute is also a unique identifier for the test. It must start with `serving_`.
 - The `server-parameters` includes the command line arguments for vLLM server.
-- The `client-parameters` includes the command line arguments for `benchmark_serving.py`.
-- The `qps_list` controls the list of qps for test. It will be used to configure the `--request-rate` parameter in `benchmark_serving.py`
+- The `client-parameters` includes the command line arguments for `vllm bench serve`.
+- The `qps_list` controls the list of qps for test. It will be used to configure the `--request-rate` parameter in `vllm bench serve`
 
 The number of this test is less stable compared to the delay and latency benchmarks (due to randomized sharegpt dataset sampling inside `benchmark_serving.py`), but a large change on this number (e.g. 5% change) still vary the output greatly.
 
@@ -119,6 +136,23 @@ If you do not see the table, please wait till the benchmark finish running.
 The json version of the table (together with the json version of the benchmark) will be also attached to the markdown file.
 The raw benchmarking results (in the format of json files) are in the `Artifacts` tab of the benchmarking.
 
+The `compare-json-results.py` helps to compare benchmark results JSON files converted using `convert-results-json-to-markdown.py`.
+When run, benchmark script generates results under `benchmark/results` folder, along with the `benchmark_results.md` and `benchmark_results.json`.
+`compare-json-results.py` compares two `benchmark_results.json` files and provides performance ratio e.g. for Output Tput, Median TTFT and Median TPOT.  
+If only one benchmark_results.json is passed, `compare-json-results.py` compares different TP and PP configurations in the benchmark_results.json instead.
+
+Here is an example using the script to compare result_a and result_b with Model, Dataset name, input/output lenght, max concurrency and qps.
+`python3 compare-json-results.py -f results_a/benchmark_results.json -f results_b/benchmark_results.json`
+
+|   | Model | Dataset Name | Input Len | Output Len | # of max concurrency | qps  | results_a/benchmark_results.json | results_b/benchmark_results.json | perf_ratio        |
+|----|---------------------------------------|--------|-----|-----|------|-----|-----------|----------|----------|
+| 0  | meta-llama/Meta-Llama-3.1-8B-Instruct | random | 128 | 128 | 1000 | 1 | 142.633982                             | 156.526018                             | 1.097396 |
+| 1  | meta-llama/Meta-Llama-3.1-8B-Instruct | random | 128 | 128 | 1000 | inf| 241.620334                             | 294.018783                             | 1.216863 |
+
+A comparison diagram will be generated below the table.
+Here is an example to compare between 96c/results_gnr_96c_091_tp2pp3 and 128c/results_gnr_128c_091_tp2pp3
+<img width="1886" height="828" alt="image" src="https://github.com/user-attachments/assets/c02a43ef-25d0-4fd6-90e5-2169a28682dd" />
+
 ## Nightly test details
 
 See [nightly-descriptions.md](nightly-descriptions.md) for the detailed description on test workload, models and docker containers of benchmarking other llm engines.
@@ -126,9 +160,9 @@ See [nightly-descriptions.md](nightly-descriptions.md) for the detailed descript
 ### Workflow
 
 - The [nightly-pipeline.yaml](nightly-pipeline.yaml) specifies the docker containers for different LLM serving engines.
-- Inside each container, we run [run-nightly-suite.sh](run-nightly-suite.sh), which will probe the serving engine of the current container.
-- The `run-nightly-suite.sh` will redirect the request to `tests/run-[llm serving engine name]-nightly.sh`, which parses the workload described in [nightly-tests.json](tests/nightly-tests.json) and performs the benchmark.
-- At last, we run [scripts/plot-nightly-results.py](scripts/plot-nightly-results.py) to collect and plot the final benchmarking results, and update the results to buildkite.
+- Inside each container, we run [scripts/run-nightly-benchmarks.sh](scripts/run-nightly-benchmarks.sh), which will probe the serving engine of the current container.
+- The `scripts/run-nightly-benchmarks.sh` will parse the workload described in [nightly-tests.json](tests/nightly-tests.json) and launch the right benchmark for the specified serving engine via `scripts/launch-server.sh`.
+- At last, we run [scripts/summary-nightly-results.py](scripts/summary-nightly-results.py) to collect and plot the final benchmarking results, and update the results to buildkite.
 
 ### Nightly tests
 
@@ -138,6 +172,6 @@ In [nightly-tests.json](tests/nightly-tests.json), we include the command line a
 
 The docker containers for benchmarking are specified in `nightly-pipeline.yaml`.
 
-WARNING: the docker versions are HARD-CODED and SHOULD BE ALIGNED WITH `nightly-descriptions.md`. The docker versions need to be hard-coded as there are several version-specific bug fixes inside `tests/run-[llm serving engine name]-nightly.sh`.
+WARNING: the docker versions are HARD-CODED and SHOULD BE ALIGNED WITH `nightly-descriptions.md`. The docker versions need to be hard-coded as there are several version-specific bug fixes inside `scripts/run-nightly-benchmarks.sh` and `scripts/launch-server.sh`.
 
 WARNING: populating `trt-llm` to latest version is not easy, as it requires updating several protobuf files in [tensorrt-demo](https://github.com/neuralmagic/tensorrt-demo.git).

.buildkite/nightly-benchmarks/nightly-annotation.md

@@ -1,3 +1,4 @@
+# Nightly benchmark annotation
 
 ## Description
 
@@ -13,15 +14,15 @@ Please download the visualization scripts in the post
 
 - Find the docker we use in `benchmarking pipeline`
 - Deploy the docker, and inside the docker:
-  - Download `nightly-benchmarks.zip`.
-  - In the same folder, run the following code:
+    - Download `nightly-benchmarks.zip`.
+    - In the same folder, run the following code:
 
-  ```console
-  export HF_TOKEN=<your HF token>
-  apt update
-  apt install -y git
-  unzip nightly-benchmarks.zip
-  VLLM_SOURCE_CODE_LOC=./ bash .buildkite/nightly-benchmarks/scripts/run-nightly-benchmarks.sh
-  ```
+    ```bash
+    export HF_TOKEN=<your HF token>
+    apt update
+    apt install -y git
+    unzip nightly-benchmarks.zip
+    VLLM_SOURCE_CODE_LOC=./ bash .buildkite/nightly-benchmarks/scripts/run-nightly-benchmarks.sh
+    ```
 
 And the results will be inside `./benchmarks/results`.

.buildkite/nightly-benchmarks/nightly-descriptions.md

@@ -13,25 +13,25 @@ Latest reproduction guilde: [github issue link](https://github.com/vllm-project/
 ## Setup
 
 - Docker images:
-  - vLLM: `vllm/vllm-openai:v0.6.2`
-  - SGLang: `lmsysorg/sglang:v0.3.2-cu121`
-  - LMDeploy: `openmmlab/lmdeploy:v0.6.1-cu12`
-  - TensorRT-LLM: `nvcr.io/nvidia/tritonserver:24.07-trtllm-python-py3`
-    - *NOTE: we uses r24.07 as the current implementation only works for this version. We are going to bump this up.*
-  - Check [nightly-pipeline.yaml](nightly-pipeline.yaml) for the concrete docker images, specs and commands we use for the benchmark.
+    - vLLM: `vllm/vllm-openai:v0.6.2`
+    - SGLang: `lmsysorg/sglang:v0.3.2-cu121`
+    - LMDeploy: `openmmlab/lmdeploy:v0.6.1-cu12`
+    - TensorRT-LLM: `nvcr.io/nvidia/tritonserver:24.07-trtllm-python-py3`
+        - *NOTE: we uses r24.07 as the current implementation only works for this version. We are going to bump this up.*
+    - Check [nightly-pipeline.yaml](nightly-pipeline.yaml) for the concrete docker images, specs and commands we use for the benchmark.
 - Hardware
-  - 8x Nvidia A100 GPUs
+    - 8x Nvidia A100 GPUs
 - Workload:
-  - Dataset
-    - ShareGPT dataset
-    - Prefill-heavy dataset (in average 462 input tokens, 16 tokens as output)
-    - Decode-heavy dataset (in average 462 input tokens, 256 output tokens)
-    - Check [nightly-tests.json](tests/nightly-tests.json) for the concrete configuration of datasets we use.
-  - Models: llama-3 8B, llama-3 70B.
-    - We do not use llama 3.1 as it is incompatible with trt-llm r24.07. ([issue](https://github.com/NVIDIA/TensorRT-LLM/issues/2105)).
-  - Average QPS (query per second): 2, 4, 8, 16, 32 and inf.
-    - Queries are randomly sampled, and arrival patterns are determined via Poisson process, but all with fixed random seed.
-  - Evaluation metrics: Throughput (higher the better), TTFT (time to the first token, lower the better), ITL (inter-token latency, lower the better).
+    - Dataset
+        - ShareGPT dataset
+        - Prefill-heavy dataset (in average 462 input tokens, 16 tokens as output)
+        - Decode-heavy dataset (in average 462 input tokens, 256 output tokens)
+        - Check [nightly-tests.json](tests/nightly-tests.json) for the concrete configuration of datasets we use.
+    - Models: llama-3 8B, llama-3 70B.
+        - We do not use llama 3.1 as it is incompatible with trt-llm r24.07. ([issue](https://github.com/NVIDIA/TensorRT-LLM/issues/2105)).
+    - Average QPS (query per second): 2, 4, 8, 16, 32 and inf.
+        - Queries are randomly sampled, and arrival patterns are determined via Poisson process, but all with fixed random seed.
+    - Evaluation metrics: Throughput (higher the better), TTFT (time to the first token, lower the better), ITL (inter-token latency, lower the better).
 
 ## Known issues
 

.buildkite/nightly-benchmarks/performance-benchmarks-descriptions.md

@@ -1,10 +1,12 @@
+# Performance benchmarks descriptions
 
 ## Latency tests
 
 - Input length: 32 tokens.
 - Output length: 128 tokens.
 - Batch size: fixed (8).
-- Models: llama-3.1 8B, llama-3 70B, mixtral 8x7B.
+- GPU Models: llama-3.1 8B, llama-3 70B, mixtral 8x7B.
+- CPU Models: llama-3.1 8B.
 - Evaluation metrics: end-to-end latency (mean, median, p99).
 
 {latency_tests_markdown_table}
@@ -14,7 +16,8 @@
 - Input length: randomly sample 200 prompts from ShareGPT dataset (with fixed random seed).
 - Output length: the corresponding output length of these 200 prompts.
 - Batch size: dynamically determined by vllm to achieve maximum throughput.
-- Models: llama-3.1 8B, llama-3 70B, mixtral 8x7B.
+- GPU Models: llama-3.1 8B, llama-3 70B, mixtral 8x7B.
+- CPU Models: llama-3.1 8B.
 - Evaluation metrics: throughput.
 
 {throughput_tests_markdown_table}
@@ -25,12 +28,18 @@
 - Output length: the corresponding output length of these 200 prompts.
 - Batch size: dynamically determined by vllm and the arrival pattern of the requests.
 - **Average QPS (query per second)**: 1, 4, 16 and inf. QPS = inf means all requests come at once. For other QPS values, the arrival time of each query is determined using a random Poisson process (with fixed random seed).
-- Models: llama-3.1 8B, llama-3 70B, mixtral 8x7B.
-- We also added a speculative decoding test for llama-3 70B, under QPS 2
+- GPU Models: llama-3.1 8B, llama-3 70B, mixtral 8x7B.
+- We also added a speculative decoding test for llama-3 70B on GPU, under QPS 2
+- CPU Models: llama-3.1 8B.
 - Evaluation metrics: throughput, TTFT (time to the first token, with mean, median and p99), ITL (inter-token latency, with mean, median and p99).
+- For CPU, we added random dataset tests to benchmark fixed input/output length with 100 prompts.
 
 {serving_tests_markdown_table}
 
+## Platform Information
+
+{platform_markdown_table}
+
 ## json version of the benchmarking tables
 
 This section contains the data of the markdown tables above in JSON format.

.buildkite/nightly-benchmarks/scripts/compare-json-results.py

@@ -0,0 +1,215 @@
+# SPDX-License-Identifier: Apache-2.0
+# SPDX-FileCopyrightText: Copyright contributors to the vLLM project
+import argparse
+import json
+import os
+
+import pandas as pd
+
+
+def compare_data_columns(
+    files, name_column, data_column, info_cols, drop_column, debug=False
+):
+    print("\ncompare_data_column: " + data_column)
+    frames = []
+    raw_data_cols = []
+    compare_frames = []
+    for file in files:
+        data_df = pd.read_json(file)
+        serving_df = data_df.dropna(subset=[drop_column], ignore_index=True)
+        # Show all info columns in the first couple columns
+        if not frames:
+            for col in info_cols:
+                if col not in serving_df.columns:
+                    print(f"Skipping missing column: {col}")
+                    continue
+                frames.append(serving_df[col])
+        # only show test name under debug mode
+        if debug is True:
+            serving_df = serving_df.rename(columns={name_column: file + "_name"})
+            frames.append(serving_df[file + "_name"])
+
+        file = "/".join(file.split("/")[:-1])
+        serving_df = serving_df.rename(columns={data_column: file})
+        frames.append(serving_df[file])
+        raw_data_cols.append(file)
+        compare_frames.append(serving_df[file])
+        if len(compare_frames) >= 2:
+            # Compare numbers among two files
+            ratio_df = compare_frames[1] / compare_frames[0]
+            frames.append(ratio_df)
+            compare_frames.pop(1)
+
+    concat_df = pd.concat(frames, axis=1)
+    print(raw_data_cols)
+    return concat_df, raw_data_cols
+
+
+def split_json_by_tp_pp(
+    input_file: str = "benchmark_results.json", output_root: str = "."
+) -> list[str]:
+    """
+    Split a benchmark JSON into separate folders by (TP Size, PP Size).
+
+    Creates: <output_root>/tp{TP}_pp{PP}/benchmark_results.json
+    Returns: list of file paths written.
+    """
+    # Load JSON data into DataFrame
+    with open(input_file, encoding="utf-8") as f:
+        data = json.load(f)
+
+    # If the JSON is a dict with a list under common keys, use that list
+    if isinstance(data, dict):
+        for key in ("results", "serving_results", "benchmarks", "data"):
+            if isinstance(data.get(key), list):
+                data = data[key]
+                break
+
+    df = pd.DataFrame(data)
+
+    # Handle alias column names
+    rename_map = {
+        "tp_size": "TP Size",
+        "tensor_parallel_size": "TP Size",
+        "pp_size": "PP Size",
+        "pipeline_parallel_size": "PP Size",
+    }
+    df.rename(
+        columns={k: v for k, v in rename_map.items() if k in df.columns}, inplace=True
+    )
+
+    # Ensure TP/PP columns exist (default to 1 if missing)
+    if "TP Size" not in df.columns:
+        df["TP Size"] = 1
+    if "PP Size" not in df.columns:
+        df["PP Size"] = 1
+
+    # make sure TP/PP are numeric ints with no NaN
+    df["TP Size"] = (
+        pd.to_numeric(df.get("TP Size", 1), errors="coerce").fillna(1).astype(int)
+    )
+    df["PP Size"] = (
+        pd.to_numeric(df.get("PP Size", 1), errors="coerce").fillna(1).astype(int)
+    )
+
+    # Split into separate folders
+    saved_paths: list[str] = []
+    for (tp, pp), group_df in df.groupby(["TP Size", "PP Size"], dropna=False):
+        folder_name = os.path.join(output_root, f"tp{int(tp)}_pp{int(pp)}")
+        os.makedirs(folder_name, exist_ok=True)
+        filepath = os.path.join(folder_name, "benchmark_results.json")
+        group_df.to_json(filepath, orient="records", indent=2, force_ascii=False)
+        print(f"Saved: {filepath}")
+        saved_paths.append(filepath)
+
+    return saved_paths
+
+
+if __name__ == "__main__":
+    parser = argparse.ArgumentParser()
+    parser.add_argument(
+        "-f", "--file", action="append", type=str, help="input file name"
+    )
+    parser.add_argument(
+        "--debug", action="store_true", help="show all information for debugging"
+    )
+    parser.add_argument(
+        "--plot",
+        action=argparse.BooleanOptionalAction,
+        default=True,
+        help="plot perf diagrams or not --no-plot --plot",
+    )
+    parser.add_argument(
+        "-x",
+        "--xaxis",
+        type=str,
+        default="# of max concurrency.",
+        help="column name to use as X Axis in comparision graph",
+    )
+    args = parser.parse_args()
+
+    drop_column = "P99"
+    name_column = "Test name"
+    info_cols = [
+        "Model",
+        "Dataset Name",
+        "Input Len",
+        "Output Len",
+        "TP Size",
+        "PP Size",
+        "# of max concurrency.",
+        "qps",
+    ]
+    data_cols_to_compare = ["Output Tput (tok/s)", "Median TTFT (ms)", "Median"]
+    html_msgs_for_data_cols = [
+        "Compare Output Tokens /n",
+        "Median TTFT /n",
+        "Median TPOT /n",
+    ]
+
+    if len(args.file) == 1:
+        files = split_json_by_tp_pp(args.file[0], output_root="splits")
+        info_cols = [c for c in info_cols if c not in ("TP Size", "PP Size")]
+    else:
+        files = args.file
+    print("comparing : " + ", ".join(files))
+    debug = args.debug
+    plot = args.plot
+    # For Plot feature, assign y axis from one of info_cols
+    y_axis_index = info_cols.index(args.xaxis) if args.xaxis in info_cols else 6
+    with open("perf_comparison.html", "w") as text_file:
+        for i in range(len(data_cols_to_compare)):
+            output_df, raw_data_cols = compare_data_columns(
+                files,
+                name_column,
+                data_cols_to_compare[i],
+                info_cols,
+                drop_column,
+                debug=debug,
+            )
+
+            # For Plot feature, insert y axis from one of info_cols
+            raw_data_cols.insert(0, info_cols[y_axis_index])
+
+            filtered_info_cols = info_cols[:-2]
+            existing_group_cols = [
+                c for c in filtered_info_cols if c in output_df.columns
+            ]
+            if not existing_group_cols:
+                raise ValueError(
+                    f"No valid group-by columns  "
+                    f"Expected subset: {filtered_info_cols}, "
+                    f"but DataFrame has: {list(output_df.columns)}"
+                )
+
+            output_df_sorted = output_df.sort_values(by=existing_group_cols)
+            output_groups = output_df_sorted.groupby(existing_group_cols, dropna=False)
+            for name, group in output_groups:
+                html = group.to_html()
+                text_file.write(html_msgs_for_data_cols[i])
+                text_file.write(html)
+
+                if plot is True:
+                    import pandas as pd
+                    import plotly.express as px
+
+                    df = group[raw_data_cols]
+                    df_sorted = df.sort_values(by=info_cols[y_axis_index])
+                    # Melt DataFrame for plotting
+                    df_melted = df_sorted.melt(
+                        id_vars=info_cols[y_axis_index],
+                        var_name="Configuration",
+                        value_name=data_cols_to_compare[i],
+                    )
+                    title = data_cols_to_compare[i] + " vs " + info_cols[y_axis_index]
+                    # Create Plotly line chart
+                    fig = px.line(
+                        df_melted,
+                        x=info_cols[y_axis_index],
+                        y=data_cols_to_compare[i],
+                        color="Configuration",
+                        title=title,
+                        markers=True,
+                    )
+                    # Export to HTML
+                    text_file.write(fig.to_html(full_html=True, include_plotlyjs="cdn"))

.buildkite/nightly-benchmarks/scripts/convert-results-json-to-markdown.py

@@ -1,15 +1,19 @@
 # SPDX-License-Identifier: Apache-2.0
 # SPDX-FileCopyrightText: Copyright contributors to the vLLM project
 
+import argparse
 import json
 import os
+import shlex
+from importlib import util
 from pathlib import Path
+from typing import Any
 
 import pandas as pd
+import psutil
+import regex as re
 from tabulate import tabulate
 
-results_folder = Path("results/")
-
 # latency results and the keys that will be printed into markdown
 latency_results = []
 latency_column_mapping = {
@@ -29,28 +33,39 @@ throughput_results = []
 throughput_results_column_mapping = {
     "test_name": "Test name",
     "gpu_type": "GPU",
-    # "num_requests": "# of req.",
-    # "total_num_tokens": "Total # of tokens",
-    # "elapsed_time": "Elapsed time (s)",
+    "num_requests": "# of req.",
+    "total_num_tokens": "Total # of tokens",
+    "elapsed_time": "Elapsed time (s)",
     "requests_per_second": "Tput (req/s)",
-    # "tokens_per_second": "Tput (tok/s)",
+    "tokens_per_second": "Tput (tok/s)",
 }
 
 # serving results and the keys that will be printed into markdown
 serving_results = []
 serving_column_mapping = {
     "test_name": "Test name",
+    "model_id": "Model",
+    "dataset_name": "Dataset Name",
+    "input_len": "Input Len",
+    "output_len": "Output Len",
+    "tp_size": "TP Size",
+    "pp_size": "PP Size",
+    "dtype": "dtype",
     "gpu_type": "GPU",
-    # "completed": "# of req.",
+    "completed": "# of req.",
+    "qps": "qps",
+    "max_concurrency": "# of max concurrency.",
     "request_throughput": "Tput (req/s)",
-    # "input_throughput": "Input Tput (tok/s)",
-    # "output_throughput": "Output Tput (tok/s)",
+    "total_token_throughput": "Total Token Tput (tok/s)",
+    "output_throughput": "Output Tput (tok/s)",
+    # "total_input_tokens": "Total input tokens",
+    # "total_output_tokens": "Total output tokens",
     "mean_ttft_ms": "Mean TTFT (ms)",
     "median_ttft_ms": "Median TTFT (ms)",
     "p99_ttft_ms": "P99 TTFT (ms)",
-    # "mean_tpot_ms": "Mean TPOT (ms)",
-    # "median_tpot_ms": "Median",
-    # "p99_tpot_ms": "P99",
+    "mean_tpot_ms": "Mean TPOT (ms)",
+    "median_tpot_ms": "Median",
+    "p99_tpot_ms": "P99",
     "mean_itl_ms": "Mean ITL (ms)",
     "median_itl_ms": "Median ITL (ms)",
     "p99_itl_ms": "P99 ITL (ms)",
@@ -75,15 +90,125 @@ def results_to_json(latency, throughput, serving):
     )
 
 
+def get_size_with_unit(bytes, suffix="B"):
+    """
+    Scale bytes to its proper format
+    e.g:
+        1253656 => '1.20MB'
+        1253656678 => '1.17GB'
+    """
+    factor = 1024
+    for unit in ["", "K", "M", "G", "T", "P"]:
+        if bytes < factor:
+            return f"{bytes:.2f}{unit}{suffix}"
+        bytes /= factor
+
+
+def _coerce(val: str) -> Any:
+    """Best-effort type coercion from string to Python types."""
+    low = val.lower()
+    if low == "null":
+        return None
+    if low == "true":
+        return True
+    if low == "false":
+        return False
+    # integers
+    if re.fullmatch(r"[+-]?\d+", val):
+        try:
+            return int(val)
+        except ValueError:
+            pass
+    # floats (keep 'inf'/'-inf'/'nan' as strings)
+    if re.fullmatch(r"[+-]?\d*\.\d+", val):
+        try:
+            return float(val)
+        except ValueError:
+            pass
+    return val
+
+
+def parse_client_command(cmd: str) -> dict[str, Any]:
+    """Parse the client_command shell string into {executable, script, args}."""
+    toks = shlex.split(cmd)
+    if len(toks) < 2:
+        raise ValueError("client_command must include an executable and a script")
+    executable, script = toks[0], toks[1]
+    args: dict[str, Any] = {}
+
+    i = 2
+    while i < len(toks):
+        t = toks[i]
+        if t.startswith("--"):
+            # --key=value or --key (value) or boolean flag
+            if "=" in t:
+                key, val = t.split("=", 1)
+                if key == "--metadata":
+                    md = {}
+                    if val:
+                        if "=" in val:
+                            k, v = val.split("=", 1)
+                            md[k] = _coerce(v)
+                        else:
+                            md[val] = True
+                    args[key] = md
+                else:
+                    args[key] = _coerce(val)
+                i += 1
+                continue
+
+            key = t
+
+            # Special: consume metadata k=v pairs until next --flag
+            if key == "--metadata":
+                i += 1
+                md = {}
+                while i < len(toks) and not toks[i].startswith("--"):
+                    pair = toks[i]
+                    if "=" in pair:
+                        k, v = pair.split("=", 1)
+                        md[k] = _coerce(v)
+                    else:
+                        md[pair] = True
+                    i += 1
+                args[key] = md
+                continue
+
+            # Standard: check if next token is a value (not a flag)
+            if i + 1 < len(toks) and not toks[i + 1].startswith("--"):
+                args[key] = _coerce(toks[i + 1])
+                i += 2
+            else:
+                # lone flag -> True
+                args[key] = True
+                i += 1
+        else:
+            # unexpected positional; skip
+            i += 1
+
+    return {"executable": executable, "script": script, "args": args}
+
+
 if __name__ == "__main__":
+    parser = argparse.ArgumentParser()
+    parser.add_argument(
+        "-r",
+        "--result",
+        type=str,
+        default="results",
+        help="Folder name for benchmark output results.",
+    )
+    args = parser.parse_args()
+    results_folder = Path(args.result)
+    if not results_folder.exists():
+        raise FileNotFoundError(f"results folder does not exist: {results_folder}")
     # collect results
     for test_file in results_folder.glob("*.json"):
         with open(test_file) as f:
             raw_result = json.loads(f.read())
 
         if "serving" in str(test_file):
-            # this result is generated via `benchmark_serving.py`
-
+            # this result is generated via `vllm bench serve` command
             # attach the benchmarking command to raw_result
             try:
                 with open(test_file.with_suffix(".commands")) as f:
@@ -91,18 +216,50 @@ if __name__ == "__main__":
             except OSError as e:
                 print(e)
                 continue
+            # Parse Server Command Arg
+            out: dict[str, Any] = {
+                "server_command": parse_client_command(command["server_command"])
+            }
+            parse_args = [
+                "--tensor-parallel-size",
+                "--pipeline-parallel-size",
+                "--dtype",
+            ]
+            col_mapping = ["tp_size", "pp_size", "dtype"]
+            for index, arg in enumerate(parse_args):
+                if arg in out["server_command"]["args"]:
+                    raw_result.update(
+                        {col_mapping[index]: out["server_command"]["args"][arg]}
+                    )
 
+            # Parse Client Command Arg
+            out: dict[str, Any] = {
+                "client_command": parse_client_command(command["client_command"])
+            }
+            parse_args = [
+                "--dataset-name",
+                "--random-input-len",
+                "--random-output-len",
+                "--request-rate",
+            ]
+            col_mapping = ["dataset_name", "input_len", "output_len", "qps"]
+
+            for index, arg in enumerate(parse_args):
+                if arg in out["client_command"]["args"]:
+                    raw_result.update(
+                        {col_mapping[index]: out["client_command"]["args"][arg]}
+                    )
+            # Add Server, Client command
             raw_result.update(command)
 
             # update the test name of this result
             raw_result.update({"test_name": test_file.stem})
-
             # add the result to raw_result
             serving_results.append(raw_result)
             continue
 
         elif "latency" in f.name:
-            # this result is generated via `benchmark_latency.py`
+            # this result is generated via `vllm bench latency` command
 
             # attach the benchmarking command to raw_result
             try:
@@ -130,7 +287,7 @@ if __name__ == "__main__":
             continue
 
         elif "throughput" in f.name:
-            # this result is generated via `benchmark_throughput.py`
+            # this result is generated via `vllm bench throughput` command
 
             # attach the benchmarking command to raw_result
             try:
@@ -155,6 +312,27 @@ if __name__ == "__main__":
     serving_results = pd.DataFrame.from_dict(serving_results)
     throughput_results = pd.DataFrame.from_dict(throughput_results)
 
+    svmem = psutil.virtual_memory()
+    platform_data = {
+        "Physical cores": [psutil.cpu_count(logical=False)],
+        "Total cores": [psutil.cpu_count(logical=True)],
+        "Total Memory": [get_size_with_unit(svmem.total)],
+    }
+
+    if util.find_spec("numa") is not None:
+        from numa import info
+
+        platform_data["Total NUMA nodes"] = [info.get_num_configured_nodes()]
+
+    if util.find_spec("cpuinfo") is not None:
+        from cpuinfo import get_cpu_info
+
+        platform_data["CPU Brand"] = [get_cpu_info()["brand_raw"]]
+
+    platform_results = pd.DataFrame.from_dict(
+        platform_data, orient="index", columns=["Platform Info"]
+    )
+
     raw_results_json = results_to_json(
         latency_results, throughput_results, serving_results
     )
@@ -165,7 +343,10 @@ if __name__ == "__main__":
             columns=latency_column_mapping
         )
     if not serving_results.empty:
-        serving_results = serving_results[list(serving_column_mapping.keys())].rename(
+        valid_columns = [
+            col for col in serving_column_mapping if col in serving_results.columns
+        ]
+        serving_results = serving_results[valid_columns].rename(
             columns=serving_column_mapping
         )
     if not throughput_results.empty:
@@ -200,9 +381,14 @@ if __name__ == "__main__":
     throughput_md_table = tabulate(
         throughput_results, headers="keys", tablefmt="pipe", showindex=False
     )
+    platform_md_table = tabulate(
+        platform_results, headers="keys", tablefmt="pipe", showindex=True
+    )
 
     # document the result
-    with open(results_folder / "benchmark_results.md", "w") as f:
+    md_file = "benchmark_results.md"
+    json_file = "benchmark_results.json"
+    with open(results_folder / md_file, "w") as f:
         results = read_markdown(
             "../.buildkite/nightly-benchmarks/"
             + "performance-benchmarks-descriptions.md"
@@ -211,12 +397,13 @@ if __name__ == "__main__":
             latency_tests_markdown_table=latency_md_table,
             throughput_tests_markdown_table=throughput_md_table,
             serving_tests_markdown_table=serving_md_table,
+            platform_markdown_table=platform_md_table,
             benchmarking_results_in_json_string=processed_results_json,
         )
         f.write(results)
 
     # document benchmarking results in json
-    with open(results_folder / "benchmark_results.json", "w") as f:
+    with open(results_folder / json_file, "w") as f:
         results = (
             latency_results.to_dict(orient="records")
             + throughput_results.to_dict(orient="records")

.buildkite/nightly-benchmarks/scripts/run-nightly-benchmarks.sh

@@ -73,7 +73,7 @@ get_current_llm_serving_engine() {
     echo "Container: vllm"
     # move to a completely irrelevant directory, to avoid import vllm from current folder
     export CURRENT_LLM_SERVING_ENGINE=vllm
-    
+
     return
   fi
 }
@@ -95,12 +95,14 @@ json2args() {
 }
 
 kill_gpu_processes() {
-  pkill -f python
-  pkill -f python3
-  pkill -f tritonserver
-  pkill -f pt_main_thread
-  pkill -f text-generation
-  pkill -f lmdeploy
+  pkill -f '[p]ython'
+  pkill -f '[p]ython3'
+  pkill -f '[t]ritonserver'
+  pkill -f '[p]t_main_thread'
+  pkill -f '[t]ext-generation'
+  pkill -f '[l]mdeploy'
+  # vLLM now names the process with VLLM prefix after https://github.com/vllm-project/vllm/pull/21445
+  pkill -f '[V]LLM'
 
   while [ "$(nvidia-smi --query-gpu=memory.used --format=csv,noheader,nounits | head -n 1)" -ge 1000 ]; do
     sleep 1
@@ -125,7 +127,7 @@ ensure_installed() {
 }
 
 run_serving_tests() {
-  # run serving tests using `benchmark_serving.py`
+  # run serving tests using `vllm bench serve` command
   # $1: a json file specifying serving test cases
 
   local serving_test_file
@@ -225,7 +227,7 @@ run_serving_tests() {
 
       if [[ "$dataset_name" = "sharegpt" ]]; then
 
-        client_command="python3 benchmark_serving.py \
+        client_command="vllm bench serve \
           --backend $backend \
           --tokenizer /tokenizer_cache \
           --model $model \
@@ -246,7 +248,7 @@ run_serving_tests() {
         sonnet_output_len=$(echo "$common_params" | jq -r '.sonnet_output_len')
         sonnet_prefix_len=$(echo "$common_params" | jq -r '.sonnet_prefix_len')
 
-        client_command="python3 benchmark_serving.py \
+        client_command="vllm bench serve \
           --backend $backend \
           --tokenizer /tokenizer_cache \
           --model $model \
@@ -265,13 +267,13 @@ run_serving_tests() {
           $client_args"
 
       else
-  
+
         echo "The dataset name must be either 'sharegpt' or 'sonnet'. Got $dataset_name."
         exit 1
 
       fi
 
-        
+
 
       echo "Running test case $test_name with qps $qps"
       echo "Client command: $client_command"
@@ -302,7 +304,7 @@ run_serving_tests() {
 }
 
 run_genai_perf_tests() {
-  # run genai-perf tests 
+  # run genai-perf tests
 
   # $1: a json file specifying genai-perf test cases
   local genai_perf_test_file
@@ -311,14 +313,14 @@ run_genai_perf_tests() {
   # Iterate over genai-perf tests
   jq -c '.[]' "$genai_perf_test_file" | while read -r params; do
     # get the test name, and append the GPU type back to it.
-    test_name=$(echo "$params" | jq -r '.test_name')    
-    
+    test_name=$(echo "$params" | jq -r '.test_name')
+
     # if TEST_SELECTOR is set, only run the test cases that match the selector
     if [[ -n "$TEST_SELECTOR" ]] && [[ ! "$test_name" =~ $TEST_SELECTOR ]]; then
       echo "Skip test case $test_name."
       continue
     fi
-    
+
     # prepend the current serving engine to the test name
     test_name=${CURRENT_LLM_SERVING_ENGINE}_${test_name}
 
@@ -369,10 +371,10 @@ run_genai_perf_tests() {
         qps=$num_prompts
         echo "now qps is $qps"
       fi
-    
+
       new_test_name=$test_name"_qps_"$qps
       backend=$CURRENT_LLM_SERVING_ENGINE
-      
+
       if [[ "$backend" == *"vllm"* ]]; then
         backend="vllm"
       fi
@@ -413,7 +415,7 @@ prepare_dataset() {
   do
     cat sonnet.txt >> sonnet_4x.txt
   done
-  
+
 }
 
 main() {

.buildkite/nightly-benchmarks/scripts/run-performance-benchmarks.sh

@@ -31,6 +31,20 @@ check_gpus() {
   echo "GPU type is $gpu_type"
 }
 
+check_cpus() {
+  # check the number of CPUs and NUMA Node and GPU type.
+  declare -g numa_count=$(lscpu | grep "NUMA node(s):" | awk '{print $3}')
+  if [[ $numa_count -gt 0 ]]; then
+    echo "NUMA found."
+    echo $numa_count
+  else
+    echo "Need at least 1 NUMA to run benchmarking."
+    exit 1
+  fi
+  declare -g gpu_type="cpu"
+  echo "GPU type is $gpu_type"
+}
+
 check_hf_token() {
   # check if HF_TOKEN is available and valid
   if [[ -z "$HF_TOKEN" ]]; then
@@ -69,6 +83,22 @@ json2args() {
   echo "$args"
 }
 
+json2envs() {
+  # transforms the JSON string to environment variables.
+  # example:
+  # input: { "VLLM_CPU_KVCACHE_SPACE": 5 }
+  # output: VLLM_CPU_KVCACHE_SPACE=5
+  local json_string=$1
+  local args=$(
+    echo "$json_string" | jq -r '
+      to_entries |
+      map((.key ) + "=" + (.value | tostring)) |
+      join(" ")
+    '
+  )
+  echo "$args"
+}
+
 wait_for_server() {
   # wait for vllm server to start
   # return 1 if vllm server crashes
@@ -96,7 +126,8 @@ kill_gpu_processes() {
   ps -aux
   lsof -t -i:8000 | xargs -r kill -9
   pgrep python3 | xargs -r kill -9
-
+  # vLLM now names the process with VLLM prefix after https://github.com/vllm-project/vllm/pull/21445
+  pgrep VLLM | xargs -r kill -9
 
   # wait until GPU memory usage smaller than 1GB
   if command -v nvidia-smi; then
@@ -134,7 +165,7 @@ upload_to_buildkite() {
 }
 
 run_latency_tests() {
-  # run latency tests using `benchmark_latency.py`
+  # run latency tests using `vllm bench latency` command
   # $1: a json file specifying latency test cases
 
   local latency_test_file
@@ -158,15 +189,26 @@ run_latency_tests() {
     # get arguments
     latency_params=$(echo "$params" | jq -r '.parameters')
     latency_args=$(json2args "$latency_params")
+    latency_environment_variables=$(echo "$params" | jq -r '.environment_variables')
+    latency_envs=$(json2envs "$latency_environment_variables")
 
     # check if there is enough GPU to run the test
     tp=$(echo "$latency_params" | jq -r '.tensor_parallel_size')
-    if [[ $gpu_count -lt $tp ]]; then
-      echo "Required tensor-parallel-size $tp but only $gpu_count GPU found. Skip testcase $test_name."
-      continue
+    if [ "$ON_CPU" == "1" ]; then
+      pp=$(echo "$latency_params" | jq -r '.pipeline_parallel_size')
+      world_size=$(($tp*$pp))
+      if [[ $numa_count -lt $world_size  && -z "${REMOTE_HOST}" ]]; then
+        echo "Required world-size $world_size but only $numa_count NUMA nodes found. Skip testcase $test_name."
+        continue
+      fi
+    else
+      if [[ $gpu_count -lt $tp ]]; then
+        echo "Required tensor-parallel-size $tp but only $gpu_count GPU found. Skip testcase $test_name."
+        continue
+      fi
     fi
 
-    latency_command="python3 benchmark_latency.py \
+    latency_command=" $latency_envs vllm bench latency \
       --output-json $RESULTS_FOLDER/${test_name}.json \
       $latency_args"
 
@@ -192,7 +234,7 @@ run_latency_tests() {
 }
 
 run_throughput_tests() {
-  # run throughput tests using `benchmark_throughput.py`
+  # run throughput tests using `vllm bench throughput`
   # $1: a json file specifying throughput test cases
 
   local throughput_test_file
@@ -216,15 +258,26 @@ run_throughput_tests() {
     # get arguments
     throughput_params=$(echo "$params" | jq -r '.parameters')
     throughput_args=$(json2args "$throughput_params")
+    throughput_environment_variables=$(echo "$params" | jq -r '.environment_variables')
+    throughput_envs=$(json2envs "$throughput_environment_variables")
 
     # check if there is enough GPU to run the test
     tp=$(echo "$throughput_params" | jq -r '.tensor_parallel_size')
-    if [[ $gpu_count -lt $tp ]]; then
-      echo "Required tensor-parallel-size $tp but only $gpu_count GPU found. Skip testcase $test_name."
-      continue
+    if [ "$ON_CPU" == "1" ]; then
+      pp=$(echo "$throughput_params" | jq -r '.pipeline_parallel_size')
+      world_size=$(($tp*$pp))
+      if [[ $numa_count -lt $world_size  && -z "${REMOTE_HOST}" ]]; then
+        echo "Required world-size $world_size but only $numa_count NUMA nodes found. Skip testcase $test_name."
+        continue
+      fi
+    else
+      if [[ $gpu_count -lt $tp ]]; then
+        echo "Required tensor-parallel-size $tp but only $gpu_count GPU found. Skip testcase $test_name."
+        continue
+      fi
     fi
 
-    throughput_command="python3 benchmark_throughput.py \
+    throughput_command=" $throughput_envs vllm bench throughput \
       --output-json $RESULTS_FOLDER/${test_name}.json \
       $throughput_args"
 
@@ -249,7 +302,7 @@ run_throughput_tests() {
 }
 
 run_serving_tests() {
-  # run serving tests using `benchmark_serving.py`
+  # run serving tests using `vllm bench serve` command
   # $1: a json file specifying serving test cases
 
   local serving_test_file
@@ -272,18 +325,36 @@ run_serving_tests() {
 
     # get client and server arguments
     server_params=$(echo "$params" | jq -r '.server_parameters')
+    server_envs=$(echo "$params" | jq -r '.server_environment_variables')
     client_params=$(echo "$params" | jq -r '.client_parameters')
     server_args=$(json2args "$server_params")
+    server_envs=$(json2envs "$server_envs")
     client_args=$(json2args "$client_params")
     qps_list=$(echo "$params" | jq -r '.qps_list')
     qps_list=$(echo "$qps_list" | jq -r '.[] | @sh')
     echo "Running over qps list $qps_list"
+    max_concurrency_list=$(echo "$params" | jq -r '.max_concurrency_list')
+    if [[ -z "$max_concurrency_list" || "$max_concurrency_list" == "null" ]]; then
+        num_prompts=$(echo "$client_params" | jq -r '.num_prompts')
+        max_concurrency_list="[$num_prompts]"
+    fi
+    max_concurrency_list=$(echo "$max_concurrency_list" | jq -r '.[] | @sh')
+    echo "Running over max concurrency list $max_concurrency_list"
 
-    # check if there is enough GPU to run the test
+    # check if there is enough resources to run the test
     tp=$(echo "$server_params" | jq -r '.tensor_parallel_size')
-    if [[ $gpu_count -lt $tp ]]; then
-      echo "Required tensor-parallel-size $tp but only $gpu_count GPU found. Skip testcase $test_name."
-      continue
+    if [ "$ON_CPU" == "1" ]; then
+      pp=$(echo "$server_params" | jq -r '.pipeline_parallel_size')
+      world_size=$(($tp*$pp))
+      if [[ $numa_count -lt $world_size  && -z "${REMOTE_HOST}" ]]; then
+        echo "Required world-size $world_size but only $numa_count NUMA nodes found. Skip testcase $test_name."
+        continue
+      fi
+    else
+      if [[ $gpu_count -lt $tp ]]; then
+        echo "Required tensor-parallel-size $tp but only $gpu_count GPU found. Skip testcase $test_name."
+        continue
+      fi
     fi
 
     # check if server model and client model is aligned
@@ -294,23 +365,33 @@ run_serving_tests() {
       continue
     fi
 
-    server_command="python3 \
+    server_command="$server_envs python3 \
       -m vllm.entrypoints.openai.api_server \
       $server_args"
 
     # run the server
     echo "Running test case $test_name"
     echo "Server command: $server_command"
-    bash -c "$server_command" &
-    server_pid=$!
-
-    # wait until the server is alive
-    if wait_for_server; then
-      echo ""
-      echo "vllm server is up and running."
+    # support remote vllm server
+    client_remote_args=""
+    if [[ -z "${REMOTE_HOST}" ]]; then
+      bash -c "$server_command" &
+      server_pid=$!
+      # wait until the server is alive
+      if wait_for_server; then
+        echo ""
+        echo "vLLM server is up and running."
+      else
+        echo ""
+        echo "vLLM failed to start within the timeout period."
+      fi
     else
-      echo ""
-      echo "vllm failed to start within the timeout period."
+      server_command="Using Remote Server $REMOTE_HOST $REMOTE_PORT"
+      if [[ ${REMOTE_PORT} ]]; then
+        client_remote_args=" --host=$REMOTE_HOST --port=$REMOTE_PORT "
+      else
+        client_remote_args=" --host=$REMOTE_HOST "
+      fi
     fi
 
     # iterate over different QPS
@@ -322,35 +403,39 @@ run_serving_tests() {
         echo "now qps is $qps"
       fi
 
-      new_test_name=$test_name"_qps_"$qps
+      # iterate over different max_concurrency
+      for max_concurrency in $max_concurrency_list; do
+        new_test_name=$test_name"_qps_"$qps"_concurrency_"$max_concurrency
+        echo " new test name $new_test_name"
+        # pass the tensor parallel size to the client so that it can be displayed
+        # on the benchmark dashboard
+        client_command="vllm bench serve \
+          --save-result \
+          --result-dir $RESULTS_FOLDER \
+          --result-filename ${new_test_name}.json \
+          --request-rate $qps \
+          --max-concurrency $max_concurrency \
+          --metadata "tensor_parallel_size=$tp" \
+          $client_args $client_remote_args "
 
-      # pass the tensor parallel size to the client so that it can be displayed
-      # on the benchmark dashboard
-      client_command="python3 benchmark_serving.py \
-        --save-result \
-        --result-dir $RESULTS_FOLDER \
-        --result-filename ${new_test_name}.json \
-        --request-rate $qps \
-        --metadata "tensor_parallel_size=$tp" \
-        $client_args"
+        echo "Running test case $test_name with qps $qps"
+        echo "Client command: $client_command"
 
-      echo "Running test case $test_name with qps $qps"
-      echo "Client command: $client_command"
+        bash -c "$client_command"
 
-      bash -c "$client_command"
-
-      # record the benchmarking commands
-      jq_output=$(jq -n \
-        --arg server "$server_command" \
-        --arg client "$client_command" \
-        --arg gpu "$gpu_type" \
-        '{
-          server_command: $server,
-          client_command: $client,
-          gpu_type: $gpu
-        }')
-      echo "$jq_output" >"$RESULTS_FOLDER/${new_test_name}.commands"
+        # record the benchmarking commands
+        jq_output=$(jq -n \
+          --arg server "$server_command" \
+          --arg client "$client_command" \
+          --arg gpu "$gpu_type" \
+          '{
+            server_command: $server,
+            client_command: $client,
+            gpu_type: $gpu
+          }')
+        echo "$jq_output" >"$RESULTS_FOLDER/${new_test_name}.commands"
 
+      done
     done
 
     # clean up
@@ -360,7 +445,14 @@ run_serving_tests() {
 }
 
 main() {
-  check_gpus
+  local ARCH
+  ARCH=''
+  if [ "$ON_CPU" == "1" ];then
+     check_cpus
+     ARCH='-cpu'
+  else
+     check_gpus
+  fi
   check_hf_token
 
   # Set to v1 to run v1 benchmark
@@ -373,7 +465,7 @@ main() {
   (which jq) || (apt-get update && apt-get -y install jq)
   (which lsof) || (apt-get update && apt-get install -y lsof)
 
-  # get the current IP address, required by benchmark_serving.py
+  # get the current IP address, required by `vllm bench serve` command
   export VLLM_HOST_IP=$(hostname -I | awk '{print $1}')
   # turn of the reporting of the status of each request, to clean up the terminal output
   export VLLM_LOGGING_LEVEL="WARNING"
@@ -386,9 +478,9 @@ main() {
   QUICK_BENCHMARK_ROOT=../.buildkite/nightly-benchmarks/
 
   # benchmarking
-  run_serving_tests $QUICK_BENCHMARK_ROOT/tests/serving-tests.json
-  run_latency_tests $QUICK_BENCHMARK_ROOT/tests/latency-tests.json
-  run_throughput_tests $QUICK_BENCHMARK_ROOT/tests/throughput-tests.json
+  run_serving_tests $QUICK_BENCHMARK_ROOT/tests/"${SERVING_JSON:-serving-tests$ARCH.json}"
+  run_latency_tests $QUICK_BENCHMARK_ROOT/tests/"${LATENCY_JSON:-latency-tests$ARCH.json}"
+  run_throughput_tests $QUICK_BENCHMARK_ROOT/tests/"${THROUGHPUT_JSON:-throughput-tests$ARCH.json}"
 
   # postprocess benchmarking results
   pip install tabulate pandas

.buildkite/nightly-benchmarks/tests/genai-perf-tests.json

@@ -11,9 +11,7 @@
         },
         "vllm_server_parameters": {
             "disable_log_stats": "",
-            "disable_log_requests": "",
             "gpu_memory_utilization": 0.9,
-            "num_scheduler_steps": 10,
             "max_num_seqs": 512,
             "dtype": "bfloat16"
         },

.buildkite/nightly-benchmarks/tests/latency-tests-cpu.json

@@ -0,0 +1,30 @@
+[
+    {
+        "test_name": "latency_llama8B_tp1",
+        "environment_variables": {
+	    "VLLM_ALLOW_LONG_MAX_MODEL_LEN": 1,
+	    "VLLM_CPU_KVCACHE_SPACE": 40
+        },
+        "parameters": {
+            "model": "meta-llama/Llama-3.1-8B-Instruct",
+            "tensor_parallel_size": 1,
+            "load_format": "dummy",
+            "num_iters_warmup": 5,
+            "num_iters": 15
+        }
+    },
+    {
+        "test_name": "latency_llama8B_tp4",
+        "environment_variables": {
+	    "VLLM_ALLOW_LONG_MAX_MODEL_LEN": 1,
+	    "VLLM_CPU_KVCACHE_SPACE": 40
+        },
+        "parameters": {
+            "model": "meta-llama/Llama-3.1-8B-Instruct",
+            "tensor_parallel_size": 4,
+            "load_format": "dummy",
+            "num_iters_warmup": 5,
+            "num_iters": 15
+        }
+    }
+]

.buildkite/nightly-benchmarks/tests/nightly-tests.json

@@ -35,9 +35,7 @@
         }, 
         "vllm_server_parameters": {
             "disable_log_stats": "",
-            "disable_log_requests": "",
             "gpu_memory_utilization": 0.9,
-            "num_scheduler_steps": 10,
             "max_num_seqs": 512,
             "dtype": "bfloat16"
         },
@@ -90,9 +88,7 @@
         }, 
         "vllm_server_parameters": {
             "disable_log_stats": "",
-            "disable_log_requests": "",
             "gpu_memory_utilization": 0.9,
-            "num_scheduler_steps": 10,
             "max_num_seqs": 512,
             "dtype": "bfloat16"
         },
@@ -145,9 +141,7 @@
         }, 
         "vllm_server_parameters": {
             "disable_log_stats": "",
-            "disable_log_requests": "",
             "gpu_memory_utilization": 0.9,
-            "num_scheduler_steps": 10,
             "max_num_seqs": 512,
             "dtype": "bfloat16"
         },
@@ -197,9 +191,7 @@
         }, 
         "vllm_server_parameters": {
             "disable_log_stats": "",
-            "disable_log_requests": "",
             "gpu_memory_utilization": 0.9,
-            "num_scheduler_steps": 10,
             "max_num_seqs": 512,
             "dtype": "bfloat16"
         },
@@ -251,9 +243,7 @@
         }, 
         "vllm_server_parameters": {
             "disable_log_stats": "",
-            "disable_log_requests": "",
             "gpu_memory_utilization": 0.9,
-            "num_scheduler_steps": 10,
             "max_num_seqs": 512,
             "dtype": "bfloat16"
         },
@@ -305,9 +295,7 @@
         }, 
         "vllm_server_parameters": {
             "disable_log_stats": "",
-            "disable_log_requests": "",
             "gpu_memory_utilization": 0.9,
-            "num_scheduler_steps": 10,
             "max_num_seqs": 512,
             "dtype": "bfloat16"
         },

.buildkite/nightly-benchmarks/tests/serving-tests-cpu-snc2.json

@@ -0,0 +1,202 @@
+[
+    {
+        "test_name": "serving_llama8B_tp1_sharegpt",
+        "qps_list": ["inf"],
+        "max_concurrency_list": [12, 16, 24, 32, 64, 128, 200],
+        "server_environment_variables": {
+            "VLLM_RPC_TIMEOUT": 100000,
+	    "VLLM_ALLOW_LONG_MAX_MODEL_LEN": 1,
+	    "VLLM_ENGINE_ITERATION_TIMEOUT_S": 120,
+	    "VLLM_CPU_SGL_KERNEL": 1,
+	    "VLLM_CPU_KVCACHE_SPACE": 40
+        },
+        "server_parameters": {
+            "model": "meta-llama/Llama-3.1-8B-Instruct",
+            "tensor_parallel_size": 1,
+	    "dtype": "bfloat16",
+	    "distributed_executor_backend": "mp",
+	    "block_size": 128,
+	    "trust_remote_code": "",
+            "disable_log_stats": "",
+	    "enforce_eager": "",
+	    "max_num_batched_tokens": 2048,
+	    "max_num_seqs": 256,
+            "load_format": "dummy"
+        },
+        "client_parameters": {
+            "model": "meta-llama/Llama-3.1-8B-Instruct",
+            "backend": "vllm",
+            "dataset_name": "sharegpt",
+            "dataset_path": "./ShareGPT_V3_unfiltered_cleaned_split.json",
+            "num_prompts": 200
+        }
+    },
+    {
+        "test_name": "serving_llama8B_tp2_sharegpt",
+        "qps_list": ["inf"],
+        "max_concurrency_list": [12, 16, 24, 32, 64, 128, 200],
+        "server_environment_variables": {
+            "VLLM_RPC_TIMEOUT": 100000,
+	    "VLLM_ALLOW_LONG_MAX_MODEL_LEN": 1,
+	    "VLLM_ENGINE_ITERATION_TIMEOUT_S": 120,
+	    "VLLM_CPU_SGL_KERNEL": 1,
+	    "VLLM_CPU_KVCACHE_SPACE": 40
+        },
+        "server_parameters": {
+            "model": "meta-llama/Llama-3.1-8B-Instruct",
+            "tensor_parallel_size": 2,
+	    "dtype": "bfloat16",
+	    "distributed_executor_backend": "mp",
+	    "block_size": 128,
+	    "trust_remote_code": "",
+            "disable_log_stats": "",
+	    "enforce_eager": "",
+	    "max_num_batched_tokens": 2048,
+	    "max_num_seqs": 256,
+            "load_format": "dummy"
+        },
+        "client_parameters": {
+            "model": "meta-llama/Llama-3.1-8B-Instruct",
+            "backend": "vllm",
+            "dataset_name": "sharegpt",
+            "dataset_path": "./ShareGPT_V3_unfiltered_cleaned_split.json",
+            "num_prompts": 200
+        }
+    },
+    {
+        "test_name": "serving_llama8B_tp4_sharegpt",
+        "qps_list": ["inf"],
+        "max_concurrency_list": [12, 16, 24, 32, 64, 128, 200],
+        "server_environment_variables": {
+            "VLLM_RPC_TIMEOUT": 100000,
+	    "VLLM_ALLOW_LONG_MAX_MODEL_LEN": 1,
+	    "VLLM_ENGINE_ITERATION_TIMEOUT_S": 120,
+	    "VLLM_CPU_SGL_KERNEL": 1,
+	    "VLLM_CPU_KVCACHE_SPACE": 40
+        },
+        "server_parameters": {
+            "model": "meta-llama/Llama-3.1-8B-Instruct",
+            "tensor_parallel_size": 4,
+	    "dtype": "bfloat16",
+	    "distributed_executor_backend": "mp",
+	    "block_size": 128,
+	    "trust_remote_code": "",
+            "disable_log_stats": "",
+	    "enforce_eager": "",
+	    "max_num_batched_tokens": 2048,
+	    "max_num_seqs": 256,
+            "load_format": "dummy"
+        },
+        "client_parameters": {
+            "model": "meta-llama/Llama-3.1-8B-Instruct",
+            "backend": "vllm",
+            "dataset_name": "sharegpt",
+            "dataset_path": "./ShareGPT_V3_unfiltered_cleaned_split.json",
+            "num_prompts": 200
+        }
+    },
+    {
+        "test_name": "serving_llama8B_tp1_random_128_128",
+        "qps_list": ["inf"],
+        "max_concurrency_list": [12, 16, 24, 32, 64, 128, 200, 1000],
+        "server_environment_variables": {
+            "VLLM_RPC_TIMEOUT": 100000,
+	    "VLLM_ALLOW_LONG_MAX_MODEL_LEN": 1,
+	    "VLLM_ENGINE_ITERATION_TIMEOUT_S": 120,
+	    "VLLM_CPU_SGL_KERNEL": 1,
+	    "VLLM_CPU_KVCACHE_SPACE": 40
+        },
+        "server_parameters": {
+            "model": "meta-llama/Llama-3.1-8B-Instruct",
+            "tensor_parallel_size": 1,
+	    "dtype": "bfloat16",
+	    "distributed_executor_backend": "mp",
+	    "block_size": 128,
+	    "trust_remote_code": "",
+	    "enable_chunked_prefill": "",
+            "disable_log_stats": "",
+	    "enforce_eager": "",
+	    "max_num_batched_tokens": 2048,
+	    "max_num_seqs": 256,
+            "load_format": "dummy"
+        },
+        "client_parameters": {
+            "model": "meta-llama/Llama-3.1-8B-Instruct",
+            "backend": "vllm",
+            "dataset_name": "random",
+	    "random-input-len": 128,
+	    "random-output-len": 128,
+	    "ignore-eos": "",
+            "num_prompts": 1000
+        }
+    },
+    {
+        "test_name": "serving_llama8B_tp2_random_128_128",
+        "qps_list": ["inf"],
+        "max_concurrency_list": [12, 16, 24, 32, 64, 128, 200, 1000],
+        "server_environment_variables": {
+            "VLLM_RPC_TIMEOUT": 100000,
+	    "VLLM_ALLOW_LONG_MAX_MODEL_LEN": 1,
+	    "VLLM_ENGINE_ITERATION_TIMEOUT_S": 120,
+	    "VLLM_CPU_SGL_KERNEL": 1,
+	    "VLLM_CPU_KVCACHE_SPACE": 40
+        },
+        "server_parameters": {
+            "model": "meta-llama/Llama-3.1-8B-Instruct",
+            "tensor_parallel_size": 2,
+	    "dtype": "bfloat16",
+	    "distributed_executor_backend": "mp",
+	    "block_size": 128,
+	    "trust_remote_code": "",
+	    "enable_chunked_prefill": "",
+            "disable_log_stats": "",
+	    "enforce_eager": "",
+	    "max_num_batched_tokens": 2048,
+	    "max_num_seqs": 256,
+            "load_format": "dummy"
+        },
+        "client_parameters": {
+            "model": "meta-llama/Llama-3.1-8B-Instruct",
+            "backend": "vllm",
+            "dataset_name": "random",
+	    "random-input-len": 128,
+	    "random-output-len": 128,
+	    "ignore-eos": "",
+            "num_prompts": 1000
+        }
+    },
+    {
+        "test_name": "serving_llama8B_tp4_random_128_128",
+        "qps_list": ["inf"],
+        "max_concurrency_list": [12, 16, 24, 32, 64, 128, 200, 1000],
+        "server_environment_variables": {
+            "VLLM_RPC_TIMEOUT": 100000,
+	    "VLLM_ALLOW_LONG_MAX_MODEL_LEN": 1,
+	    "VLLM_ENGINE_ITERATION_TIMEOUT_S": 120,
+	    "VLLM_CPU_SGL_KERNEL": 1,
+	    "VLLM_CPU_KVCACHE_SPACE": 40
+        },
+        "server_parameters": {
+            "model": "meta-llama/Llama-3.1-8B-Instruct",
+            "tensor_parallel_size": 4,
+	    "dtype": "bfloat16",
+	    "distributed_executor_backend": "mp",
+	    "block_size": 128,
+	    "trust_remote_code": "",
+	    "enable_chunked_prefill": "",
+            "disable_log_stats": "",
+	    "enforce_eager": "",
+	    "max_num_batched_tokens": 2048,
+	    "max_num_seqs": 256,
+            "load_format": "dummy"
+        },
+        "client_parameters": {
+            "model": "meta-llama/Llama-3.1-8B-Instruct",
+            "backend": "vllm",
+            "dataset_name": "random",
+	    "random-input-len": 128,
+	    "random-output-len": 128,
+            "num_prompts": 1000
+        }
+    }
+]

.buildkite/nightly-benchmarks/tests/serving-tests-cpu-snc3.json

@@ -0,0 +1,205 @@
+[
+    {
+        "test_name": "serving_llama8B_pp1_sharegpt",
+        "qps_list": ["inf"],
+        "max_concurrency_list": [12, 16, 24, 32, 64, 128, 200],
+        "server_environment_variables": {
+            "VLLM_RPC_TIMEOUT": 100000,
+	    "VLLM_ALLOW_LONG_MAX_MODEL_LEN": 1,
+	    "VLLM_ENGINE_ITERATION_TIMEOUT_S": 120,
+	    "VLLM_CPU_SGL_KERNEL": 1,
+	    "VLLM_CPU_KVCACHE_SPACE": 40
+        },
+        "server_parameters": {
+            "model": "meta-llama/Llama-3.1-8B-Instruct",
+            "pipeline_parallel_size": 1,
+	    "dtype": "bfloat16",
+	    "distributed_executor_backend": "mp",
+	    "block_size": 128,
+	    "trust_remote_code": "",
+            "disable_log_stats": "",
+	    "enforce_eager": "",
+	    "max_num_batched_tokens": 2048,
+	    "max_num_seqs": 256,
+            "load_format": "dummy"
+        },
+        "client_parameters": {
+            "model": "meta-llama/Llama-3.1-8B-Instruct",
+            "backend": "vllm",
+            "dataset_name": "sharegpt",
+            "dataset_path": "./ShareGPT_V3_unfiltered_cleaned_split.json",
+            "num_prompts": 200
+        }
+    },
+    {
+        "test_name": "serving_llama8B_pp3_sharegpt",
+        "qps_list": ["inf"],
+        "max_concurrency_list": [12, 16, 24, 32, 64, 128, 200],
+        "server_environment_variables": {
+            "VLLM_RPC_TIMEOUT": 100000,
+	    "VLLM_ALLOW_LONG_MAX_MODEL_LEN": 1,
+	    "VLLM_ENGINE_ITERATION_TIMEOUT_S": 120,
+	    "VLLM_CPU_SGL_KERNEL": 1,
+	    "VLLM_CPU_KVCACHE_SPACE": 40
+        },
+        "server_parameters": {
+            "model": "meta-llama/Llama-3.1-8B-Instruct",
+            "pipeline_parallel_size": 3,
+	    "dtype": "bfloat16",
+	    "distributed_executor_backend": "mp",
+	    "block_size": 128,
+	    "trust_remote_code": "",
+            "disable_log_stats": "",
+	    "enforce_eager": "",
+	    "max_num_batched_tokens": 2048,
+	    "max_num_seqs": 256,
+            "load_format": "dummy"
+        },
+        "client_parameters": {
+            "model": "meta-llama/Llama-3.1-8B-Instruct",
+            "backend": "vllm",
+            "dataset_name": "sharegpt",
+            "dataset_path": "./ShareGPT_V3_unfiltered_cleaned_split.json",
+            "num_prompts": 200
+        }
+    },
+    {
+        "test_name": "serving_llama8B_tp2pp3_sharegpt",
+        "qps_list": ["inf"],
+        "max_concurrency_list": [12, 16, 24, 32, 64, 128, 200],
+        "server_environment_variables": {
+            "VLLM_RPC_TIMEOUT": 100000,
+	    "VLLM_ALLOW_LONG_MAX_MODEL_LEN": 1,
+	    "VLLM_ENGINE_ITERATION_TIMEOUT_S": 120,
+	    "VLLM_CPU_SGL_KERNEL": 1,
+	    "VLLM_CPU_KVCACHE_SPACE": 40
+        },
+        "server_parameters": {
+            "model": "meta-llama/Llama-3.1-8B-Instruct",
+            "tensor_parallel_size": 2,
+            "pipeline_parallel_size": 3,
+	    "dtype": "bfloat16",
+	    "distributed_executor_backend": "mp",
+	    "block_size": 128,
+	    "trust_remote_code": "",
+            "disable_log_stats": "",
+	    "enforce_eager": "",
+	    "max_num_batched_tokens": 2048,
+	    "max_num_seqs": 256,
+            "load_format": "dummy"
+        },
+        "client_parameters": {
+            "model": "meta-llama/Llama-3.1-8B-Instruct",
+            "backend": "vllm",
+            "dataset_name": "sharegpt",
+            "dataset_path": "./ShareGPT_V3_unfiltered_cleaned_split.json",
+            "num_prompts": 200
+        }
+    },
+    {
+        "test_name": "serving_llama8B_pp1_random_128_128",
+        "qps_list": ["inf"],
+        "max_concurrency_list": [12, 16, 24, 32, 64, 128, 200, 1000],
+        "server_environment_variables": {
+            "VLLM_RPC_TIMEOUT": 100000,
+	    "VLLM_ALLOW_LONG_MAX_MODEL_LEN": 1,
+	    "VLLM_ENGINE_ITERATION_TIMEOUT_S": 120,
+	    "VLLM_CPU_SGL_KERNEL": 1,
+	    "VLLM_CPU_KVCACHE_SPACE": 40
+        },
+        "server_parameters": {
+            "model": "meta-llama/Llama-3.1-8B-Instruct",
+            "pipeline_parallel_size": 1,
+	    "dtype": "bfloat16",
+	    "distributed_executor_backend": "mp",
+	    "block_size": 128,
+	    "trust_remote_code": "",
+	    "enable_chunked_prefill": "",
+            "disable_log_stats": "",
+	    "enforce_eager": "",
+	    "max_num_batched_tokens": 2048,
+	    "max_num_seqs": 256,
+            "load_format": "dummy"
+        },
+        "client_parameters": {
+            "model": "meta-llama/Llama-3.1-8B-Instruct",
+            "backend": "vllm",
+            "dataset_name": "random",
+	    "random-input-len": 128,
+	    "random-output-len": 128,
+	    "ignore-eos": "",
+            "num_prompts": 1000
+        }
+    },
+    {
+        "test_name": "serving_llama8B_pp3_random_128_128",
+        "qps_list": ["inf"],
+        "max_concurrency_list": [12, 16, 24, 32, 64, 128, 200, 1000],
+        "server_environment_variables": {
+            "VLLM_RPC_TIMEOUT": 100000,
+	    "VLLM_ALLOW_LONG_MAX_MODEL_LEN": 1,
+	    "VLLM_ENGINE_ITERATION_TIMEOUT_S": 120,
+	    "VLLM_CPU_SGL_KERNEL": 1,
+	    "VLLM_CPU_KVCACHE_SPACE": 40
+        },
+        "server_parameters": {
+            "model": "meta-llama/Llama-3.1-8B-Instruct",
+            "pipeline_parallel_size": 3,
+	    "dtype": "bfloat16",
+	    "distributed_executor_backend": "mp",
+	    "block_size": 128,
+	    "trust_remote_code": "",
+	    "enable_chunked_prefill": "",
+            "disable_log_stats": "",
+	    "enforce_eager": "",
+	    "max_num_batched_tokens": 2048,
+	    "max_num_seqs": 256,
+            "load_format": "dummy"
+        },
+        "client_parameters": {
+            "model": "meta-llama/Llama-3.1-8B-Instruct",
+            "backend": "vllm",
+            "dataset_name": "random",
+	    "random-input-len": 128,
+	    "random-output-len": 128,
+	    "ignore-eos": "",
+            "num_prompts": 1000
+        }
+    },
+    {
+        "test_name": "serving_llama8B_tp2pp3_random_128_128",
+        "qps_list": ["inf"],
+        "max_concurrency_list": [12, 16, 24, 32, 64, 128, 200, 1000],
+        "server_environment_variables": {
+            "VLLM_RPC_TIMEOUT": 100000,
+	    "VLLM_ALLOW_LONG_MAX_MODEL_LEN": 1,
+	    "VLLM_ENGINE_ITERATION_TIMEOUT_S": 120,
+	    "VLLM_CPU_SGL_KERNEL": 1,
+	    "VLLM_CPU_KVCACHE_SPACE": 40
+        },
+        "server_parameters": {
+            "model": "meta-llama/Llama-3.1-8B-Instruct",
+            "tensor_parallel_size": 2,
+            "pipeline_parallel_size": 3,
+	    "dtype": "bfloat16",
+	    "distributed_executor_backend": "mp",
+	    "block_size": 128,
+	    "trust_remote_code": "",
+	    "enable_chunked_prefill": "",
+            "disable_log_stats": "",
+	    "enforce_eager": "",
+	    "max_num_batched_tokens": 2048,
+	    "max_num_seqs": 256,
+            "load_format": "dummy"
+        },
+        "client_parameters": {
+            "model": "meta-llama/Llama-3.1-8B-Instruct",
+            "backend": "vllm",
+            "dataset_name": "random",
+	    "random-input-len": 128,
+	    "random-output-len": 128,
+	    "ignore-eos": "",
+            "num_prompts": 1000
+        }
+    }
+]

.buildkite/nightly-benchmarks/tests/serving-tests-cpu.json

@@ -0,0 +1,168 @@
+[
+    {
+        "test_name": "serving_llama8B_tp1_sharegpt",
+        "qps_list": [1, 4, 16, "inf"],
+        "max_concurrency_list": [12, 16, 24, 32, 64, 128, 200],
+        "server_environment_variables": {
+            "VLLM_RPC_TIMEOUT": 100000,
+	    "VLLM_ALLOW_LONG_MAX_MODEL_LEN": 1,
+	    "VLLM_ENGINE_ITERATION_TIMEOUT_S": 120,
+	    "VLLM_CPU_SGL_KERNEL": 1,
+	    "VLLM_CPU_KVCACHE_SPACE": 40
+        },
+        "server_parameters": {
+            "model": "meta-llama/Llama-3.1-8B-Instruct",
+            "tensor_parallel_size": 1,
+	    "dtype": "bfloat16",
+	    "distributed_executor_backend": "mp",
+	    "block_size": 128,
+	    "trust_remote_code": "",
+            "disable_log_stats": "",
+	    "enforce_eager": "",
+	    "max_num_batched_tokens": 2048,
+	    "max_num_seqs": 256,
+            "load_format": "dummy"
+        },
+        "client_parameters": {
+            "model": "meta-llama/Llama-3.1-8B-Instruct",
+            "backend": "vllm",
+            "dataset_name": "sharegpt",
+            "dataset_path": "./ShareGPT_V3_unfiltered_cleaned_split.json",
+            "num_prompts": 200
+        }
+    },
+    {
+        "test_name": "serving_llama8B_tp2_sharegpt",
+        "qps_list": [1, 4, 16, "inf"],
+        "max_concurrency_list": [12, 16, 24, 32, 64, 128, 200],
+        "server_environment_variables": {
+            "VLLM_RPC_TIMEOUT": 100000,
+	    "VLLM_ALLOW_LONG_MAX_MODEL_LEN": 1,
+	    "VLLM_ENGINE_ITERATION_TIMEOUT_S": 120,
+	    "VLLM_CPU_SGL_KERNEL": 1,
+	    "VLLM_CPU_KVCACHE_SPACE": 40
+        },
+        "server_parameters": {
+            "model": "meta-llama/Llama-3.1-8B-Instruct",
+            "tensor_parallel_size": 2,
+	    "dtype": "bfloat16",
+	    "distributed_executor_backend": "mp",
+	    "block_size": 128,
+	    "trust_remote_code": "",
+            "disable_log_stats": "",
+	    "enforce_eager": "",
+	    "max_num_batched_tokens": 2048,
+	    "max_num_seqs": 256,
+            "load_format": "dummy"
+        },
+        "client_parameters": {
+            "model": "meta-llama/Llama-3.1-8B-Instruct",
+            "backend": "vllm",
+            "dataset_name": "sharegpt",
+            "dataset_path": "./ShareGPT_V3_unfiltered_cleaned_split.json",
+            "num_prompts": 200
+        }
+    },
+    {
+        "test_name": "serving_llama8B_tp4_sharegpt",
+        "qps_list": [1, 4, 16, "inf"],
+        "max_concurrency_list": [12, 16, 24, 32, 64, 128, 200],
+        "server_environment_variables": {
+            "VLLM_RPC_TIMEOUT": 100000,
+	    "VLLM_ALLOW_LONG_MAX_MODEL_LEN": 1,
+	    "VLLM_ENGINE_ITERATION_TIMEOUT_S": 120,
+	    "VLLM_CPU_SGL_KERNEL": 1,
+	    "VLLM_CPU_KVCACHE_SPACE": 40
+        },
+        "server_parameters": {
+            "model": "meta-llama/Llama-3.1-8B-Instruct",
+            "tensor_parallel_size": 4,
+	    "dtype": "bfloat16",
+	    "distributed_executor_backend": "mp",
+	    "block_size": 128,
+	    "trust_remote_code": "",
+            "disable_log_stats": "",
+	    "enforce_eager": "",
+	    "max_num_batched_tokens": 2048,
+	    "max_num_seqs": 256,
+            "load_format": "dummy"
+        },
+        "client_parameters": {
+            "model": "meta-llama/Llama-3.1-8B-Instruct",
+            "backend": "vllm",
+            "dataset_name": "sharegpt",
+            "dataset_path": "./ShareGPT_V3_unfiltered_cleaned_split.json",
+            "num_prompts": 200
+        }
+    },
+    {
+        "test_name": "serving_llama8B_tp4_random_1024_128",
+        "qps_list": [1, 4, 16, "inf"],
+        "max_concurrency_list": [12, 16, 24, 32, 64, 128, 200],
+        "server_environment_variables": {
+            "VLLM_RPC_TIMEOUT": 100000,
+	    "VLLM_ALLOW_LONG_MAX_MODEL_LEN": 1,
+	    "VLLM_ENGINE_ITERATION_TIMEOUT_S": 120,
+	    "VLLM_CPU_SGL_KERNEL": 1,
+	    "VLLM_CPU_KVCACHE_SPACE": 40
+        },
+        "server_parameters": {
+            "model": "meta-llama/Llama-3.1-8B-Instruct",
+            "tensor_parallel_size": 4,
+	    "dtype": "bfloat16",
+	    "distributed_executor_backend": "mp",
+	    "block_size": 128,
+	    "trust_remote_code": "",
+	    "enable_chunked_prefill": "",
+            "disable_log_stats": "",
+	    "enforce_eager": "",
+	    "max_num_batched_tokens": 2048,
+	    "max_num_seqs": 256,
+            "load_format": "dummy"
+        },
+        "client_parameters": {
+            "model": "meta-llama/Llama-3.1-8B-Instruct",
+            "backend": "vllm",
+            "dataset_name": "random",
+	    "random-input-len": 1024,
+	    "random-output-len": 128,
+	    "ignore-eos": "",
+            "num_prompts": 100
+        }
+    },
+    {
+        "test_name": "serving_llama8B_pp6_random_1024_128",
+        "qps_list": [1, 4, 16, "inf"],
+        "max_concurrency_list": [12, 16, 24, 32, 64, 128, 200],
+        "server_environment_variables": {
+            "VLLM_RPC_TIMEOUT": 100000,
+	    "VLLM_ALLOW_LONG_MAX_MODEL_LEN": 1,
+	    "VLLM_ENGINE_ITERATION_TIMEOUT_S": 120,
+	    "VLLM_CPU_SGL_KERNEL": 1,
+	    "VLLM_CPU_KVCACHE_SPACE": 40
+        },
+        "server_parameters": {
+            "model": "meta-llama/Llama-3.1-8B-Instruct",
+            "pipeline_parallel_size": 6,
+	    "dtype": "bfloat16",
+	    "distributed_executor_backend": "mp",
+	    "block_size": 128,
+	    "trust_remote_code": "",
+	    "enable_chunked_prefill": "",
+            "disable_log_stats": "",
+	    "enforce_eager": "",
+	    "max_num_batched_tokens": 2048,
+	    "max_num_seqs": 256,
+            "load_format": "dummy"
+        },
+        "client_parameters": {
+            "model": "meta-llama/Llama-3.1-8B-Instruct",
+            "backend": "vllm",
+            "dataset_name": "random",
+	    "random-input-len": 1024,
+	    "random-output-len": 128,
+	    "ignore-eos": "",
+            "num_prompts": 100
+        }
+    }
+]

.buildkite/nightly-benchmarks/tests/serving-tests.json

@@ -7,7 +7,6 @@
             "tensor_parallel_size": 1,
             "swap_space": 16,
             "disable_log_stats": "",
-            "disable_log_requests": "",
             "load_format": "dummy"
         },
         "client_parameters": {
@@ -26,7 +25,6 @@
             "tensor_parallel_size": 4,
             "swap_space": 16,
             "disable_log_stats": "",
-            "disable_log_requests": "",
             "load_format": "dummy"
         },
         "client_parameters": {
@@ -45,7 +43,6 @@
             "tensor_parallel_size": 2,
             "swap_space": 16,
             "disable_log_stats": "",
-            "disable_log_requests": "",
             "load_format": "dummy"
         },
         "client_parameters": {
@@ -60,8 +57,7 @@
         "test_name": "serving_llama70B_tp4_sharegpt_specdecode",
         "qps_list": [2],
         "server_parameters": {
-            "model": "meta-llama/Meta-Llama-3.1-70B-Instruct",
-            "disable_log_requests": "", 
+            "model": "meta-llama/Meta-Llama-3.1-70B-Instruct", 
             "tensor_parallel_size": 4,
             "swap_space": 16,
             "speculative_config": {

.buildkite/nightly-benchmarks/tests/throughput-tests-cpu.json

@@ -0,0 +1,32 @@
+[
+    {
+        "test_name": "throughput_llama8B_tp1",
+        "environment_variables": {
+	    "VLLM_ALLOW_LONG_MAX_MODEL_LEN": 1,
+	    "VLLM_CPU_KVCACHE_SPACE": 40
+        },
+        "parameters": {
+            "model": "meta-llama/Llama-3.1-8B-Instruct",
+            "tensor_parallel_size": 1,
+            "load_format": "dummy",
+            "dataset": "./ShareGPT_V3_unfiltered_cleaned_split.json",
+            "num_prompts": 200,
+            "backend": "vllm"
+        }
+    },
+    {
+        "test_name": "throughput_llama8B_tp4",
+        "environment_variables": {
+	    "VLLM_ALLOW_LONG_MAX_MODEL_LEN": 1,
+	    "VLLM_CPU_KVCACHE_SPACE": 40
+        },
+        "parameters": {
+            "model": "meta-llama/Llama-3.1-8B-Instruct",
+            "tensor_parallel_size": 4,
+            "load_format": "dummy",
+            "dataset": "./ShareGPT_V3_unfiltered_cleaned_split.json",
+            "num_prompts": 200,
+            "backend": "vllm"
+        }
+    }
+]

.buildkite/release-pipeline.yaml

@@ -1,4 +1,20 @@
 steps:
+  # aarch64 + CUDA builds
+  - label: "Build arm64 wheel - CUDA 12.8"
+    id: build-wheel-arm64-cuda-12-8
+    agents:
+      queue: arm64_cpu_queue_postmerge
+    commands:
+      # #NOTE: torch_cuda_arch_list is derived from upstream PyTorch build files here:
+      # https://github.com/pytorch/pytorch/blob/main/.ci/aarch64_linux/aarch64_ci_build.sh#L7
+      - "DOCKER_BUILDKIT=1 docker build --build-arg max_jobs=16 --build-arg USE_SCCACHE=1 --build-arg GIT_REPO_CHECK=1 --build-arg CUDA_VERSION=12.8.1 --build-arg torch_cuda_arch_list='8.7 9.0 10.0+PTX' --tag vllm-ci:build-image --target build --progress plain -f docker/Dockerfile ."
+      - "mkdir artifacts"
+      - "docker run --rm -v $(pwd)/artifacts:/artifacts_host vllm-ci:build-image bash -c 'cp -r dist /artifacts_host && chmod -R a+rw /artifacts_host'"
+      - "bash .buildkite/scripts/upload-wheels.sh"
+    env:
+      DOCKER_BUILDKIT: "1"
+
+  # x86 + CUDA builds
   - label: "Build wheel - CUDA 12.8"
     id: build-wheel-cuda-12-8
     agents:
@@ -52,7 +68,7 @@ steps:
       queue: cpu_queue_postmerge
     commands:
       - "aws ecr-public get-login-password --region us-east-1 | docker login --username AWS --password-stdin public.ecr.aws/q9t5s3a7"
-      - "DOCKER_BUILDKIT=1 docker build --build-arg max_jobs=16 --build-arg USE_SCCACHE=1 --build-arg GIT_REPO_CHECK=1 --build-arg CUDA_VERSION=12.8.1 --tag public.ecr.aws/q9t5s3a7/vllm-release-repo:$BUILDKITE_COMMIT --target vllm-openai --progress plain -f docker/Dockerfile ."
+      - "DOCKER_BUILDKIT=1 docker build --build-arg max_jobs=16 --build-arg USE_SCCACHE=1 --build-arg GIT_REPO_CHECK=1 --build-arg CUDA_VERSION=12.8.1 --build-arg INSTALL_KV_CONNECTORS=true --tag public.ecr.aws/q9t5s3a7/vllm-release-repo:$BUILDKITE_COMMIT --target vllm-openai --progress plain -f docker/Dockerfile ."
       - "docker push public.ecr.aws/q9t5s3a7/vllm-release-repo:$BUILDKITE_COMMIT"
 
   - label: "Annotate release workflow"
@@ -101,7 +117,8 @@ steps:
       queue: cpu_queue_postmerge
     commands:
       - "aws ecr-public get-login-password --region us-east-1 | docker login --username AWS --password-stdin public.ecr.aws/q9t5s3a7"
-      - "DOCKER_BUILDKIT=1 docker build --build-arg max_jobs=16 --build-arg GIT_REPO_CHECK=1 --tag public.ecr.aws/q9t5s3a7/vllm-cpu-release-repo:$(buildkite-agent meta-data get release-version) --tag public.ecr.aws/q9t5s3a7/vllm-cpu-release-repo:latest --progress plain --target vllm-openai -f docker/Dockerfile.cpu ."
+      - "DOCKER_BUILDKIT=1 docker build --build-arg max_jobs=16 --build-arg GIT_REPO_CHECK=1 --build-arg VLLM_CPU_AVX512BF16=true --build-arg VLLM_CPU_AVX512VNNI=true --tag public.ecr.aws/q9t5s3a7/vllm-cpu-release-repo:$(buildkite-agent meta-data get release-version) --tag public.ecr.aws/q9t5s3a7/vllm-cpu-release-repo:latest --progress plain --target vllm-openai -f docker/Dockerfile.cpu ."
+      - "docker push public.ecr.aws/q9t5s3a7/vllm-cpu-release-repo:latest"
       - "docker push public.ecr.aws/q9t5s3a7/vllm-cpu-release-repo:$(buildkite-agent meta-data get release-version)"
     env:
       DOCKER_BUILDKIT: "1"
@@ -117,6 +134,7 @@ steps:
     commands:
       - "aws ecr-public get-login-password --region us-east-1 | docker login --username AWS --password-stdin public.ecr.aws/q9t5s3a7"
       - "DOCKER_BUILDKIT=1 docker build --build-arg max_jobs=16 --build-arg GIT_REPO_CHECK=1 --tag public.ecr.aws/q9t5s3a7/vllm-neuron-release-repo:$(buildkite-agent meta-data get release-version) --tag public.ecr.aws/q9t5s3a7/vllm-neuron-release-repo:latest --progress plain -f docker/Dockerfile.neuron ."
+      - "docker push public.ecr.aws/q9t5s3a7/vllm-neuron-release-repo:latest"
       - "docker push public.ecr.aws/q9t5s3a7/vllm-neuron-release-repo:$(buildkite-agent meta-data get release-version)"
     env:
       DOCKER_BUILDKIT: "1"

.buildkite/scripts/hardware_ci/run-amd-test.sh

@@ -107,10 +107,8 @@ fi
 
 if [[ $commands == *" kernels/attention"* ]]; then
   commands="${commands} \
-  --ignore=kernels/attention/stest_attention_selector.py \
-  --ignore=kernels/attention/test_blocksparse_attention.py \
-  --ignore=kernels/attention/test_encoder_decoder_attn.py \
   --ignore=kernels/attention/test_attention_selector.py \
+  --ignore=kernels/attention/test_encoder_decoder_attn.py \
   --ignore=kernels/attention/test_flash_attn.py \
   --ignore=kernels/attention/test_flashinfer.py \
   --ignore=kernels/attention/test_prefix_prefill.py \
@@ -123,7 +121,6 @@ fi
 if [[ $commands == *" kernels/quantization"* ]]; then
   commands="${commands} \
   --ignore=kernels/quantization/test_int8_quant.py \
-  --ignore=kernels/quantization/test_aqlm.py \
   --ignore=kernels/quantization/test_machete_mm.py \
   --ignore=kernels/quantization/test_block_fp8.py \
   --ignore=kernels/quantization/test_block_int8.py \

.buildkite/scripts/hardware_ci/run-cpu-test.sh

@@ -6,15 +6,16 @@ set -ex
 
 # allow to bind to different cores
 CORE_RANGE=${CORE_RANGE:-48-95}
+# used for TP/PP E2E test
 OMP_CORE_RANGE=${OMP_CORE_RANGE:-48-95}
 NUMA_NODE=${NUMA_NODE:-1}
 
 export CMAKE_BUILD_PARALLEL_LEVEL=32
 
 # Setup cleanup
-remove_docker_container() { 
-    set -e; 
-    docker rm -f cpu-test-"$NUMA_NODE" cpu-test-"$NUMA_NODE"-avx2 || true; 
+remove_docker_container() {
+    set -e;
+    docker rm -f cpu-test-"$NUMA_NODE" cpu-test-"$NUMA_NODE"-avx2 || true;
 }
 trap remove_docker_container EXIT
 remove_docker_container
@@ -24,13 +25,22 @@ numactl -C "$CORE_RANGE" -N "$NUMA_NODE" docker build --tag cpu-test-"$NUMA_NODE
 numactl -C "$CORE_RANGE" -N "$NUMA_NODE" docker build --build-arg VLLM_CPU_DISABLE_AVX512="true" --tag cpu-test-"$NUMA_NODE"-avx2 --target vllm-test -f docker/Dockerfile.cpu .
 
 # Run the image, setting --shm-size=4g for tensor parallel.
-docker run -itd --cpuset-cpus="$CORE_RANGE" --cpuset-mems="$NUMA_NODE" --entrypoint /bin/bash -v ~/.cache/huggingface:/root/.cache/huggingface --privileged=true -e HF_TOKEN --env VLLM_CPU_KVCACHE_SPACE=4 --env VLLM_CPU_OMP_THREADS_BIND="$OMP_CORE_RANGE" --shm-size=4g --name cpu-test-"$NUMA_NODE" cpu-test-"$NUMA_NODE"
-docker run -itd --cpuset-cpus="$CORE_RANGE" --cpuset-mems="$NUMA_NODE" --entrypoint /bin/bash -v ~/.cache/huggingface:/root/.cache/huggingface --privileged=true -e HF_TOKEN --env VLLM_CPU_KVCACHE_SPACE=4 --env VLLM_CPU_OMP_THREADS_BIND="$OMP_CORE_RANGE" --shm-size=4g --name cpu-test-"$NUMA_NODE"-avx2 cpu-test-"$NUMA_NODE"-avx2
+docker run -itd --cpuset-cpus="$CORE_RANGE" --cpuset-mems="$NUMA_NODE" --entrypoint /bin/bash -v ~/.cache/huggingface:/root/.cache/huggingface --privileged=true -e HF_TOKEN --env VLLM_CPU_KVCACHE_SPACE=4 --env VLLM_CPU_CI_ENV=1 -e E2E_OMP_THREADS="$OMP_CORE_RANGE" --shm-size=4g --name cpu-test-"$NUMA_NODE" cpu-test-"$NUMA_NODE"
+docker run -itd --cpuset-cpus="$CORE_RANGE" --cpuset-mems="$NUMA_NODE" --entrypoint /bin/bash -v ~/.cache/huggingface:/root/.cache/huggingface --privileged=true -e HF_TOKEN --env VLLM_CPU_KVCACHE_SPACE=4 --env VLLM_CPU_CI_ENV=1 -e E2E_OMP_THREADS="$OMP_CORE_RANGE" --shm-size=4g --name cpu-test-"$NUMA_NODE"-avx2 cpu-test-"$NUMA_NODE"-avx2
 
 function cpu_tests() {
   set -e
   export NUMA_NODE=$2
 
+  # list packages
+  docker exec cpu-test-"$NUMA_NODE"-avx2 bash -c "
+    set -e
+    pip list"
+
+  docker exec cpu-test-"$NUMA_NODE" bash -c "
+    set -e
+    pip list"
+
   # offline inference
   docker exec cpu-test-"$NUMA_NODE"-avx2 bash -c "
     set -e
@@ -39,9 +49,16 @@ function cpu_tests() {
   # Run basic model test
   docker exec cpu-test-"$NUMA_NODE" bash -c "
     set -e
-    pytest -v -s tests/kernels/attention/test_cache.py -m cpu_model
-    pytest -v -s tests/kernels/attention/test_mla_decode_cpu.py -m cpu_model
-    pytest -v -s tests/models/language/generation -m cpu_model
+    # Note: disable until supports V1
+    # pytest -v -s tests/kernels/attention/test_cache.py -m cpu_model
+    # pytest -v -s tests/kernels/attention/test_mla_decode_cpu.py -m cpu_model
+
+    # Note: disable Bart until supports V1
+    pytest -v -s tests/models/language/generation -m cpu_model \
+                --ignore=tests/models/language/generation/test_bart.py
+    VLLM_CPU_SGL_KERNEL=1 pytest -v -s tests/models/language/generation -m cpu_model \
+                --ignore=tests/models/language/generation/test_bart.py
+
     pytest -v -s tests/models/language/pooling -m cpu_model
     pytest -v -s tests/models/multimodal/generation \
                 --ignore=tests/models/multimodal/generation/test_mllama.py \
@@ -52,41 +69,34 @@ function cpu_tests() {
   docker exec cpu-test-"$NUMA_NODE" bash -c "
     set -e
     pytest -s -v \
-    tests/quantization/test_compressed_tensors.py::test_compressed_tensors_w8a8_static_setup \
-    tests/quantization/test_compressed_tensors.py::test_compressed_tensors_w8a8_dynamic_per_token"
+    tests/quantization/test_compressed_tensors.py::test_compressed_tensors_w8a8_logprobs[False-10-32-neuralmagic/Llama-3.2-1B-quantized.w8a8]"
 
+  # Note: disable it until supports V1
   # Run AWQ test
-  docker exec cpu-test-"$NUMA_NODE" bash -c "
-    set -e
-    VLLM_USE_V1=0 pytest -s -v \
-    tests/quantization/test_ipex_quant.py"
-
-  # Run chunked-prefill and prefix-cache test
-  docker exec cpu-test-"$NUMA_NODE" bash -c "
-    set -e
-    pytest -s -v -k cpu_model \
-    tests/basic_correctness/test_chunked_prefill.py"  
-
-  # online serving
-  docker exec cpu-test-"$NUMA_NODE" bash -c "
-    set -e
-    python3 -m vllm.entrypoints.openai.api_server --model facebook/opt-125m --dtype half & 
-    timeout 600 bash -c 'until curl localhost:8000/v1/models; do sleep 1; done' || exit 1
-    python3 benchmarks/benchmark_serving.py \
-      --backend vllm \
-      --dataset-name random \
-      --model facebook/opt-125m \
-      --num-prompts 20 \
-      --endpoint /v1/completions \
-      --tokenizer facebook/opt-125m"
+  # docker exec cpu-test-"$NUMA_NODE" bash -c "
+  #   set -e
+  #   VLLM_USE_V1=0 pytest -s -v \
+  #   tests/quantization/test_ipex_quant.py"
 
   # Run multi-lora tests
   docker exec cpu-test-"$NUMA_NODE" bash -c "
     set -e
     pytest -s -v \
     tests/lora/test_qwen2vl.py"
+
+  # online serving
+  docker exec cpu-test-"$NUMA_NODE" bash -c '
+    set -e
+    VLLM_CPU_OMP_THREADS_BIND=$E2E_OMP_THREADS VLLM_CPU_SGL_KERNEL=1 vllm serve meta-llama/Llama-3.2-3B-Instruct -tp=2 -pp=2 &
+    timeout 600 bash -c "until curl localhost:8000/v1/models; do sleep 1; done" || exit 1
+    vllm bench serve \
+      --backend vllm \
+      --dataset-name random \
+      --model meta-llama/Llama-3.2-3B-Instruct \
+      --num-prompts 20 \
+      --endpoint /v1/completions'
 }
 
 # All of CPU tests are expected to be finished less than 40 mins.
 export -f cpu_tests
-timeout 1h bash -c "cpu_tests $CORE_RANGE $NUMA_NODE"
+timeout 1.5h bash -c "cpu_tests $CORE_RANGE $NUMA_NODE"

.buildkite/scripts/hardware_ci/run-gh200-test.sh

@@ -16,8 +16,7 @@ DOCKER_BUILDKIT=1 docker build . \
   --build-arg max_jobs=66 \
   --build-arg nvcc_threads=2 \
   --build-arg RUN_WHEEL_CHECK=false \
-  --build-arg torch_cuda_arch_list="9.0+PTX" \
-  --build-arg vllm_fa_cmake_gpu_arches="90-real"
+  --build-arg torch_cuda_arch_list="9.0+PTX"
 
 # Setup cleanup
 remove_docker_container() { docker rm -f gh200-test || true; }

.buildkite/scripts/hardware_ci/run-hpu-test.sh

@@ -2,10 +2,32 @@
 
 # This script build the CPU docker image and run the offline inference inside the container.
 # It serves a sanity check for compilation and basic model usage.
-set -ex
+set -exuo pipefail
 
 # Try building the docker image
-docker build -t hpu-test-env -f docker/Dockerfile.hpu .
+cat <<EOF | docker build -t hpu-plugin-v1-test-env -f - .
+FROM gaudi-base-image:latest
+
+COPY ./ /workspace/vllm
+
+WORKDIR /workspace/vllm
+
+ENV no_proxy=localhost,127.0.0.1
+ENV PT_HPU_ENABLE_LAZY_COLLECTIVES=true
+
+RUN VLLM_TARGET_DEVICE=empty pip install .
+RUN pip install git+https://github.com/vllm-project/vllm-gaudi.git
+
+# install development dependencies (for testing)
+RUN python3 -m pip install -e tests/vllm_test_utils
+
+WORKDIR /workspace/
+
+RUN git clone https://github.com/vllm-project/vllm-gaudi.git
+
+RUN ln -s /workspace/vllm/tests && ln -s /workspace/vllm/examples && ln -s /workspace/vllm/benchmarks
+
+EOF
 
 # Setup cleanup
 # certain versions of HPU software stack have a bug that can
@@ -14,13 +36,21 @@ docker build -t hpu-test-env -f docker/Dockerfile.hpu .
 # functions, while other platforms only need one remove_docker_container
 # function.
 EXITCODE=1
-remove_docker_containers() { docker rm -f hpu-test || true; docker rm -f hpu-test-tp2 || true; }
-remove_docker_containers_and_exit() { remove_docker_containers; exit $EXITCODE; }
-trap remove_docker_containers_and_exit EXIT
+remove_docker_containers() { docker rm -f hpu-plugin-v1-test || true; }
+trap 'remove_docker_containers; exit $EXITCODE;' EXIT
 remove_docker_containers
 
-# Run the image and launch offline inference
-docker run --runtime=habana --name=hpu-test --network=host -e HABANA_VISIBLE_DEVICES=all -e VLLM_SKIP_WARMUP=true --entrypoint="" hpu-test-env python3 examples/offline_inference/basic/generate.py --model facebook/opt-125m
-docker run --runtime=habana --name=hpu-test-tp2 --network=host -e HABANA_VISIBLE_DEVICES=all -e VLLM_SKIP_WARMUP=true --entrypoint="" hpu-test-env python3 examples/offline_inference/basic/generate.py --model facebook/opt-125m --tensor-parallel-size 2
+echo "Running HPU plugin v1 test"
+docker run --rm --runtime=habana --name=hpu-plugin-v1-test --network=host \
+  -e HABANA_VISIBLE_DEVICES=all \
+  hpu-plugin-v1-test-env \
+  /bin/bash "/workspace/vllm-gaudi/tests/upstream_tests/ci_tests.sh"
 
 EXITCODE=$?
+if [ $EXITCODE -eq 0 ]; then
+  echo "Test with basic model passed"
+else
+  echo "Test with basic model FAILED with exit code: $EXITCODE" >&2
+fi
+
+# The trap will handle the container removal and final exit.

.buildkite/scripts/hardware_ci/run-neuron-test.sh

@@ -54,10 +54,11 @@ docker run --rm -it --device=/dev/neuron0 --network bridge \
        --name "${container_name}" \
        ${image_name} \
        /bin/bash -c "
+            set -e; # Exit on first error
             python3 /workspace/vllm/examples/offline_inference/neuron.py;
             python3 -m pytest /workspace/vllm/tests/neuron/1_core/ -v --capture=tee-sys;
             for f in /workspace/vllm/tests/neuron/2_core/*.py; do
-                echo 'Running test file: '$f;
+                echo \"Running test file: \$f\";
                 python3 -m pytest \$f -v --capture=tee-sys;
             done
        "

.buildkite/scripts/hardware_ci/run-tpu-v1-test-part2.sh

@@ -0,0 +1,167 @@
+#!/bin/bash
+
+set -xu
+
+
+remove_docker_container() { 
+    docker rm -f tpu-test || true;
+}
+
+trap remove_docker_container EXIT
+
+# Remove the container that might not be cleaned up in the previous run.
+remove_docker_container
+
+# Build the docker image.
+docker build -f docker/Dockerfile.tpu -t vllm-tpu .
+
+# Set up cleanup.
+cleanup_docker() {
+  # Get Docker's root directory
+  docker_root=$(docker info -f '{{.DockerRootDir}}')
+  if [ -z "$docker_root" ]; then
+    echo "Failed to determine Docker root directory."
+    exit 1
+  fi
+  echo "Docker root directory: $docker_root"
+  # Check disk usage of the filesystem where Docker's root directory is located
+  disk_usage=$(df "$docker_root" | tail -1 | awk '{print $5}' | sed 's/%//')
+  # Define the threshold
+  threshold=70
+  if [ "$disk_usage" -gt "$threshold" ]; then
+    echo "Disk usage is above $threshold%. Cleaning up Docker images and volumes..."
+    # Remove dangling images (those that are not tagged and not used by any container)
+    docker image prune -f
+    # Remove unused volumes / force the system prune for old images as well.
+    docker volume prune -f && docker system prune --force --filter "until=72h" --all
+    echo "Docker images and volumes cleanup completed."
+  else
+    echo "Disk usage is below $threshold%. No cleanup needed."
+  fi
+}
+cleanup_docker
+
+# For HF_TOKEN.
+source /etc/environment
+
+docker run --privileged --net host --shm-size=16G -it \
+    -e "HF_TOKEN=$HF_TOKEN" --name tpu-test \
+    vllm-tpu /bin/bash -c '
+set -e # Exit immediately if a command exits with a non-zero status.
+set -u # Treat unset variables as an error.
+
+echo "--- Starting script inside Docker container ---"
+
+# Create results directory
+RESULTS_DIR=$(mktemp -d)
+# If mktemp fails, set -e will cause the script to exit.
+echo "Results will be stored in: $RESULTS_DIR"
+
+# Install dependencies
+echo "--- Installing Python dependencies ---"
+python3 -m pip install --progress-bar off git+https://github.com/thuml/depyf.git \
+    && python3 -m pip install --progress-bar off pytest pytest-asyncio tpu-info \
+    && python3 -m pip install --progress-bar off lm_eval[api]==0.4.4 \
+    && python3 -m pip install --progress-bar off hf-transfer
+echo "--- Python dependencies installed ---"
+export VLLM_USE_V1=1
+export VLLM_XLA_CHECK_RECOMPILATION=1
+export VLLM_XLA_CACHE_PATH=
+echo "Using VLLM V1"
+
+echo "--- Hardware Information ---"
+# tpu-info
+echo "--- Starting Tests ---"
+set +e
+overall_script_exit_code=0
+
+# --- Test Definitions ---
+# If a test fails, this function will print logs and will not cause the main script to exit.
+run_test() {
+    local test_num=$1
+    local test_name=$2
+    local test_command=$3
+    local log_file="$RESULTS_DIR/test_${test_num}.log"
+    local actual_exit_code
+
+    echo "--- TEST_$test_num: Running $test_name ---"
+    
+    # Execute the test command.
+    eval "$test_command" > >(tee -a "$log_file") 2> >(tee -a "$log_file" >&2)
+    actual_exit_code=$?
+
+    echo "TEST_${test_num}_COMMAND_EXIT_CODE: $actual_exit_code" # This goes to main log
+    echo "TEST_${test_num}_COMMAND_EXIT_CODE: $actual_exit_code" >> "$log_file" # Also to per-test log
+
+    if [ "$actual_exit_code" -ne 0 ]; then
+        echo "TEST_$test_num ($test_name) FAILED with exit code $actual_exit_code." >&2
+        echo "--- Log for failed TEST_$test_num ($test_name) ---" >&2
+        if [ -f "$log_file" ]; then
+            cat "$log_file" >&2
+        else
+            echo "Log file $log_file not found for TEST_$test_num ($test_name)." >&2
+        fi
+        echo "--- End of log for TEST_$test_num ($test_name) ---" >&2
+        return "$actual_exit_code" # Return the failure code
+    else
+        echo "TEST_$test_num ($test_name) PASSED."
+        return 0 # Return success
+    fi
+}
+
+# Helper function to call run_test and update the overall script exit code
+run_and_track_test() {
+    local test_num_arg="$1"
+    local test_name_arg="$2"
+    local test_command_arg="$3"
+
+    # Run the test
+    run_test "$test_num_arg" "$test_name_arg" "$test_command_arg"
+    local test_specific_exit_code=$?
+
+    # If the test failed, set the overall script exit code to 1
+    if [ "$test_specific_exit_code" -ne 0 ]; then
+        # No need for extra echo here, run_test already logged the failure.
+        overall_script_exit_code=1
+    fi
+}
+
+# --- Actual Test Execution ---
+run_and_track_test 1 "test_struct_output_generate.py" \
+    "python3 -m pytest -s -v /workspace/vllm/tests/v1/entrypoints/llm/test_struct_output_generate.py -k \"not test_structured_output_with_reasoning_matrices\""
+run_and_track_test 2 "test_moe_pallas.py" \
+    "python3 -m pytest -s -v /workspace/vllm/tests/tpu/test_moe_pallas.py"
+run_and_track_test 3 "test_lora.py" \
+    "VLLM_XLA_CHECK_RECOMPILATION=0 python3 -m pytest -s -v /workspace/vllm/tests/tpu/lora/test_lora.py"
+run_and_track_test 4 "test_tpu_qkv_linear.py" \
+    "python3 -m pytest -s -v /workspace/vllm/tests/v1/tpu/test_tpu_qkv_linear.py"
+run_and_track_test 5 "test_spmd_model_weight_loading.py" \
+    "python3 -m pytest -s -v /workspace/vllm/tests/v1/tpu/test_spmd_model_weight_loading.py"
+run_and_track_test 6 "test_kv_cache_update_kernel.py" \
+    "python3 -m pytest -s -v /workspace/vllm/tests/v1/tpu/test_kv_cache_update_kernel.py"
+run_and_track_test 7 "test_tpu_int8.py" \
+    "python3 -m pytest -s -v /workspace/vllm/tests/v1/tpu/test_tpu_int8.py"
+
+# After all tests have been attempted, exit with the overall status.
+if [ "$overall_script_exit_code" -ne 0 ]; then
+    echo "--- One or more tests FAILED. Overall script exiting with failure code 1. ---"
+else
+    echo "--- All tests have completed and PASSED. Overall script exiting with success code 0. ---"
+fi
+exit "$overall_script_exit_code"
+' # IMPORTANT: This is the closing single quote for the bash -c "..." command. Ensure it is present and correct.
+
+# Capture the exit code of the docker run command
+DOCKER_RUN_EXIT_CODE=$?
+
+# The trap will run for cleanup.
+# Exit the main script with the Docker run command's exit code.
+if [ "$DOCKER_RUN_EXIT_CODE" -ne 0 ]; then
+    echo "Docker run command failed with exit code $DOCKER_RUN_EXIT_CODE."
+    exit "$DOCKER_RUN_EXIT_CODE"
+else
+    echo "Docker run command completed successfully."
+    exit 0
+fi
+# TODO: This test fails because it uses RANDOM_SEED sampling
+# pytest -v -s /workspace/vllm/tests/tpu/test_custom_dispatcher.py \

.buildkite/scripts/hardware_ci/run-tpu-v1-test.sh

@@ -5,7 +5,6 @@ set -xu
 
 remove_docker_container() { 
     docker rm -f tpu-test || true; 
-    docker rm -f vllm-tpu || true;
 }
 
 trap remove_docker_container EXIT
@@ -62,7 +61,8 @@ echo "Results will be stored in: $RESULTS_DIR"
 echo "--- Installing Python dependencies ---"
 python3 -m pip install --progress-bar off git+https://github.com/thuml/depyf.git \
     && python3 -m pip install --progress-bar off pytest pytest-asyncio tpu-info \
-    && python3 -m pip install --progress-bar off lm_eval[api]==0.4.4
+    && python3 -m pip install --progress-bar off lm_eval[api]==0.4.4 \
+    && python3 -m pip install --progress-bar off hf-transfer
 echo "--- Python dependencies installed ---"
 export VLLM_USE_V1=1
 export VLLM_XLA_CHECK_RECOMPILATION=1
@@ -70,7 +70,7 @@ export VLLM_XLA_CACHE_PATH=
 echo "Using VLLM V1"
 
 echo "--- Hardware Information ---"
-tpu-info
+# tpu-info
 echo "--- Starting Tests ---"
 set +e
 overall_script_exit_code=0
@@ -149,16 +149,6 @@ run_and_track_test 9 "test_multimodal.py" \
     "python3 -m pytest -s -v /workspace/vllm/tests/v1/tpu/test_multimodal.py"
 run_and_track_test 10 "test_pallas.py" \
     "python3 -m pytest -s -v /workspace/vllm/tests/v1/tpu/test_pallas.py"
-run_and_track_test 11 "test_struct_output_generate.py" \
-    "python3 -m pytest -s -v /workspace/vllm/tests/v1/entrypoints/llm/test_struct_output_generate.py -k \"not test_structured_output_with_reasoning_matrices\""
-run_and_track_test 12 "test_moe_pallas.py" \
-    "python3 -m pytest -s -v /workspace/vllm/tests/tpu/test_moe_pallas.py"
-run_and_track_test 13 "test_lora.py" \
-    "VLLM_XLA_CHECK_RECOMPILATION=0 python3 -m pytest -s -v /workspace/vllm/tests/tpu/lora/test_lora.py"
-run_and_track_test 14 "test_tpu_qkv_linear.py" \
-    "python3 -m pytest -s -v /workspace/vllm/tests/v1/tpu/test_tpu_qkv_linear.py"
-run_and_track_test 15 "test_spmd_model_weight_loading.py" \
-    "python3 -m pytest -s -v /workspace/vllm/tests/v1/tpu/test_spmd_model_weight_loading.py"
 
 # After all tests have been attempted, exit with the overall status.
 if [ "$overall_script_exit_code" -ne 0 ]; then

.buildkite/scripts/hardware_ci/run-xpu-test.sh

@@ -11,8 +11,8 @@ container_name="xpu_${BUILDKITE_COMMIT}_$(tr -dc A-Za-z0-9 < /dev/urandom | head
 docker build -t ${image_name} -f docker/Dockerfile.xpu .
 
 # Setup cleanup
-remove_docker_container() { 
-  docker rm -f "${container_name}" || true; 
+remove_docker_container() {
+  docker rm -f "${container_name}" || true;
   docker image rm -f "${image_name}" || true;
   docker system prune -f || true;
 }
@@ -26,6 +26,18 @@ docker run \
     --name "${container_name}" \
     "${image_name}" \
     sh -c '
-    VLLM_USE_V1=0 python3 examples/offline_inference/basic/generate.py --model facebook/opt-125m
-    VLLM_USE_V1=0 python3 examples/offline_inference/basic/generate.py --model facebook/opt-125m -tp 2
+    VLLM_USE_V1=1 python3 examples/offline_inference/basic/generate.py --model facebook/opt-125m --block-size 64 --enforce-eager
+    VLLM_USE_V1=1 python3 examples/offline_inference/basic/generate.py --model facebook/opt-125m --block-size 64 --enforce-eager -tp 2 --distributed-executor-backend ray
+    VLLM_USE_V1=1 python3 examples/offline_inference/basic/generate.py --model facebook/opt-125m --block-size 64 --enforce-eager -tp 2 --distributed-executor-backend mp
+    cd tests
+    pytest -v -s v1/core
+    pytest -v -s v1/engine
+    pytest -v -s v1/sample --ignore=v1/sample/test_logprobs.py --ignore=v1/sample/test_logprobs_e2e.py
+    pytest -v -s v1/worker --ignore=v1/worker/test_gpu_model_runner.py
+    pytest -v -s v1/structured_output
+    pytest -v -s v1/spec_decode --ignore=v1/spec_decode/test_max_len.py --ignore=v1/spec_decode/test_eagle.py
+    pytest -v -s v1/kv_connector/unit --ignore=v1/kv_connector/unit/test_multi_connector.py --ignore=v1/kv_connector/unit/test_nixl_connector.py
+    pytest -v -s v1/test_serial_utils.py
+    pytest -v -s v1/test_utils.py
+    pytest -v -s v1/test_metrics_reader.py
 '

.buildkite/scripts/run-benchmarks.sh

@@ -11,10 +11,10 @@ cd "$(dirname "${BASH_SOURCE[0]}")/../.."
 (which wget && which curl) || (apt-get update && apt-get install -y wget curl)
 
 # run python-based benchmarks and upload the result to buildkite
-python3 benchmarks/benchmark_latency.py --output-json latency_results.json 2>&1 | tee benchmark_latency.txt
+vllm bench latency --output-json latency_results.json 2>&1 | tee benchmark_latency.txt
 bench_latency_exit_code=$?
 
-python3 benchmarks/benchmark_throughput.py --input-len 256 --output-len 256 --output-json throughput_results.json 2>&1 | tee benchmark_throughput.txt
+vllm bench throughput --input-len 256 --output-len 256 --output-json throughput_results.json 2>&1 | tee benchmark_throughput.txt
 bench_throughput_exit_code=$?
 
 # run server-based benchmarks and upload the result to buildkite
@@ -24,7 +24,7 @@ wget https://huggingface.co/datasets/anon8231489123/ShareGPT_Vicuna_unfiltered/r
 
 # wait for server to start, timeout after 600 seconds
 timeout 600 bash -c 'until curl localhost:8000/v1/models; do sleep 1; done' || exit 1
-python3 benchmarks/benchmark_serving.py \
+vllm bench serve \
     --backend vllm \
     --dataset-name sharegpt \
     --dataset-path ./ShareGPT_V3_unfiltered_cleaned_split.json \

.buildkite/scripts/tpu/config_v6e_1.env

@@ -1,11 +1,11 @@
 # Environment config
 TEST_NAME=llama8b
-CONTAINER_NAME=vllm-tpu
+CONTAINER_NAME=tpu-test
 
 # vllm config
 MODEL=meta-llama/Llama-3.1-8B-Instruct
-MAX_NUM_SEQS=512
-MAX_NUM_BATCHED_TOKENS=512
+MAX_NUM_SEQS=256
+MAX_NUM_BATCHED_TOKENS=1024
 TENSOR_PARALLEL_SIZE=1
 MAX_MODEL_LEN=2048
 DOWNLOAD_DIR=/mnt/disks/persist

.buildkite/scripts/tpu/docker_run_bm.sh

@@ -12,8 +12,6 @@ source /etc/environment
 source $ENV_FILE
 
 remove_docker_container() { 
-    docker rm -f tpu-test || true; 
-    docker rm -f vllm-tpu || true;
     docker rm -f $CONTAINER_NAME || true;
 }
 
@@ -22,16 +20,6 @@ trap remove_docker_container EXIT
 # Remove the container that might not be cleaned up in the previous run.
 remove_docker_container
 
-# Build docker image.
-# TODO: build the image outside the script and share the image with other
-# tpu test if building time is too long.
-DOCKER_BUILDKIT=1 docker build \
-  --build-arg max_jobs=16 \
-  --build-arg USE_SCCACHE=1 \
-  --build-arg GIT_REPO_CHECK=0 \
-  --tag vllm/vllm-tpu-bm \
-  --progress plain -f docker/Dockerfile.tpu .
-
 LOG_ROOT=$(mktemp -d)
 # If mktemp fails, set -e will cause the script to exit.
 echo "Results will be stored in: $LOG_ROOT"
@@ -68,7 +56,7 @@ docker run \
 
 echo "run script..."
 echo
-docker exec "$CONTAINER_NAME" /bin/bash -c ".buildkite/scripts/hardware_ci/run_bm.sh"
+docker exec "$CONTAINER_NAME" /bin/bash -c ".buildkite/scripts/tpu/run_bm.sh"
 
 echo "copy result back..."
 VLLM_LOG="$LOG_ROOT/$TEST_NAME"_vllm_log.txt

.buildkite/scripts/tpu/quantized_v6e_1.env

@@ -0,0 +1,14 @@
+# Environment config
+TEST_NAME=llama8bw8a8
+CONTAINER_NAME=tpu-test
+
+# vllm config
+MODEL=RedHatAI/Meta-Llama-3.1-8B-Instruct-quantized.w8a8
+MAX_NUM_SEQS=128
+MAX_NUM_BATCHED_TOKENS=1024
+TENSOR_PARALLEL_SIZE=1
+MAX_MODEL_LEN=2048
+DOWNLOAD_DIR=/mnt/disks/persist
+EXPECTED_THROUGHPUT=10.0
+INPUT_LEN=1800
+OUTPUT_LEN=128

.buildkite/scripts/tpu/run_bm.sh

@@ -44,7 +44,6 @@ echo
 
 VLLM_USE_V1=1 vllm serve $MODEL \
  --seed 42 \
- --disable-log-requests \
  --max-num-seqs $MAX_NUM_SEQS \
  --max-num-batched-tokens $MAX_NUM_BATCHED_TOKENS \
  --tensor-parallel-size $TENSOR_PARALLEL_SIZE \
@@ -77,7 +76,7 @@ done
 echo "run benchmark test..."
 echo "logging to $BM_LOG"
 echo
-python benchmarks/benchmark_serving.py \
+vllm bench serve \
     --backend vllm \
     --model $MODEL  \
     --dataset-name sonnet \

.buildkite/test-pipeline.yaml

@@ -31,15 +31,15 @@
 steps:
 ##### fast check tests  #####
 
-- label: Documentation Build # 2min
-  mirror_hardwares: [amdexperimental]
-  working_dir: "/vllm-workspace/test_docs"
-  fast_check: true
-  no_gpu: True
+- label: Pytorch Nightly Dependency Override Check # 2min
+  # if this test fails, it means the nightly torch version is not compatible with some
+  # of the dependencies. Please check the error message and add the package to whitelist
+  # in /vllm/tools/generate_nightly_torch_test.py
+  soft_fail: true
+  source_file_dependencies:
+  - requirements/nightly_torch_test.txt
   commands:
-  - pip install -r ../requirements/docs.txt
-  # TODO: add `--strict` once warnings in docstrings are fixed
-  - mkdocs build
+  - bash standalone_tests/pytorch_nightly_dependency.sh
 
 - label: Async Engine, Inputs, Utils, Worker Test # 24min
   mirror_hardwares: [amdexperimental]
@@ -47,20 +47,20 @@ steps:
   - vllm/
   - tests/mq_llm_engine
   - tests/async_engine
-  - tests/test_inputs
+  - tests/test_inputs.py
+  - tests/test_outputs.py
   - tests/multimodal
-  - tests/test_utils
+  - tests/utils_
   - tests/worker
   - tests/standalone_tests/lazy_imports.py
   commands:
   - python3 standalone_tests/lazy_imports.py
   - pytest -v -s mq_llm_engine # MQLLMEngine
   - pytest -v -s async_engine # AsyncLLMEngine
-  - NUM_SCHEDULER_STEPS=4 pytest -v -s async_engine/test_async_llm_engine.py
   - pytest -v -s test_inputs.py
   - pytest -v -s test_outputs.py
   - pytest -v -s multimodal
-  - pytest -v -s test_utils.py # Utils
+  - pytest -v -s utils_ # Utils
   - pytest -v -s worker # Worker
 
 - label: Python-only Installation Test
@@ -72,7 +72,7 @@ steps:
   - bash standalone_tests/python_only_compile.sh
 
 - label: Basic Correctness Test # 30min
-  mirror_hardwares: [amdexperimental, amdproduction]
+  mirror_hardwares: [amdexperimental]
   fast_check: true
   torch_nightly: true
   source_file_dependencies:
@@ -98,7 +98,7 @@ steps:
   - VLLM_ATTENTION_BACKEND=FLASH_ATTN pytest -v -s basic_correctness/test_chunked_prefill.py
 
 - label: Core Test # 10min
-  mirror_hardwares: [amdexperimental, amdproduction]
+  mirror_hardwares: [amdexperimental]
   fast_check: true
   source_file_dependencies:
   - vllm/core
@@ -107,7 +107,7 @@ steps:
   commands:
   - pytest -v -s core
 
-- label: Entrypoints Test # 40min
+- label: Entrypoints Test (LLM) # 40min
   mirror_hardwares: [amdexperimental]
   working_dir: "/vllm-workspace/tests"
   fast_check: true
@@ -115,19 +115,28 @@ steps:
   source_file_dependencies:
   - vllm/
   - tests/entrypoints/llm
-  - tests/entrypoints/openai
-  - tests/entrypoints/test_chat_utils
   - tests/entrypoints/offline_mode
   commands:
   - export VLLM_WORKER_MULTIPROC_METHOD=spawn
-  - pytest -v -s entrypoints/llm --ignore=entrypoints/llm/test_lazy_outlines.py --ignore=entrypoints/llm/test_generate.py --ignore=entrypoints/llm/test_generate_multiple_loras.py --ignore=entrypoints/llm/test_guided_generate.py --ignore=entrypoints/llm/test_collective_rpc.py
+  - pytest -v -s entrypoints/llm --ignore=entrypoints/llm/test_lazy_outlines.py --ignore=entrypoints/llm/test_generate.py --ignore=entrypoints/llm/test_generate_multiple_loras.py --ignore=entrypoints/llm/test_collective_rpc.py
   - pytest -v -s entrypoints/llm/test_lazy_outlines.py # it needs a clean process
   - pytest -v -s entrypoints/llm/test_generate.py # it needs a clean process
   - pytest -v -s entrypoints/llm/test_generate_multiple_loras.py # it needs a clean process
-  - VLLM_USE_V1=0 pytest -v -s entrypoints/llm/test_guided_generate.py # it needs a clean process
+  - VLLM_USE_V1=0 pytest -v -s entrypoints/offline_mode # Needs to avoid interference with other tests
+
+- label: Entrypoints Test (API Server) # 40min
+  mirror_hardwares: [amdexperimental]
+  working_dir: "/vllm-workspace/tests"
+  fast_check: true
+  torch_nightly: true
+  source_file_dependencies:
+  - vllm/
+  - tests/entrypoints/openai
+  - tests/entrypoints/test_chat_utils
+  commands:
+  - export VLLM_WORKER_MULTIPROC_METHOD=spawn
   - pytest -v -s entrypoints/openai --ignore=entrypoints/openai/test_chat_with_tool_reasoning.py --ignore=entrypoints/openai/test_oot_registration.py --ignore=entrypoints/openai/test_tensorizer_entrypoint.py --ignore=entrypoints/openai/correctness/
   - pytest -v -s entrypoints/test_chat_utils.py
-  - VLLM_USE_V1=0 pytest -v -s entrypoints/offline_mode # Needs to avoid interference with other tests
 
 - label: Distributed Tests (4 GPUs) # 10min
   mirror_hardwares: [amdexperimental]
@@ -139,12 +148,14 @@ steps:
   - tests/distributed/test_utils
   - tests/distributed/test_pynccl
   - tests/distributed/test_events
-  - tests/spec_decode/e2e/test_integration_dist_tp4
   - tests/compile/test_basic_correctness
   - examples/offline_inference/rlhf.py
   - examples/offline_inference/rlhf_colocate.py
   - tests/examples/offline_inference/data_parallel.py
   - tests/v1/test_async_llm_dp.py
+  - tests/v1/test_external_lb_dp.py
+  - tests/v1/test_internal_lb_dp.py
+  - tests/v1/test_hybrid_lb_dp.py
   - tests/v1/engine/test_engine_core_client.py
   commands:
   # test with tp=2 and external_dp=2
@@ -153,14 +164,16 @@ steps:
   # test with tp=2 and pp=2
   - PP_SIZE=2 torchrun --nproc-per-node=4 distributed/test_torchrun_example.py
   # test with internal dp
-  - python3 ../examples/offline_inference/data_parallel.py
+  - python3 ../examples/offline_inference/data_parallel.py --enforce-eager
   - TP_SIZE=2 DP_SIZE=2 pytest -v -s v1/test_async_llm_dp.py
+  - TP_SIZE=2 DP_SIZE=2 pytest -v -s v1/test_external_lb_dp.py
+  - TP_SIZE=1 DP_SIZE=4 pytest -v -s v1/test_internal_lb_dp.py
+  - TP_SIZE=1 DP_SIZE=4 pytest -v -s v1/test_hybrid_lb_dp.py
   - pytest -v -s v1/engine/test_engine_core_client.py::test_kv_cache_events_dp
   - pytest -v -s distributed/test_utils.py
   - pytest -v -s compile/test_basic_correctness.py
   - pytest -v -s distributed/test_pynccl.py
   - pytest -v -s distributed/test_events.py
-  - pytest -v -s spec_decode/e2e/test_integration_dist_tp4.py
   # TODO: create a dedicated test section for multi-GPU example tests
   # when we have multiple distributed example tests
   - pushd ../examples/offline_inference
@@ -168,8 +181,25 @@ steps:
   - VLLM_ALLOW_INSECURE_SERIALIZATION=1 RAY_DEDUP_LOGS=0 python3 rlhf_colocate.py
   - popd
 
+- label: EPLB Algorithm Test
+  working_dir: "/vllm-workspace/tests"
+  source_file_dependencies:
+  - vllm/distributed/eplb
+  - tests/distributed/test_eplb_algo.py
+  commands:
+  - pytest -v -s distributed/test_eplb_algo.py
+
+- label: EPLB Execution Test # 5min
+  working_dir: "/vllm-workspace/tests"
+  num_gpus: 4
+  source_file_dependencies:
+  - vllm/distributed/eplb
+  - tests/distributed/test_eplb_execute.py
+  commands:
+  - pytest -v -s distributed/test_eplb_execute.py
+
 - label: Metrics, Tracing Test # 10min
-  mirror_hardwares: [amdexperimental, amdproduction]
+  mirror_hardwares: [amdexperimental]
   num_gpus: 2
   source_file_dependencies:
   - vllm/
@@ -177,13 +207,18 @@ steps:
   - tests/tracing
   commands:
   - pytest -v -s metrics
+  - "pip install \
+      'opentelemetry-sdk>=1.26.0' \
+      'opentelemetry-api>=1.26.0' \
+      'opentelemetry-exporter-otlp>=1.26.0' \
+      'opentelemetry-semantic-conventions-ai>=0.4.1'"
   - pytest -v -s tracing
 
 ##### fast check tests  #####
 #####  1 GPU test  #####
 
 - label: Regression Test # 5min
-  mirror_hardwares: [amdexperimental, amdproduction]
+  mirror_hardwares: [amdexperimental]
   source_file_dependencies:
   - vllm/
   - tests/test_regression
@@ -193,7 +228,7 @@ steps:
   working_dir: "/vllm-workspace/tests" # optional
 
 - label: Engine Test # 10min
-  mirror_hardwares: [amdexperimental, amdproduction]
+  mirror_hardwares: [amdexperimental]
   source_file_dependencies:
   - vllm/
   - tests/engine
@@ -218,10 +253,12 @@ steps:
     - pytest -v -s v1/engine
     - pytest -v -s v1/entrypoints
     - pytest -v -s v1/sample
+    - pytest -v -s v1/logits_processors
     - pytest -v -s v1/worker
     - pytest -v -s v1/structured_output
     - pytest -v -s v1/spec_decode
     - pytest -v -s v1/kv_connector/unit
+    - pytest -v -s v1/metrics
     - pytest -v -s v1/test_serial_utils.py
     - pytest -v -s v1/test_utils.py
     - pytest -v -s v1/test_oracle.py
@@ -230,7 +267,7 @@ steps:
     # VLLM_USE_FLASHINFER_SAMPLER or not on H100.
     - pytest -v -s v1/e2e
     # Integration test for streaming correctness (requires special branch).
-    - pip install -U git+https://github.com/robertgshaw2-neuralmagic/lm-evaluation-harness.git@streaming-api
+    - pip install -U git+https://github.com/robertgshaw2-redhat/lm-evaluation-harness.git@streaming-api
     - pytest -v -s entrypoints/openai/correctness/test_lmeval.py::test_lm_eval_accuracy_v1_engine
 
 - label: Examples Test # 25min
@@ -248,7 +285,7 @@ steps:
     - python3 offline_inference/llm_engine_example.py
     - python3 offline_inference/audio_language.py --seed 0
     - python3 offline_inference/vision_language.py --seed 0
-    - python3 offline_inference/vision_language_embedding.py --seed 0
+    - python3 offline_inference/vision_language_pooling.py --seed 0
     - python3 offline_inference/vision_language_multi_image.py --seed 0
     - VLLM_USE_V1=0 python3 others/tensorize_vllm_model.py --model facebook/opt-125m serialize --serialized-directory /tmp/ --suffix v1 && python3 others/tensorize_vllm_model.py --model facebook/opt-125m deserialize --path-to-tensors /tmp/vllm/facebook/opt-125m/v1/model.tensors
     - python3 offline_inference/encoder_decoder.py
@@ -259,13 +296,22 @@ steps:
     - VLLM_USE_V1=0 python3 offline_inference/profiling.py --model facebook/opt-125m run_num_steps --num-steps 2
 
 - label: Prefix Caching Test # 9min
-  mirror_hardwares: [amdexperimental, amdproduction]
+  mirror_hardwares: [amdexperimental]
   source_file_dependencies:
   - vllm/
   - tests/prefix_caching
   commands:
     - pytest -v -s prefix_caching
 
+
+- label: Platform Tests (CUDA)
+  mirror_hardwares: [amdexperimental]
+  source_file_dependencies:
+  - vllm/
+  - tests/cuda
+  commands:
+    - pytest -v -s cuda/test_cuda_context.py
+
 - label: Samplers Test # 36min
   mirror_hardwares: [amdexperimental]
   source_file_dependencies:
@@ -277,19 +323,8 @@ steps:
     - pytest -v -s samplers
     - VLLM_USE_FLASHINFER_SAMPLER=1 pytest -v -s samplers
 
-- label: Speculative decoding tests # 40min
-  mirror_hardwares: [amdexperimental]
-  source_file_dependencies:
-  - vllm/spec_decode
-  - tests/spec_decode
-  - vllm/model_executor/models/eagle.py
-  commands:
-    - pytest -v -s spec_decode/e2e/test_multistep_correctness.py
-    - VLLM_ATTENTION_BACKEND=FLASH_ATTN pytest -v -s spec_decode --ignore=spec_decode/e2e/test_multistep_correctness.py --ignore=spec_decode/e2e/test_mtp_correctness.py
-    - pytest -v -s spec_decode/e2e/test_eagle_correctness.py
-
 - label: LoRA Test %N # 15min each
-  mirror_hardwares: [amdexperimental, amdproduction]
+  mirror_hardwares: [amdexperimental]
   source_file_dependencies:
   - vllm/lora
   - tests/lora
@@ -297,7 +332,7 @@ steps:
   parallelism: 4
 
 - label: PyTorch Compilation Unit Tests
-  mirror_hardwares: [amdexperimental, amdproduction]
+  mirror_hardwares: [amdexperimental]
   torch_nightly: true
   source_file_dependencies:
     - vllm/
@@ -305,12 +340,14 @@ steps:
   commands:
     - pytest -v -s compile/test_pass_manager.py
     - pytest -v -s compile/test_fusion.py
+    - pytest -v -s compile/test_fusion_attn.py
     - pytest -v -s compile/test_silu_mul_quant_fusion.py
     - pytest -v -s compile/test_sequence_parallelism.py
     - pytest -v -s compile/test_async_tp.py
+    - pytest -v -s compile/test_fusion_all_reduce.py
 
 - label: PyTorch Fullgraph Smoke Test # 9min
-  mirror_hardwares: [amdexperimental, amdproduction]
+  mirror_hardwares: [amdexperimental]
   torch_nightly: true
   source_file_dependencies:
   - vllm/
@@ -323,7 +360,7 @@ steps:
   - pytest -v -s compile/piecewise/test_full_cudagraph.py
 
 - label: PyTorch Fullgraph Test # 18min
-  mirror_hardwares: [amdexperimental, amdproduction]
+  mirror_hardwares: [amdexperimental]
   torch_nightly: true
   source_file_dependencies:
   - vllm/
@@ -332,7 +369,7 @@ steps:
   - pytest -v -s compile/test_full_graph.py
 
 - label: Kernels Core Operation Test
-  mirror_hardwares: [amdexperimental, amdproduction]
+  mirror_hardwares: [amdexperimental]
   source_file_dependencies:
   - csrc/
   - tests/kernels/core
@@ -340,7 +377,7 @@ steps:
     - pytest -v -s kernels/core
 
 - label: Kernels Attention Test %N
-  mirror_hardwares: [amdexperimental, amdproduction]
+  mirror_hardwares: [amdexperimental]
   source_file_dependencies:
   - csrc/attention/
   - vllm/attention
@@ -351,23 +388,25 @@ steps:
   parallelism: 2
 
 - label: Kernels Quantization Test %N
-  mirror_hardwares: [amdexperimental, amdproduction]
+  mirror_hardwares: [amdexperimental]
   source_file_dependencies:
   - csrc/quantization/
   - vllm/model_executor/layers/quantization
   - tests/kernels/quantization
   commands:
-    - pytest -v -s kernels/quantization  --shard-id=$$BUILDKITE_PARALLEL_JOB --num-shards=$$BUILDKITE_PARALLEL_JOB_COUNT
+    - pytest -v -s kernels/quantization --shard-id=$$BUILDKITE_PARALLEL_JOB --num-shards=$$BUILDKITE_PARALLEL_JOB_COUNT
   parallelism: 2
 
-- label: Kernels MoE Test
+- label: Kernels MoE Test %N
   mirror_hardwares: [amdexperimental]
   source_file_dependencies:
+  - csrc/quantization/cutlass_w8a8/moe/
   - csrc/moe/
   - tests/kernels/moe
   - vllm/model_executor/layers/fused_moe/
   commands:
-    - pytest -v -s kernels/moe
+    - pytest -v -s kernels/moe --shard-id=$$BUILDKITE_PARALLEL_JOB --num-shards=$$BUILDKITE_PARALLEL_JOB_COUNT
+  parallelism: 2
 
 - label: Kernels Mamba Test
   mirror_hardwares: [amdexperimental]
@@ -378,8 +417,7 @@ steps:
     - pytest -v -s kernels/mamba
 
 - label: Tensorizer Test # 11min
-  mirror_hardwares: [amdexperimental, amdproduction]
-  soft_fail: true
+  mirror_hardwares: [amdexperimental]
   source_file_dependencies:
   - vllm/model_executor/model_loader
   - tests/tensorizer_loader
@@ -391,8 +429,7 @@ steps:
     - pytest -v -s entrypoints/openai/test_tensorizer_entrypoint.py
 
 - label: Model Executor Test
-  mirror_hardwares: [amdexperimental, amdproduction]
-  soft_fail: true
+  mirror_hardwares: [amdexperimental]
   source_file_dependencies:
   - vllm/model_executor
   - tests/model_executor
@@ -402,7 +439,7 @@ steps:
     - pytest -v -s model_executor
 
 - label: Benchmarks # 9min
-  mirror_hardwares: [amdexperimental, amdproduction]
+  mirror_hardwares: [amdexperimental]
   working_dir: "/vllm-workspace/.buildkite"
   source_file_dependencies:
   - benchmarks/
@@ -410,7 +447,7 @@ steps:
   - bash scripts/run-benchmarks.sh
 
 - label: Benchmarks CLI Test # 10min
-  mirror_hardwares: [amdexperimental, amdproduction]
+  mirror_hardwares: [amdexperimental]
   source_file_dependencies:
   - vllm/
   - tests/benchmarks/
@@ -470,7 +507,7 @@ steps:
 #####  models test  #####
 
 - label: Basic Models Test # 24min
-  mirror_hardwares: [amdexperimental, amdproduction]
+  mirror_hardwares: [amdexperimental]
   torch_nightly: true
   source_file_dependencies:
   - vllm/
@@ -489,11 +526,22 @@ steps:
   - vllm/
   - tests/models/language
   commands:
-    # Install causal-conv1d for plamo2 models here, as it is not compatible with pip-compile.
-    - pip install 'git+https://github.com/Dao-AILab/causal-conv1d@v1.5.0.post8'
     - pip freeze | grep -E 'torch'
     - pytest -v -s models/language -m core_model
 
+- label: Language Models Test (Hybrid) # 35 min
+  mirror_hardwares: [amdexperimental]
+  torch_nightly: true
+  source_file_dependencies:
+  - vllm/
+  - tests/models/language/generation
+  commands:
+    # Install fast path packages for testing against transformers
+    # Note: also needed to run plamo2 model in vLLM
+    - uv pip install --system --no-build-isolation 'git+https://github.com/state-spaces/mamba@v2.2.5'
+    - uv pip install --system --no-build-isolation 'git+https://github.com/Dao-AILab/causal-conv1d@v1.5.2'
+    - pytest -v -s models/language/generation -m hybrid_model
+
 - label: Language Models Test (Extended Generation) # 1hr20min
   mirror_hardwares: [amdexperimental]
   optional: true
@@ -503,7 +551,7 @@ steps:
   commands:
     # Install causal-conv1d for plamo2 models here, as it is not compatible with pip-compile.
     - pip install 'git+https://github.com/Dao-AILab/causal-conv1d@v1.5.0.post8'
-    - pytest -v -s models/language/generation -m 'not core_model'
+    - pytest -v -s models/language/generation -m '(not core_model) and (not hybrid_model)'
 
 - label: Language Models Test (Extended Pooling)  # 36min
   mirror_hardwares: [amdexperimental]
@@ -524,7 +572,8 @@ steps:
     - pip install git+https://github.com/TIGER-AI-Lab/Mantis.git
     - pip freeze | grep -E 'torch'
     - pytest -v -s models/multimodal/processing
-    - pytest -v -s --ignore models/multimodal/generation/test_whisper.py models/multimodal -m core_model
+    - pytest -v -s --ignore models/multimodal/generation/test_whisper.py --ignore models/multimodal/test_tensor_schema.py models/multimodal -m core_model
+    - pytest -v -s models/multimodal/test_tensor_schema.py -m core_model  # Needs mp_method="spawn"
     - cd .. && pytest -v -s tests/models/multimodal/generation/test_whisper.py -m core_model  # Otherwise, mp_method="spawn" doesn't work
 
 - label: Multi-Modal Models Test (Extended) 1
@@ -548,7 +597,7 @@ steps:
     - pytest -v -s models/multimodal/generation/test_common.py -m 'split(group=0) and not core_model'
 
 - label: Multi-Modal Models Test (Extended) 3
-  mirror_hardwares: [amdexperimental, amdproduction]
+  mirror_hardwares: [amdexperimental]
   optional: true
   source_file_dependencies:
   - vllm/
@@ -558,7 +607,7 @@ steps:
     - pytest -v -s models/multimodal/generation/test_common.py -m 'split(group=1) and not core_model'
 
 - label: Quantized Models Test
-  mirror_hardwares: [amdexperimental, amdproduction]
+  mirror_hardwares: [amdexperimental]
   source_file_dependencies:
   - vllm/model_executor/layers/quantization
   - tests/models/quantization
@@ -567,7 +616,7 @@ steps:
 
 # This test is used only in PR development phase to test individual models and should never run on main
 - label: Custom Models Test
-  mirror_hardwares: [amdexperimental, amdproduction]
+  mirror_hardwares: [amdexperimental]
   optional: true
   commands:
     - echo 'Testing custom models...'
@@ -575,11 +624,53 @@ steps:
     # e.g. pytest -v -s models/encoder_decoder/vision_language/test_mllama.py
     # *To avoid merge conflicts, remember to REMOVE (not just comment out) them before merging the PR*
 
+- label: Transformers Nightly Models Test
+  working_dir: "/vllm-workspace/"
+  optional: true
+  commands:
+    - pip install --upgrade git+https://github.com/huggingface/transformers
+    - pytest -v -s tests/models/test_initialization.py
+    - pytest -v -s tests/models/multimodal/processing/
+    - pytest -v -s tests/models/multimodal/test_mapping.py
+    - python3 examples/offline_inference/basic/chat.py
+    - python3 examples/offline_inference/audio_language.py --model-type whisper
+    - python3 examples/offline_inference/vision_language.py --model-type qwen2_5_vl
+
+- label: Blackwell Test
+  working_dir: "/vllm-workspace/"
+  gpu: b200
+  # optional: true
+  source_file_dependencies:
+  - csrc/quantization/fp4/
+  - csrc/attention/mla/
+  - csrc/quantization/cutlass_w8a8/moe/
+  - vllm/model_executor/layers/fused_moe/cutlass_moe.py
+  - vllm/model_executor/layers/fused_moe/flashinfer_cutlass_moe.py
+  - vllm/model_executor/layers/fused_moe/flashinfer_cutlass_prepare_finalize.py
+  - vllm/v1/attention/backends/flashinfer.py
+  - vllm/compilation/fusion.py
+  commands:
+    - nvidia-smi
+    - python3 examples/offline_inference/basic/chat.py
+    # Attention
+    # num_heads2 broken by https://github.com/flashinfer-ai/flashinfer/issues/1353
+    - pytest -v -s tests/kernels/attention/test_flashinfer.py -k 'not num_heads2'
+    - pytest -v -s tests/kernels/attention/test_flashinfer_trtllm_attention.py
+    - pytest -v -s tests/kernels/test_cutlass_mla_decode.py
+    # Quantization
+    - pytest -v -s tests/kernels/quantization/test_cutlass_scaled_mm.py -k 'fp8'
+    - pytest -v -s tests/kernels/quantization/test_nvfp4_quant.py
+    - pytest -v -s tests/kernels/quantization/test_nvfp4_scaled_mm.py
+    - pytest -v -s tests/kernels/quantization/test_flashinfer_nvfp4_scaled_mm.py
+    - pytest -v -s tests/kernels/moe/test_nvfp4_moe.py
+    # Fusion
+    - pytest -v -s tests/compile/test_fusion_all_reduce.py
+
 #####  1 GPU test  #####
 #####  multi gpus test  #####
 
 - label: Distributed Comm Ops Test # 7min
-  mirror_hardwares: [amdexperimental, amdproduction]
+  mirror_hardwares: [amdexperimental]
   working_dir: "/vllm-workspace/tests"
   num_gpus: 2
   source_file_dependencies:
@@ -600,13 +691,18 @@ steps:
   - vllm/executor/
   - vllm/model_executor/models/
   - tests/distributed/
+  - tests/examples/offline_inference/data_parallel.py
   commands:
   - # the following commands are for the first node, with ip 192.168.10.10 (ray environment already set up)
     - VLLM_TEST_SAME_HOST=0 torchrun --nnodes 2 --nproc-per-node=2 --rdzv_backend=c10d --rdzv_endpoint=192.168.10.10 distributed/test_same_node.py | grep 'Same node test passed'
+    - NUM_NODES=2 torchrun --nnodes 2 --nproc-per-node=2 --rdzv_backend=c10d --rdzv_endpoint=192.168.10.10 distributed/test_node_count.py | grep 'Node count test passed'
+    - python3 ../examples/offline_inference/data_parallel.py --dp-size=2 --tp-size=1 --node-size=2 --node-rank=0 --master-addr=192.168.10.10 --master-port=12345 --enforce-eager --trust-remote-code
     - VLLM_MULTI_NODE=1 pytest -v -s distributed/test_multi_node_assignment.py
     - VLLM_MULTI_NODE=1 pytest -v -s distributed/test_pipeline_parallel.py
   - # the following commands are for the second node, with ip 192.168.10.11 (ray environment already set up)
     - VLLM_TEST_SAME_HOST=0 torchrun --nnodes 2 --nproc-per-node=2 --rdzv_backend=c10d --rdzv_endpoint=192.168.10.10 distributed/test_same_node.py | grep 'Same node test passed'
+    - NUM_NODES=2 torchrun --nnodes 2 --nproc-per-node=2 --rdzv_backend=c10d --rdzv_endpoint=192.168.10.10 distributed/test_node_count.py | grep 'Node count test passed'
+    - python3 ../examples/offline_inference/data_parallel.py --dp-size=2 --tp-size=1 --node-size=2 --node-rank=1 --master-addr=192.168.10.10 --master-port=12345 --enforce-eager --trust-remote-code
 
 - label: Distributed Tests (2 GPUs) # 40min
   mirror_hardwares: [amdexperimental]
@@ -624,10 +720,12 @@ steps:
   - vllm/worker/model_runner.py
   - entrypoints/llm/test_collective_rpc.py
   - tests/v1/test_async_llm_dp.py
+  - tests/v1/test_external_lb_dp.py
   - tests/v1/entrypoints/openai/test_multi_api_servers.py
   - vllm/v1/engine/
   commands:
   - TP_SIZE=1 DP_SIZE=2 pytest -v -s v1/test_async_llm_dp.py
+  - TP_SIZE=1 DP_SIZE=2 pytest -v -s v1/test_external_lb_dp.py
   - DP_SIZE=2 pytest -v -s v1/entrypoints/openai/test_multi_api_servers.py
   - pytest -v -s entrypoints/llm/test_collective_rpc.py
   - pytest -v -s ./compile/test_basic_correctness.py
@@ -642,10 +740,9 @@ steps:
   - pytest -v -s distributed/test_sequence_parallel.py
   # this test fails consistently.
   # TODO: investigate and fix
-  # - pytest -v -s spec_decode/e2e/test_integration_dist_tp2.py
   - VLLM_USE_V1=0 CUDA_VISIBLE_DEVICES=0,1 pytest -v -s test_sharded_state_loader.py
-  - VLLM_USE_V1=0 CUDA_VISIBLE_DEVICES=0,1 pytest -v -s kv_transfer/test_disagg.py
   - CUDA_VISIBLE_DEVICES=0,1 pytest -v -s v1/shutdown
+  - pytest -v -s models/multimodal/generation/test_maverick.py
 
 - label: Plugin Tests (2 GPUs) # 40min
   mirror_hardwares: [amdexperimental]
@@ -668,29 +765,8 @@ steps:
   - pytest -v -s models/test_oot_registration.py # it needs a clean process
   - pytest -v -s plugins/lora_resolvers # unit tests for in-tree lora resolver plugins
 
-- label: Multi-step Tests (4 GPUs) # 36min
-  mirror_hardwares: [amdexperimental]
-  working_dir: "/vllm-workspace/tests"
-  num_gpus: 4
-  source_file_dependencies:
-  - vllm/model_executor/layers/sampler.py
-  - vllm/sequence.py
-  - vllm/worker/worker_base.py
-  - vllm/worker/worker.py
-  - vllm/worker/multi_step_worker.py
-  - vllm/worker/model_runner_base.py
-  - vllm/worker/model_runner.py
-  - vllm/worker/multi_step_model_runner.py
-  - vllm/engine
-  - tests/multi_step
-  commands:
-  # this test is quite flaky
-  # TODO: investigate and fix.
-  # - pytest -v -s multi_step/test_correctness_async_llm.py
-  - pytest -v -s multi_step/test_correctness_llm.py
-
 - label: Pipeline Parallelism Test # 45min
-  mirror_hardwares: [amdexperimental, amdproduction]
+  mirror_hardwares: [amdexperimental]
   working_dir: "/vllm-workspace/tests"
   num_gpus: 4
   source_file_dependencies:
@@ -704,7 +780,7 @@ steps:
   - pytest -v -s distributed/test_pipeline_parallel.py
 
 - label: LoRA TP Test (Distributed)
-  mirror_hardwares: [amdexperimental, amdproduction]
+  mirror_hardwares: [amdexperimental]
   num_gpus: 4
   source_file_dependencies:
   - vllm/lora
@@ -717,6 +793,7 @@ steps:
     # requires multi-GPU testing for validation.
     - pytest -v -s -x lora/test_chatglm3_tp.py
     - pytest -v -s -x lora/test_llama_tp.py
+    - pytest -v -s -x lora/test_multi_loras_with_tp.py
 
 
 - label: Weight Loading Multiple GPU Test  # 33min
@@ -730,7 +807,7 @@ steps:
     - bash weight_loading/run_model_weight_loading_test.sh -c weight_loading/models.txt
 
 - label: Weight Loading Multiple GPU Test - Large Models # optional
-  mirror_hardwares: [amdexperimental] 
+  mirror_hardwares: [amdexperimental]
   working_dir: "/vllm-workspace/tests"
   num_gpus: 2
   gpu: a100

.gemini/config.yaml

@@ -0,0 +1,6 @@
+# https://developers.google.com/gemini-code-assist/docs/customize-gemini-behavior-github
+have_fun: false  # Just review the code
+code_review:
+  comment_severity_threshold: HIGH  # Reduce quantity of comments
+  pull_request_opened:
+    summary: false  # Don't summarize the PR in a separate comment

.github/CODEOWNERS

@@ -9,14 +9,18 @@
 /vllm/worker/worker_base.py @zhuohan123 @youkaichao @alexm-redhat @comaniac @njhill
 /vllm/worker/worker.py @zhuohan123 @youkaichao @alexm-redhat @comaniac @njhill
 /vllm/model_executor/layers/sampler.py @zhuohan123 @youkaichao @alexm-redhat @comaniac @njhill
-/vllm/model_executor/layers/quantization @mgoin @robertgshaw2-redhat @tlrmchlsmth
-/vllm/model_executor/guided_decoding @mgoin @russellb @aarnphm
+/vllm/model_executor/layers/quantization @mgoin @robertgshaw2-redhat @tlrmchlsmth @yewentao256
 /vllm/multimodal @DarkLight1337 @ywang96
 /vllm/vllm_flash_attn @LucasWilkinson
 /vllm/lora @jeejeelee
 /vllm/reasoning @aarnphm
 /vllm/entrypoints @aarnphm
-CMakeLists.txt @tlrmchlsmth
+/vllm/compilation @zou3519 @youkaichao @ProExpertProg
+CMakeLists.txt @tlrmchlsmth @LucasWilkinson
+
+# Any change to the VllmConfig changes can have a large user-facing impact,
+# so spam a lot of people
+/vllm/config @simon-mo @WoosukKwon @youkaichao @robertgshaw2-redhat @mgoin @tlrmchlsmth @houseroad @hmellor @yewentao256 @ProExpertProg
 
 # vLLM V1
 /vllm/v1 @WoosukKwon @robertgshaw2-redhat @njhill @ywang96 @comaniac @alexm-redhat
@@ -30,21 +34,41 @@ CMakeLists.txt @tlrmchlsmth
 /tests/distributed/test_pipeline_parallel.py @youkaichao
 /tests/distributed/test_same_node.py @youkaichao
 /tests/entrypoints @DarkLight1337 @robertgshaw2-redhat @simon-mo @aarnphm
-/tests/entrypoints/llm/test_guided_generate.py @mgoin @russellb @aarnphm
-/tests/kernels @tlrmchlsmth @WoosukKwon
-/tests/model_executor/test_guided_processors.py @mgoin @russellb
+/tests/kernels @tlrmchlsmth @WoosukKwon @yewentao256
 /tests/models @DarkLight1337 @ywang96
-/tests/multi_step @alexm-redhat @comaniac
 /tests/multimodal @DarkLight1337 @ywang96
 /tests/prefix_caching @comaniac @KuntaiDu
-/tests/quantization @mgoin @robertgshaw2-redhat
-/tests/spec_decode @njhill @LiuXiaoxuanPKU
+/tests/quantization @mgoin @robertgshaw2-redhat @yewentao256
 /tests/test_inputs.py @DarkLight1337 @ywang96
 /tests/v1/entrypoints/llm/test_struct_output_generate.py @mgoin @russellb @aarnphm
 /tests/v1/structured_output @mgoin @russellb @aarnphm
-/tests/weight_loading @mgoin @youkaichao
+/tests/weight_loading @mgoin @youkaichao @yewentao256
 /tests/lora @jeejeelee
 
 # Docs
 /docs @hmellor
 mkdocs.yaml @hmellor
+
+# CPU
+/vllm/v1/worker/^cpu @bigPYJ1151
+/csrc/cpu @bigPYJ1151
+/vllm/platforms/cpu.py @bigPYJ1151
+/cmake/cpu_extension.cmake @bigPYJ1151
+/docker/Dockerfile.cpu @bigPYJ1151
+
+# Intel GPU
+/vllm/v1/worker/^xpu @jikunshang
+/vllm/platforms/xpu.py @jikunshang
+/docker/Dockerfile.xpu @jikunshang
+
+# Qwen-specific files
+/vllm/attention/backends/dual_chunk_flash_attn.py @sighingnow
+/vllm/model_executor/models/qwen* @sighingnow
+
+# Mistral-specific files
+/vllm/model_executor/models/mistral*.py @patrickvonplaten
+/vllm/model_executor/models/mixtral*.py @patrickvonplaten
+/vllm/model_executor/models/voxtral*.py @patrickvonplaten
+/vllm/model_executor/models/pixtral*.py @patrickvonplaten
+/vllm/transformers_utils/configs/mistral.py @patrickvonplaten
+/vllm/transformers_utils/tokenizers/mistral.py @patrickvonplaten

.github/ISSUE_TEMPLATE/750-RFC.yml

@@ -46,7 +46,7 @@ body:
 - type: markdown
   attributes:
     value: >
-      Thanks for contributing 🎉!
+      Thanks for contributing 🎉! The vLLM core team hosts a biweekly RFC review session at 9:30AM Pacific Time, while most RFCs can be discussed online, you can optionally sign up for a slot to discuss your RFC online [here](https://docs.google.com/document/d/1CiLVBZeIVfR7_PNAKVSusxpceywkoOOB78qoWqHvSZc/edit).
 - type: checkboxes
   id: askllm
   attributes:

.github/PULL_REQUEST_TEMPLATE.md

@@ -1,10 +1,5 @@
-## Essential Elements of an Effective PR Description Checklist
-- [ ] The purpose of the PR, such as "Fix some issue (link existing issues this PR will resolve)".
-- [ ] The test plan, such as providing test command.
-- [ ] The test results, such as pasting the results comparison before and after, or e2e results
-- [ ] (Optional) The necessary documentation update, such as updating `supported_models.md` and `examples` for a new model.
-
-PLEASE FILL IN THE PR DESCRIPTION HERE ENSURING ALL CHECKLIST ITEMS ABOVE HAVE BEEN CONSIDERED.
+<!-- markdownlint-disable -->
+PLEASE FILL IN THE PR DESCRIPTION HERE ENSURING ALL CHECKLIST ITEMS (AT THE BOTTOM) HAVE BEEN CONSIDERED.
 
 ## Purpose
 
@@ -14,5 +9,14 @@ PLEASE FILL IN THE PR DESCRIPTION HERE ENSURING ALL CHECKLIST ITEMS ABOVE HAVE B
 
 ## (Optional) Documentation Update
 
-<!--- pyml disable-next-line no-emphasis-as-heading -->
+---
+<details>
+<summary> Essential Elements of an Effective PR Description Checklist </summary>
+
+- [ ] The purpose of the PR, such as "Fix some issue (link existing issues this PR will resolve)".
+- [ ] The test plan, such as providing test command.
+- [ ] The test results, such as pasting the results comparison before and after, or e2e results
+- [ ] (Optional) The necessary documentation update, such as updating `supported_models.md` and `examples` for a new model.
+</details>
+
 **BEFORE SUBMITTING, PLEASE READ <https://docs.vllm.ai/en/latest/contributing>** (anything written below this line will be removed by GitHub Actions)

.github/mergify.yml

@@ -27,6 +27,22 @@ pull_request_rules:
       add:
         - ci/build
 
+- name: label-deepseek
+  description: Automatically apply deepseek label
+  conditions:
+    - or:
+      - files~=^examples/.*deepseek.*\.py
+      - files~=^tests/.*deepseek.*\.py
+      - files~=^vllm/entrypoints/openai/tool_parsers/.*deepseek.*\.py
+      - files~=^vllm/model_executor/models/.*deepseek.*\.py
+      - files~=^vllm/reasoning/.*deepseek.*\.py
+      - files~=^vllm/transformers_utils/.*deepseek.*\.py
+      - title~=(?i)DeepSeek
+  actions:
+    label:
+      add:
+        - deepseek
+
 - name: label-frontend
   description: Automatically apply frontend label
   conditions:
@@ -45,6 +61,7 @@ pull_request_rules:
       - files~=^vllm/entrypoints/openai/tool_parsers/llama.*\.py
       - files~=^vllm/model_executor/models/.*llama.*\.py
       - files~=^vllm/transformers_utils/configs/.*llama.*\.py
+      - title~=(?i)llama
   actions:
     label:
       add:
@@ -57,14 +74,84 @@ pull_request_rules:
       - files~=^vllm/multimodal/
       - files~=^tests/multimodal/
       - files~=^tests/models/multimodal/
-      - files~=^tests/models/*/audio_language/
-      - files~=^tests/models/*/vision_language/
       - files=tests/models/test_vision.py
   actions:
     label:
       add:
         - multi-modality
 
+- name: label-new-model
+  description: Automatically apply new-model label
+  conditions:
+    - and:
+      - files~=^vllm/model_executor/models/
+      - files=vllm/model_executor/models/registry.py
+  actions:
+    label:
+      add:
+        - new-model
+
+- name: label-performance
+  description: Automatically apply performance label
+  conditions:
+    - or:
+      - files~=^benchmarks/
+      - files~=^vllm/benchmarks/
+      - files~=^tests/benchmarks/
+      - files~=^\.buildkite/nightly-benchmarks/
+  actions:
+    label:
+      add:
+        - performance
+
+- name: label-qwen
+  description: Automatically apply qwen label
+  conditions:
+    - or:
+      - files~=^examples/.*qwen.*\.py
+      - files~=^tests/.*qwen.*\.py
+      - files~=^vllm/model_executor/models/.*qwen.*\.py
+      - files~=^vllm/reasoning/.*qwen.*\.py
+      - title~=(?i)Qwen
+  actions:
+    label:
+      add:
+        - qwen
+
+- name: label-gpt-oss
+  description: Automatically apply gpt-oss label
+  conditions:
+    - or:
+      - files~=^examples/.*gpt[-_]?oss.*\.py
+      - files~=^tests/.*gpt[-_]?oss.*\.py
+      - files~=^vllm/model_executor/models/.*gpt[-_]?oss.*\.py
+      - files~=^vllm/model_executor/layers/.*gpt[-_]?oss.*\.py
+      - title~=(?i)gpt[-_]?oss
+  actions:
+    label:
+      add:
+        - gpt-oss
+
+- name: label-rocm
+  description: Automatically apply rocm label
+  conditions:
+    - or:
+      - files~=^csrc/rocm/
+      - files~=^docker/Dockerfile.rocm
+      - files~=^requirements/rocm.*\.txt
+      - files~=^vllm/attention/backends/rocm.*\.py
+      - files~=^vllm/attention/ops/rocm.*\.py
+      - files~=^vllm/model_executor/layers/fused_moe/rocm.*\.py
+      - files~=^vllm/v1/attention/backends/mla/rocm.*\.py
+      - files~=^tests/kernels/.*_rocm.*\.py
+      - files=vllm/platforms/rocm.py
+      - title~=(?i)AMD
+      - title~=(?i)ROCm
+  actions:
+    label:
+      add:
+        - rocm
+
 - name: label-structured-output
   description: Automatically apply structured-output label
   conditions:
@@ -76,9 +163,6 @@ pull_request_rules:
       - files=examples/offline_inference/structured_outputs.py
       - files=examples/online_serving/openai_chat_completion_structured_outputs.py
       - files=examples/online_serving/openai_chat_completion_structured_outputs_with_reasoning.py
-      - files~=^vllm/model_executor/guided_decoding/
-      - files=tests/model_executor/test_guided_processors.py
-      - files=tests/entrypoints/llm/test_guided_generate.py
       - files~=^tests/v1/structured_output/
       - files=tests/v1/entrypoints/llm/test_guided_generate.py
       - files~=^vllm/v1/structured_output/
@@ -91,9 +175,12 @@ pull_request_rules:
   description: Automatically apply speculative-decoding label
   conditions:
     - or:
-      - files~=^vllm/spec_decode/
-      - files=vllm/model_executor/layers/spec_decode_base_sampler.py
-      - files~=^tests/spec_decode/
+      - files~=^vllm/v1/spec_decode/
+      - files~=^tests/v1/spec_decode/
+      - files~=^examples/.*(spec_decode|mlpspeculator|eagle|speculation).*\.py
+      - files~=^vllm/model_executor/models/.*eagle.*\.py
+      - files=vllm/model_executor/models/mlp_speculator.py
+      - files~=^vllm/transformers_utils/configs/(eagle|medusa|mlp_speculator)\.py
   actions:
     label:
       add:

.github/scripts/cleanup_pr_body.sh

@@ -15,11 +15,11 @@ NEW=/tmp/new_pr_body.txt
 gh pr view --json body --template "{{.body}}" "${PR_NUMBER}" > "${OLD}"
 cp "${OLD}" "${NEW}"
 
-# Remove "FIX #xxxx (*link existing issues this PR will resolve*)"
-sed -i '/FIX #xxxx.*$/d' "${NEW}"
+# Remove markdown comments (like the <!-- markdownlint-disable --> at the start)
+sed -i '/<!--.*-->$/d' "${NEW}"
 
-# Remove "FILL IN THE PR DESCRIPTION HERE"
-sed -i '/FILL IN THE PR DESCRIPTION HERE/d' "${NEW}"
+# Remove "PLEASE FILL IN THE PR DESCRIPTION HERE ENSURING ALL CHECKLIST ITEMS (AT THE BOTTOM) HAVE BEEN CONSIDERED."
+sed -i '/PLEASE FILL IN THE PR DESCRIPTION HERE.*$/d' "${NEW}"
 
 # Remove all lines after and including "**BEFORE SUBMITTING, PLEASE READ THE CHECKLIST BELOW AND FILL IN THE DESCRIPTION ABOVE**"
 sed -i '/\*\*BEFORE SUBMITTING, PLEASE READ.*\*\*/,$d' "${NEW}"

.github/workflows/lint-and-deploy.yaml

@@ -2,6 +2,10 @@ name: Lint and Deploy Charts
 
 on: pull_request
 
+concurrency:
+  group: ${{ github.workflow }}-${{ github.ref }}
+  cancel-in-progress: true
+
 permissions:
   contents: read
 
@@ -68,7 +72,7 @@ jobs:
           export AWS_ACCESS_KEY_ID=minioadmin
           export AWS_SECRET_ACCESS_KEY=minioadmin
           sleep 30 && kubectl -n ns-vllm logs -f "$(kubectl -n ns-vllm get pods | awk '/deployment/ {print $1;exit}')" &
-          helm install --wait --wait-for-jobs --timeout 5m0s --debug --create-namespace --namespace=ns-vllm test-vllm examples/online_serving/chart-helm -f examples/online_serving/chart-helm/values.yaml --set secrets.s3endpoint=http://minio:9000 --set secrets.s3bucketname=testbucket --set secrets.s3accesskeyid=$AWS_ACCESS_KEY_ID --set secrets.s3accesskey=$AWS_SECRET_ACCESS_KEY --set resources.requests.cpu=1 --set resources.requests.memory=4Gi --set resources.limits.cpu=2 --set resources.limits.memory=5Gi --set image.env[0].name=VLLM_CPU_KVCACHE_SPACE --set image.env[1].name=VLLM_LOGGING_LEVEL --set-string image.env[0].value="1" --set-string image.env[1].value="DEBUG" --set-string extraInit.s3modelpath="opt-125m/" --set-string 'resources.limits.nvidia\.com/gpu=0' --set-string 'resources.requests.nvidia\.com/gpu=0' --set-string image.repository="vllm-cpu-env"
+          helm install --wait --wait-for-jobs --timeout 5m0s --debug --create-namespace --namespace=ns-vllm test-vllm examples/online_serving/chart-helm -f examples/online_serving/chart-helm/values.yaml --set secrets.s3endpoint=http://minio:9000 --set secrets.s3bucketname=testbucket --set secrets.s3accesskeyid=$AWS_ACCESS_KEY_ID --set secrets.s3accesskey=$AWS_SECRET_ACCESS_KEY --set resources.requests.cpu=1 --set resources.requests.memory=4Gi --set resources.limits.cpu=2 --set resources.limits.memory=5Gi --set image.env[0].name=VLLM_CPU_KVCACHE_SPACE --set image.env[1].name=VLLM_LOGGING_LEVEL --set image.env[2].name=VLLM_CPU_CI_ENV --set-string image.env[0].value="1" --set-string image.env[1].value="DEBUG" --set-string image.env[2].value="1" --set-string extraInit.s3modelpath="opt-125m/" --set-string 'resources.limits.nvidia\.com/gpu=0' --set-string 'resources.requests.nvidia\.com/gpu=0' --set-string image.repository="vllm-cpu-env"
 
       - name: curl test
         run: |

.github/workflows/matchers/markdownlint.json

@@ -0,0 +1,17 @@
+{
+  "problemMatcher": [
+    {
+      "owner": "markdownlint",
+      "pattern": [
+        {
+          "regexp": "^([^:]*):(\\d+):?(\\d+)?\\s([\\w-\\/]*)\\s(.*)$",
+          "file": 1,
+          "line": 2,
+          "column": 3,
+          "code": 4,
+          "message": 5
+        }
+      ]
+    }
+  ]
+}

.github/workflows/pre-commit.yml

@@ -5,6 +5,10 @@ on:
   push:
     branches: [main]
 
+concurrency:
+  group: ${{ github.workflow }}-${{ github.ref }}
+  cancel-in-progress: ${{ github.event_name == 'pull_request' }}
+
 permissions:
   contents: read
 
@@ -17,6 +21,7 @@ jobs:
       with:
         python-version: "3.12"
     - run: echo "::add-matcher::.github/workflows/matchers/actionlint.json"
+    - run: echo "::add-matcher::.github/workflows/matchers/markdownlint.json"
     - run: echo "::add-matcher::.github/workflows/matchers/mypy.json"
     - uses: pre-commit/action@2c7b3805fd2a0fd8c1884dcaebf91fc102a13ecd # v3.0.1
       with:

.github/workflows/scripts/build.sh

@@ -15,7 +15,6 @@ $python_executable -m pip install -r requirements/build.txt -r requirements/cuda
 export MAX_JOBS=1
 # Make sure release wheels are built for the following architectures
 export TORCH_CUDA_ARCH_LIST="7.0 7.5 8.0 8.6 8.9 9.0+PTX"
-export VLLM_FA_CMAKE_GPU_ARCHES="80-real;90-real"
 
 bash tools/check_repo.sh
 

.gitignore

@@ -4,6 +4,9 @@
 # vllm-flash-attn built from source
 vllm/vllm_flash_attn/*
 
+# triton jit 
+.triton
+
 # Byte-compiled / optimized / DLL files
 __pycache__/
 *.py[cod]
@@ -146,7 +149,9 @@ venv.bak/
 
 # mkdocs documentation
 /site
-docs/examples
+docs/argparse
+docs/examples/*
+!docs/examples/README.md
 
 # mypy
 .mypy_cache/
@@ -200,5 +205,8 @@ benchmarks/**/*.json
 actionlint
 shellcheck*/
 
-# Ingore moe/marlin_moe gen code
+# Ignore moe/marlin_moe gen code
 csrc/moe/marlin_moe_wna16/kernel_*
+
+# Ignore ep_kernels_workspace folder
+ep_kernels_workspace/

.markdownlint.yaml

@@ -0,0 +1,13 @@
+MD007:
+  indent: 4
+MD013: false
+MD024:
+  siblings_only: true
+MD033: false
+MD042: false
+MD045: false
+MD046: false
+MD051: false
+MD052: false
+MD053: false
+MD059: false

.pre-commit-config.yaml

@@ -20,12 +20,10 @@ repos:
     args: [--output-format, github, --fix]
   - id: ruff-format
     files: ^(.buildkite|benchmarks|examples)/.*
-- repo: https://github.com/codespell-project/codespell
-  rev: v2.4.1
+- repo: https://github.com/crate-ci/typos
+  rev: v1.34.0
   hooks:
-  - id: codespell
-    additional_dependencies: ['tomli']
-    args: ['--toml', 'pyproject.toml']
+  - id: typos
 - repo: https://github.com/PyCQA/isort
   rev: 6.0.1
   hooks:
@@ -37,12 +35,12 @@ repos:
     exclude: 'csrc/(moe/topk_softmax_kernels.cu|quantization/gguf/(ggml-common.h|dequantize.cuh|vecdotq.cuh|mmq.cuh|mmvq.cuh))|vllm/third_party/.*'
     types_or: [c++, cuda]
     args: [--style=file, --verbose]
-- repo: https://github.com/jackdewinter/pymarkdown
-  rev: v0.9.29
+- repo: https://github.com/igorshubovych/markdownlint-cli
+  rev: v0.45.0
   hooks:
-  - id: pymarkdown
+  - id: markdownlint
     exclude: '.*\.inc\.md'
-    args: [fix]
+    stages: [manual] # Only run in CI
 - repo: https://github.com/rhysd/actionlint
   rev: v1.7.7
   hooks:
@@ -55,6 +53,11 @@ repos:
       files: ^requirements/test\.(in|txt)$
 - repo: local
   hooks:
+  - id: format-torch-nightly-test
+    name: reformat nightly_torch_test.txt to be in sync with test.in
+    language: python
+    entry: python tools/generate_nightly_torch_test.py
+    files: ^requirements/test\.(in|txt)$
   - id: mypy-local
     name: Run mypy for local Python installation
     entry: tools/mypy.sh 0 "local"
@@ -117,6 +120,11 @@ repos:
     entry: python tools/check_spdx_header.py
     language: python
     types: [python]
+  - id: check-root-lazy-imports
+    name: Check root lazy imports
+    entry: python tools/check_init_lazy_imports.py
+    language: python
+    types: [python]
   - id: check-filenames
     name: Check for spaces in all filenames
     entry: bash
@@ -145,10 +153,24 @@ repos:
     types: [python]
     pass_filenames: false
     additional_dependencies: [regex]
+  - id: check-pickle-imports
+    name: Prevent new pickle/cloudpickle imports
+    entry: python tools/check_pickle_imports.py
+    language: python
+    types: [python]
+    pass_filenames: false
+    additional_dependencies: [pathspec, regex]
+  - id: validate-config
+    name: Validate configuration has default values and that each field has a docstring
+    entry: python tools/validate_config.py
+    language: python
+    types: [python]
+    pass_filenames: true
+    files: vllm/config.py|tests/test_config.py|vllm/entrypoints/openai/cli_args.py
   # Keep `suggestion` last
   - id: suggestion
     name: Suggestion
-    entry: bash -c 'echo "To bypass pre-commit hooks, add --no-verify to git commit."'
+    entry: bash -c 'echo "To bypass all the pre-commit hooks, add --no-verify to git commit. To skip a specific hook, prefix the commit command with SKIP=<hook-id>."'
     language: system
     verbose: true
     pass_filenames: false

.readthedocs.yaml

@@ -7,6 +7,9 @@ build:
   os: ubuntu-22.04
   tools:
     python: "3.12"
+  jobs:
+    post_checkout:
+      - git fetch --unshallow || true
 
 mkdocs:
   configuration: mkdocs.yaml

CMakeLists.txt

@@ -45,7 +45,7 @@ set(HIP_SUPPORTED_ARCHS "gfx906;gfx908;gfx90a;gfx942;gfx950;gfx1030;gfx1100;gfx1
 # requirements.txt files and should be kept consistent.  The ROCm torch
 # versions are derived from docker/Dockerfile.rocm
 #
-set(TORCH_SUPPORTED_VERSION_CUDA "2.7.0")
+set(TORCH_SUPPORTED_VERSION_CUDA "2.7.1")
 set(TORCH_SUPPORTED_VERSION_ROCM "2.7.0")
 
 #
@@ -171,7 +171,6 @@ if(NVCC_THREADS AND VLLM_GPU_LANG STREQUAL "CUDA")
   list(APPEND VLLM_GPU_FLAGS "--threads=${NVCC_THREADS}")
 endif()
 
-
 #
 # Use FetchContent for C++ dependencies that are compiled as part of vLLM's build process.
 # setup.py will override FETCHCONTENT_BASE_DIR to play nicely with sccache.
@@ -232,7 +231,6 @@ endif()
 
 set(VLLM_EXT_SRC
   "csrc/mamba/mamba_ssm/selective_scan_fwd.cu"
-  "csrc/mamba/causal_conv1d/causal_conv1d.cu"
   "csrc/cache_kernels.cu"
   "csrc/attention/paged_attention_v1.cu"
   "csrc/attention/paged_attention_v2.cu"
@@ -251,7 +249,6 @@ set(VLLM_EXT_SRC
   "csrc/quantization/gguf/gguf_kernel.cu"
   "csrc/quantization/activation_kernels.cu"
   "csrc/cuda_utils_kernels.cu"
-  "csrc/prepare_inputs/advance_step.cu"
   "csrc/custom_all_reduce.cu"
   "csrc/torch_bindings.cpp")
 
@@ -259,7 +256,7 @@ if(VLLM_GPU_LANG STREQUAL "CUDA")
   SET(CUTLASS_ENABLE_HEADERS_ONLY ON CACHE BOOL "Enable only the header library")
 
   # Set CUTLASS_REVISION. Used for FetchContent. Also fixes some bogus messages when building.
-  set(CUTLASS_REVISION "v3.9.2" CACHE STRING "CUTLASS revision to use")
+  set(CUTLASS_REVISION "v4.0.0" CACHE STRING "CUTLASS revision to use")
 
   # Use the specified CUTLASS source directory for compilation if VLLM_CUTLASS_SRC_DIR is provided
   if (DEFINED ENV{VLLM_CUTLASS_SRC_DIR})
@@ -289,7 +286,6 @@ if(VLLM_GPU_LANG STREQUAL "CUDA")
   FetchContent_MakeAvailable(cutlass)
 
   list(APPEND VLLM_EXT_SRC
-    "csrc/quantization/aqlm/gemm_kernels.cu"
     "csrc/quantization/awq/gemm_kernels.cu"
     "csrc/permute_cols.cu"
     "csrc/quantization/cutlass_w8a8/scaled_mm_entry.cu"
@@ -298,7 +294,8 @@ if(VLLM_GPU_LANG STREQUAL "CUDA")
     "csrc/quantization/fp4/nvfp4_blockwise_moe_kernel.cu"
     "csrc/sparse/cutlass/sparse_scaled_mm_entry.cu"
     "csrc/cutlass_extensions/common.cpp"
-    "csrc/attention/mla/cutlass_mla_entry.cu")
+    "csrc/attention/mla/cutlass_mla_entry.cu"
+    "csrc/quantization/fp8/per_token_group_quant.cu")
 
   set_gencode_flags_for_srcs(
     SRCS "${VLLM_EXT_SRC}"
@@ -352,6 +349,10 @@ if(VLLM_GPU_LANG STREQUAL "CUDA")
     set_gencode_flags_for_srcs(
       SRCS "${MARLIN_TEMPLATE_KERNEL_SRC}"
       CUDA_ARCHS "${MARLIN_ARCHS}")
+    if(${CMAKE_CUDA_COMPILER_VERSION} VERSION_GREATER_EQUAL 12.8)
+      set_source_files_properties(${MARLIN_TEMPLATE_KERNEL_SRC}
+        PROPERTIES COMPILE_FLAGS "-static-global-template-stub=false")
+    endif()
 
     list(APPEND VLLM_EXT_SRC ${MARLIN_TEMPLATE_KERNEL_SRC})
 
@@ -365,7 +366,12 @@ if(VLLM_GPU_LANG STREQUAL "CUDA")
     set_gencode_flags_for_srcs(
       SRCS "${MARLIN_SRCS}"
       CUDA_ARCHS "${MARLIN_ARCHS}")
+    if(${CMAKE_CUDA_COMPILER_VERSION} VERSION_GREATER_EQUAL 12.8)
+      set_source_files_properties("csrc/quantization/gptq_marlin/gptq_marlin.cu"
+        PROPERTIES COMPILE_FLAGS "-static-global-template-stub=false")
+    endif()
     list(APPEND VLLM_EXT_SRC "${MARLIN_SRCS}")
+
     message(STATUS "Building Marlin kernels for archs: ${MARLIN_ARCHS}")
   else()
     message(STATUS "Not building Marlin kernels as no compatible archs found"
@@ -393,7 +399,7 @@ if(VLLM_GPU_LANG STREQUAL "CUDA")
   # The cutlass_scaled_mm kernels for Hopper (c3x, i.e. CUTLASS 3.x) require
   # CUDA 12.0 or later
   cuda_archs_loose_intersection(SCALED_MM_ARCHS "9.0a;" "${CUDA_ARCHS}")
-  if(${CMAKE_CUDA_COMPILER_VERSION} VERSION_GREATER 12.0 AND SCALED_MM_ARCHS)
+  if(${CMAKE_CUDA_COMPILER_VERSION} VERSION_GREATER_EQUAL 12.0 AND SCALED_MM_ARCHS)
     set(SRCS
        "csrc/quantization/cutlass_w8a8/scaled_mm_c3x_sm90.cu"
        "csrc/quantization/cutlass_w8a8/c3x/scaled_mm_sm90_fp8.cu"
@@ -409,7 +415,7 @@ if(VLLM_GPU_LANG STREQUAL "CUDA")
     list(APPEND SCALED_MM_3X_ARCHS "${SCALED_MM_ARCHS}")
     message(STATUS "Building scaled_mm_c3x_sm90 for archs: ${SCALED_MM_ARCHS}")
   else()
-    if (NOT ${CMAKE_CUDA_COMPILER_VERSION} VERSION_GREATER 12.0 AND SCALED_MM_ARCHS)
+    if (NOT ${CMAKE_CUDA_COMPILER_VERSION} VERSION_GREATER_EQUAL 12.0 AND SCALED_MM_ARCHS)
       message(STATUS "Not building scaled_mm_c3x_sm90 as CUDA Compiler version is "
                      "not >= 12.0, we recommend upgrading to CUDA 12.0 or "
                      "later if you intend on running FP8 quantized models on "
@@ -420,10 +426,41 @@ if(VLLM_GPU_LANG STREQUAL "CUDA")
     endif()
   endif()
 
-  # The cutlass_scaled_mm kernels for Blackwell (c3x, i.e. CUTLASS 3.x) require
+
+  # The cutlass_scaled_mm kernels for Geforce Blackwell SM120 (c3x, i.e. CUTLASS 3.x) require
   # CUDA 12.8 or later
-  cuda_archs_loose_intersection(SCALED_MM_ARCHS "10.0a;10.1a;12.0a" "${CUDA_ARCHS}")
-  if(${CMAKE_CUDA_COMPILER_VERSION} VERSION_GREATER 12.8 AND SCALED_MM_ARCHS)
+  cuda_archs_loose_intersection(SCALED_MM_ARCHS "12.0;12.0a" "${CUDA_ARCHS}")
+  if(${CMAKE_CUDA_COMPILER_VERSION} VERSION_GREATER_EQUAL 12.8 AND SCALED_MM_ARCHS)
+    set(SRCS
+      "csrc/quantization/cutlass_w8a8/scaled_mm_c3x_sm120.cu"
+      "csrc/quantization/cutlass_w8a8/c3x/scaled_mm_sm120_fp8.cu"
+      "csrc/quantization/cutlass_w8a8/c3x/scaled_mm_blockwise_sm120_fp8.cu"
+    )
+    set_gencode_flags_for_srcs(
+      SRCS "${SRCS}"
+      CUDA_ARCHS "${SCALED_MM_ARCHS}")
+    list(APPEND VLLM_EXT_SRC "${SRCS}")
+    list(APPEND VLLM_GPU_FLAGS "-DENABLE_SCALED_MM_SM120=1")
+    # Let scaled_mm_c2x know it doesn't need to build these arches
+    list(APPEND SCALED_MM_3X_ARCHS "${SCALED_MM_ARCHS}")
+    message(STATUS "Building scaled_mm_c3x_sm120 for archs: ${SCALED_MM_ARCHS}")
+  else()
+    if (NOT ${CMAKE_CUDA_COMPILER_VERSION} VERSION_GREATER_EQUAL 12.8 AND SCALED_MM_ARCHS)
+      message(STATUS "Not building scaled_mm_c3x_sm120 as CUDA Compiler version is "
+                     "not >= 12.8, we recommend upgrading to CUDA 12.8 or "
+                     "later if you intend on running FP8 quantized models on "
+                     "Blackwell.")
+    else()
+      message(STATUS "Not building scaled_mm_c3x_120 as no compatible archs found "
+                     "in CUDA target architectures")
+    endif()
+  endif()
+
+
+  # The cutlass_scaled_mm kernels for Blackwell SM100 (c3x, i.e. CUTLASS 3.x)
+  # require CUDA 12.8 or later
+  cuda_archs_loose_intersection(SCALED_MM_ARCHS "10.0a;10.1a" "${CUDA_ARCHS}")
+  if(${CMAKE_CUDA_COMPILER_VERSION} VERSION_GREATER_EQUAL 12.8 AND SCALED_MM_ARCHS)
     set(SRCS
       "csrc/quantization/cutlass_w8a8/scaled_mm_c3x_sm100.cu"
       "csrc/quantization/cutlass_w8a8/c3x/scaled_mm_sm100_fp8.cu"
@@ -438,7 +475,7 @@ if(VLLM_GPU_LANG STREQUAL "CUDA")
     list(APPEND SCALED_MM_3X_ARCHS "${SCALED_MM_ARCHS}")
     message(STATUS "Building scaled_mm_c3x_sm100 for archs: ${SCALED_MM_ARCHS}")
   else()
-    if (NOT ${CMAKE_CUDA_COMPILER_VERSION} VERSION_GREATER 12.8 AND SCALED_MM_ARCHS)
+    if (NOT ${CMAKE_CUDA_COMPILER_VERSION} VERSION_GREATER_EQUAL 12.8 AND SCALED_MM_ARCHS)
       message(STATUS "Not building scaled_mm_c3x_sm100 as CUDA Compiler version is "
                      "not >= 12.8, we recommend upgrading to CUDA 12.8 or "
                      "later if you intend on running FP8 quantized models on "
@@ -481,7 +518,7 @@ if(VLLM_GPU_LANG STREQUAL "CUDA")
   # The 2:4 sparse kernels cutlass_scaled_sparse_mm and cutlass_compressor
   # require CUDA 12.2 or later (and only work on Hopper).
   cuda_archs_loose_intersection(SCALED_MM_ARCHS "9.0a;" "${CUDA_ARCHS}")
-  if(${CMAKE_CUDA_COMPILER_VERSION} VERSION_GREATER 12.2 AND SCALED_MM_ARCHS)
+  if(${CMAKE_CUDA_COMPILER_VERSION} VERSION_GREATER_EQUAL 12.2 AND SCALED_MM_ARCHS)
     set(SRCS "csrc/sparse/cutlass/sparse_scaled_mm_c3x.cu")
     set_gencode_flags_for_srcs(
       SRCS "${SRCS}"
@@ -490,7 +527,7 @@ if(VLLM_GPU_LANG STREQUAL "CUDA")
     list(APPEND VLLM_GPU_FLAGS "-DENABLE_SPARSE_SCALED_MM_C3X=1")
     message(STATUS "Building sparse_scaled_mm_c3x for archs: ${SCALED_MM_ARCHS}")
   else()
-    if (NOT ${CMAKE_CUDA_COMPILER_VERSION} VERSION_GREATER 12.2 AND SCALED_MM_ARCHS)
+    if (NOT ${CMAKE_CUDA_COMPILER_VERSION} VERSION_GREATER_EQUAL 12.2 AND SCALED_MM_ARCHS)
       message(STATUS "Not building sparse_scaled_mm_c3x kernels as CUDA Compiler version is "
                      "not >= 12.2, we recommend upgrading to CUDA 12.2 or later "
                      "if you intend on running FP8 sparse quantized models on Hopper.")
@@ -500,9 +537,28 @@ if(VLLM_GPU_LANG STREQUAL "CUDA")
     endif()
   endif()
 
+  # The nvfp4_scaled_mm_sm120 kernels for Geforce Blackwell SM120 require
+  # CUDA 12.8 or later
+  cuda_archs_loose_intersection(FP4_ARCHS "12.0;12.0a" "${CUDA_ARCHS}")
+  if(${CMAKE_CUDA_COMPILER_VERSION} VERSION_GREATER_EQUAL 12.8 AND FP4_ARCHS)
+    set(SRCS
+      "csrc/quantization/fp4/nvfp4_quant_kernels.cu"
+      "csrc/quantization/fp4/nvfp4_scaled_mm_sm120_kernels.cu")
+    set_gencode_flags_for_srcs(
+      SRCS "${SRCS}"
+      CUDA_ARCHS "${FP4_ARCHS}")
+    list(APPEND VLLM_EXT_SRC "${SRCS}")
+    list(APPEND VLLM_GPU_FLAGS "-DENABLE_NVFP4_SM120=1")
+    message(STATUS "Building NVFP4 for archs: ${FP4_ARCHS}")
+  else()
+    message(STATUS "Not building NVFP4 as no compatible archs were found.")
+    # clear FP4_ARCHS
+    set(FP4_ARCHS)
+  endif()
+
   # FP4 Archs and flags
   cuda_archs_loose_intersection(FP4_ARCHS "10.0a" "${CUDA_ARCHS}")
-  if(${CMAKE_CUDA_COMPILER_VERSION} VERSION_GREATER 12.8 AND FP4_ARCHS)
+  if(${CMAKE_CUDA_COMPILER_VERSION} VERSION_GREATER_EQUAL 12.8 AND FP4_ARCHS)
     set(SRCS
       "csrc/quantization/fp4/nvfp4_quant_kernels.cu"
       "csrc/quantization/fp4/nvfp4_experts_quant.cu"
@@ -512,7 +568,8 @@ if(VLLM_GPU_LANG STREQUAL "CUDA")
       SRCS "${SRCS}"
       CUDA_ARCHS "${FP4_ARCHS}")
     list(APPEND VLLM_EXT_SRC "${SRCS}")
-    list(APPEND VLLM_GPU_FLAGS "-DENABLE_NVFP4=1")
+    list(APPEND VLLM_GPU_FLAGS "-DENABLE_NVFP4_SM100=1")
+    list(APPEND VLLM_GPU_FLAGS "-DENABLE_CUTLASS_MOE_SM100=1")
     message(STATUS "Building NVFP4 for archs: ${FP4_ARCHS}")
   else()
     message(STATUS "Not building NVFP4 as no compatible archs were found.")
@@ -522,9 +579,10 @@ if(VLLM_GPU_LANG STREQUAL "CUDA")
 
   # CUTLASS MLA Archs and flags
   cuda_archs_loose_intersection(MLA_ARCHS "10.0a" "${CUDA_ARCHS}")
-  if(${CMAKE_CUDA_COMPILER_VERSION} VERSION_GREATER 12.8 AND MLA_ARCHS)
+  if(${CMAKE_CUDA_COMPILER_VERSION} VERSION_GREATER_EQUAL 12.8 AND MLA_ARCHS)
     set(SRCS
-      "csrc/attention/mla/cutlass_mla_kernels.cu")
+      "csrc/attention/mla/cutlass_mla_kernels.cu"
+      "csrc/attention/mla/sm100_cutlass_mla_kernel.cu")
     set_gencode_flags_for_srcs(
       SRCS "${SRCS}"
       CUDA_ARCHS "${MLA_ARCHS}")
@@ -542,13 +600,12 @@ if(VLLM_GPU_LANG STREQUAL "CUDA")
 
   # CUTLASS MoE kernels
 
-  # The MoE kernel cutlass_moe_mm requires CUDA 12.3 or later (and only works
+  # The MoE kernel cutlass_moe_mm requires CUDA 12.3 or later (and ONLY works
   # on Hopper). get_cutlass_(pplx_)moe_mm_data should only be compiled
   # if it's possible to compile MoE kernels that use its output.
-  cuda_archs_loose_intersection(SCALED_MM_ARCHS "9.0a;10.0a" "${CUDA_ARCHS}")
+  cuda_archs_loose_intersection(SCALED_MM_ARCHS "9.0a" "${CUDA_ARCHS}")
   if(${CMAKE_CUDA_COMPILER_VERSION} VERSION_GREATER_EQUAL 12.3 AND SCALED_MM_ARCHS)
-    set(SRCS "csrc/quantization/cutlass_w8a8/moe/grouped_mm_c3x.cu"
-             "csrc/quantization/cutlass_w8a8/moe/moe_data.cu")
+    set(SRCS "csrc/quantization/cutlass_w8a8/moe/grouped_mm_c3x_sm90.cu")
     set_gencode_flags_for_srcs(
       SRCS "${SRCS}"
       CUDA_ARCHS "${SCALED_MM_ARCHS}")
@@ -562,6 +619,66 @@ if(VLLM_GPU_LANG STREQUAL "CUDA")
                      "if you intend on running FP8 quantized MoE models on Hopper.")
     else()
       message(STATUS "Not building grouped_mm_c3x as no compatible archs found "
+                     "in CUDA target architectures.")
+    endif()
+  endif()
+
+  cuda_archs_loose_intersection(SCALED_MM_ARCHS "10.0a" "${CUDA_ARCHS}")
+  if(${CMAKE_CUDA_COMPILER_VERSION} VERSION_GREATER_EQUAL 12.8 AND SCALED_MM_ARCHS)
+    set(SRCS "csrc/quantization/cutlass_w8a8/moe/grouped_mm_c3x_sm100.cu")
+    set_gencode_flags_for_srcs(
+      SRCS "${SRCS}"
+      CUDA_ARCHS "${SCALED_MM_ARCHS}")
+    list(APPEND VLLM_EXT_SRC "${SRCS}")
+    list(APPEND VLLM_GPU_FLAGS "-DENABLE_CUTLASS_MOE_SM100=1")
+    message(STATUS "Building grouped_mm_c3x for archs: ${SCALED_MM_ARCHS}")
+  else()
+    if (NOT ${CMAKE_CUDA_COMPILER_VERSION} VERSION_GREATER_EQUAL 12.8 AND SCALED_MM_ARCHS)
+      message(STATUS "Not building grouped_mm_c3x kernels as CUDA Compiler version is "
+                     "not >= 12.8, we recommend upgrading to CUDA 12.8 or later "
+                     "if you intend on running FP8 quantized MoE models on Blackwell.")
+    else()
+      message(STATUS "Not building grouped_mm_c3x as no compatible archs found "
+                     "in CUDA target architectures.")
+    endif()
+  endif()
+
+  # moe_data.cu is used by all CUTLASS MoE kernels.
+  cuda_archs_loose_intersection(CUTLASS_MOE_DATA_ARCHS "9.0a;10.0a" "${CUDA_ARCHS}")
+  if(${CMAKE_CUDA_COMPILER_VERSION} VERSION_GREATER_EQUAL 12.3 AND CUTLASS_MOE_DATA_ARCHS)
+    set(SRCS "csrc/quantization/cutlass_w8a8/moe/moe_data.cu")
+    set_gencode_flags_for_srcs(
+      SRCS "${SRCS}"
+      CUDA_ARCHS "${CUTLASS_MOE_DATA_ARCHS}")
+    list(APPEND VLLM_EXT_SRC "${SRCS}")
+    message(STATUS "Building moe_data for archs: ${CUTLASS_MOE_DATA_ARCHS}")
+  else()
+    if (NOT ${CMAKE_CUDA_COMPILER_VERSION} VERSION_GREATER_EQUAL 12.3 AND CUTLASS_MOE_DATA_ARCHS)
+      message(STATUS "Not building moe_data as CUDA Compiler version is "
+                     "not >= 12.3, we recommend upgrading to CUDA 12.3 or later "
+                     "if you intend on running FP8 quantized MoE models on Hopper or Blackwell.")
+    else()
+      message(STATUS "Not building moe_data as no compatible archs found "
+                     "in CUDA target architectures.")
+    endif()
+  endif()
+
+  cuda_archs_loose_intersection(SCALED_MM_ARCHS "10.0a" "${CUDA_ARCHS}")
+  if(${CMAKE_CUDA_COMPILER_VERSION} VERSION_GREATER_EQUAL 12.8 AND SCALED_MM_ARCHS)
+    set(SRCS "csrc/quantization/cutlass_w8a8/moe/blockwise_scaled_group_mm_sm100.cu")
+    set_gencode_flags_for_srcs(
+      SRCS "${SRCS}"
+      CUDA_ARCHS "${SCALED_MM_ARCHS}")
+    list(APPEND VLLM_EXT_SRC "${SRCS}")
+    list(APPEND VLLM_GPU_FLAGS "-DENABLE_CUTLASS_MOE_SM100=1")
+    message(STATUS "Building blockwise_scaled_group_mm_sm100 for archs: ${SCALED_MM_ARCHS}")
+  else()
+    if (NOT ${CMAKE_CUDA_COMPILER_VERSION} VERSION_GREATER_EQUAL 12.8 AND SCALED_MM_ARCHS)
+      message(STATUS "Not building blockwise_scaled_group_mm_sm100 kernels as CUDA Compiler version is "
+                     "not >= 12.8, we recommend upgrading to CUDA 12.8 or later "
+                     "if you intend on running FP8 quantized MoE models on Blackwell.")
+    else()
+      message(STATUS "Not building blockwise_scaled_group_mm_sm100 as no compatible archs found "
                      "in CUDA target architectures")
     endif()
   endif()
@@ -572,7 +689,7 @@ if(VLLM_GPU_LANG STREQUAL "CUDA")
   # The machete kernels only work on hopper and require CUDA 12.0 or later.
   # Only build Machete kernels if we are building for something compatible with sm90a
   cuda_archs_loose_intersection(MACHETE_ARCHS "9.0a" "${CUDA_ARCHS}")
-  if(${CMAKE_CUDA_COMPILER_VERSION} VERSION_GREATER 12.0 AND MACHETE_ARCHS)
+  if(${CMAKE_CUDA_COMPILER_VERSION} VERSION_GREATER_EQUAL 12.0 AND MACHETE_ARCHS)
     #
     # For the Machete kernels we automatically generate sources for various
     # preselected input type pairs and schedules.
@@ -624,7 +741,7 @@ if(VLLM_GPU_LANG STREQUAL "CUDA")
 
     message(STATUS "Building Machete kernels for archs: ${MACHETE_ARCHS}")
   else()
-    if (NOT ${CMAKE_CUDA_COMPILER_VERSION} VERSION_GREATER 12.0
+    if (NOT ${CMAKE_CUDA_COMPILER_VERSION} VERSION_GREATER_EQUAL 12.0
         AND MACHETE_ARCHS)
       message(STATUS "Not building Machete kernels as CUDA Compiler version is "
                      "not >= 12.0, we recommend upgrading to CUDA 12.0 or "
@@ -638,6 +755,14 @@ if(VLLM_GPU_LANG STREQUAL "CUDA")
 # if CUDA endif
 endif()
 
+if (VLLM_GPU_LANG STREQUAL "HIP")
+  # Add QuickReduce kernels
+  list(APPEND VLLM_EXT_SRC
+    "csrc/custom_quickreduce.cu"
+  )
+# if ROCM endif
+endif()
+
 message(STATUS "Enabling C extension.")
 define_gpu_extension_target(
   _C
@@ -670,6 +795,14 @@ if(VLLM_GPU_LANG STREQUAL "CUDA")
   list(APPEND VLLM_MOE_EXT_SRC "csrc/moe/moe_wna16.cu")
 endif()
 
+if(VLLM_GPU_LANG STREQUAL "CUDA")
+  set(MOE_PERMUTE_SRC
+      "csrc/moe/permute_unpermute_kernels/moe_permute_unpermute_kernel.cu"
+      "csrc/moe/moe_permute_unpermute_op.cu")
+
+  list(APPEND VLLM_MOE_EXT_SRC "${MOE_PERMUTE_SRC}")
+endif()
+
 set_gencode_flags_for_srcs(
   SRCS "${VLLM_MOE_EXT_SRC}"
   CUDA_ARCHS "${CUDA_ARCHS}")
@@ -728,6 +861,10 @@ if(VLLM_GPU_LANG STREQUAL "CUDA")
     set_gencode_flags_for_srcs(
       SRCS "${MOE_WNAA16_MARLIN_SRC}"
       CUDA_ARCHS "${MARLIN_MOE_ARCHS}")
+    if(${CMAKE_CUDA_COMPILER_VERSION} VERSION_GREATER_EQUAL 12.8)
+      set_source_files_properties(${MOE_WNAA16_MARLIN_SRC}
+        PROPERTIES COMPILE_FLAGS "-static-global-template-stub=false")
+    endif()
 
     list(APPEND VLLM_MOE_EXT_SRC ${MOE_WNAA16_MARLIN_SRC})
 
@@ -738,17 +875,6 @@ if(VLLM_GPU_LANG STREQUAL "CUDA")
   endif()
 endif()
 
-if(VLLM_GPU_LANG STREQUAL "CUDA")
-  set(MOE_PERMUTE_SRC
-      "csrc/moe/permute_unpermute_kernels/moe_permute_unpermute_kernel.cu"
-      "csrc/moe/moe_permute_unpermute_op.cu")
-
-  set_gencode_flags_for_srcs(
-    SRCS "${MARLIN_PERMUTE_SRC}"
-    CUDA_ARCHS "${MOE_PERMUTE_ARCHS}")
-
-  list(APPEND VLLM_MOE_EXT_SRC "${MOE_PERMUTE_SRC}")
-endif()
 message(STATUS "Enabling moe extension.")
 define_gpu_extension_target(
   _moe_C

README.md

@@ -1,3 +1,4 @@
+<!-- markdownlint-disable MD001 MD041 -->
 <p align="center">
   <picture>
     <source media="(prefers-color-scheme: dark)" srcset="https://raw.githubusercontent.com/vllm-project/vllm/main/docs/assets/logos/vllm-logo-text-dark.png">
@@ -16,14 +17,16 @@ Easy, fast, and cheap LLM serving for everyone
 ---
 
 *Latest News* 🔥
+
+- [2025/08] We hosted [vLLM Beijing Meetup](https://mp.weixin.qq.com/s/dgkWg1WFpWGO2jCdTqQHxA) focusing on large-scale LLM deployment! Please find the meetup slides [here](https://drive.google.com/drive/folders/1Pid6NSFLU43DZRi0EaTcPgXsAzDvbBqF) and the recording [here](https://www.chaspark.com/#/live/1166916873711665152).
 - [2025/05] We hosted [NYC vLLM Meetup](https://lu.ma/c1rqyf1f)! Please find the meetup slides [here](https://docs.google.com/presentation/d/1_q_aW_ioMJWUImf1s1YM-ZhjXz8cUeL0IJvaquOYBeA/edit?usp=sharing).
 - [2025/05] vLLM is now a hosted project under PyTorch Foundation! Please find the announcement [here](https://pytorch.org/blog/pytorch-foundation-welcomes-vllm/).
-- [2025/04] We hosted [Asia Developer Day](https://www.sginnovate.com/event/limited-availability-morning-evening-slots-remaining-inaugural-vllm-asia-developer-day)! Please find the meetup slides from the vLLM team [here](https://docs.google.com/presentation/d/19cp6Qu8u48ihB91A064XfaXruNYiBOUKrBxAmDOllOo/edit?usp=sharing).
 - [2025/01] We are excited to announce the alpha release of vLLM V1: A major architectural upgrade with 1.7x speedup! Clean code, optimized execution loop, zero-overhead prefix caching, enhanced multimodal support, and more. Please check out our blog post [here](https://blog.vllm.ai/2025/01/27/v1-alpha-release.html).
 
 <details>
 <summary>Previous News</summary>
 
+- [2025/04] We hosted [Asia Developer Day](https://www.sginnovate.com/event/limited-availability-morning-evening-slots-remaining-inaugural-vllm-asia-developer-day)! Please find the meetup slides from the vLLM team [here](https://docs.google.com/presentation/d/19cp6Qu8u48ihB91A064XfaXruNYiBOUKrBxAmDOllOo/edit?usp=sharing).
 - [2025/03] We hosted [vLLM x Ollama Inference Night](https://lu.ma/vllm-ollama)! Please find the meetup slides from the vLLM team [here](https://docs.google.com/presentation/d/16T2PDD1YwRnZ4Tu8Q5r6n53c5Lr5c73UV9Vd2_eBo4U/edit?usp=sharing).
 - [2025/03] We hosted [the first vLLM China Meetup](https://mp.weixin.qq.com/s/n77GibL2corAtQHtVEAzfg)! Please find the meetup slides from vLLM team [here](https://docs.google.com/presentation/d/1REHvfQMKGnvz6p3Fd23HhSO4c8j5WPGZV0bKYLwnHyQ/edit?usp=sharing).
 - [2025/03] We hosted [the East Coast vLLM Meetup](https://lu.ma/7mu4k4xx)! Please find the meetup slides [here](https://docs.google.com/presentation/d/1NHiv8EUFF1NLd3fEYODm56nDmL26lEeXCaDgyDlTsRs/edit#slide=id.g31441846c39_0_0).
@@ -46,6 +49,7 @@ Easy, fast, and cheap LLM serving for everyone
 </details>
 
 ---
+
 ## About
 
 vLLM is a fast and easy-to-use library for LLM inference and serving.
@@ -63,13 +67,11 @@ vLLM is fast with:
 - Speculative decoding
 - Chunked prefill
 
-**Performance benchmark**: We include a performance benchmark at the end of [our blog post](https://blog.vllm.ai/2024/09/05/perf-update.html). It compares the performance of vLLM against other LLM serving engines ([TensorRT-LLM](https://github.com/NVIDIA/TensorRT-LLM), [SGLang](https://github.com/sgl-project/sglang) and [LMDeploy](https://github.com/InternLM/lmdeploy)). The implementation is under [nightly-benchmarks folder](.buildkite/nightly-benchmarks/) and you can [reproduce](https://github.com/vllm-project/vllm/issues/8176) this benchmark using our one-click runnable script.
-
 vLLM is flexible and easy to use with:
 
 - Seamless integration with popular Hugging Face models
 - High-throughput serving with various decoding algorithms, including *parallel sampling*, *beam search*, and more
-- Tensor parallelism and pipeline parallelism support for distributed inference
+- Tensor, pipeline, data and expert parallelism support for distributed inference
 - Streaming outputs
 - OpenAI-compatible API server
 - Support NVIDIA GPUs, AMD CPUs and GPUs, Intel CPUs and GPUs, PowerPC CPUs, TPU, and AWS Neuron
@@ -77,6 +79,7 @@ vLLM is flexible and easy to use with:
 - Multi-LoRA support
 
 vLLM seamlessly supports most popular open-source models on HuggingFace, including:
+
 - Transformer-like LLMs (e.g., Llama)
 - Mixture-of-Expert LLMs (e.g., Mixtral, Deepseek-V2 and V3)
 - Embedding Models (e.g., E5-Mistral)
@@ -93,6 +96,7 @@ pip install vllm
 ```
 
 Visit our [documentation](https://docs.vllm.ai/en/latest/) to learn more.
+
 - [Installation](https://docs.vllm.ai/en/latest/getting_started/installation.html)
 - [Quickstart](https://docs.vllm.ai/en/latest/getting_started/quickstart.html)
 - [List of Supported Models](https://docs.vllm.ai/en/latest/models/supported_models.html)
@@ -109,6 +113,7 @@ vLLM is a community project. Our compute resources for development and testing a
 <!-- Note: Please sort them in alphabetical order. -->
 <!-- Note: Please keep these consistent with docs/community/sponsors.md -->
 Cash Donations:
+
 - a16z
 - Dropbox
 - Sequoia Capital
@@ -116,6 +121,8 @@ Cash Donations:
 - ZhenFund
 
 Compute Resources:
+
+- Alibaba Cloud
 - AMD
 - Anyscale
 - AWS
@@ -154,11 +161,13 @@ If you use vLLM for your research, please cite our [paper](https://arxiv.org/abs
 
 ## Contact Us
 
-- For technical questions and feature requests, please use GitHub [Issues](https://github.com/vllm-project/vllm/issues) or [Discussions](https://github.com/vllm-project/vllm/discussions)
+<!-- --8<-- [start:contact-us] -->
+- For technical questions and feature requests, please use GitHub [Issues](https://github.com/vllm-project/vllm/issues)
 - For discussing with fellow users, please use the [vLLM Forum](https://discuss.vllm.ai)
-- coordinating contributions and development, please use [Slack](https://slack.vllm.ai)
+- For coordinating contributions and development, please use [Slack](https://slack.vllm.ai)
 - For security disclosures, please use GitHub's [Security Advisories](https://github.com/vllm-project/vllm/security/advisories) feature
 - For collaborations and partnerships, please contact us at [vllm-questions@lists.berkeley.edu](mailto:vllm-questions@lists.berkeley.edu)
+<!-- --8<-- [end:contact-us] -->
 
 ## Media Kit
 

RELEASE.md

@@ -52,3 +52,39 @@ After branch cut, we approach finalizing the release branch with clear criteria
 * Release branch specific changes (e.g. change version identifiers or CI fixes)
 
 Please note: **No feature work allowed for cherry picks**. All PRs that are considered for cherry-picks need to be merged on trunk, the only exception are Release branch specific changes.
+
+## Manual validations
+
+### E2E Performance Validation
+
+Before each release, we perform end-to-end performance validation to ensure no regressions are introduced. This validation uses the [vllm-benchmark workflow](https://github.com/pytorch/pytorch-integration-testing/actions/workflows/vllm-benchmark.yml) on PyTorch CI.
+
+**Current Coverage:**
+
+* Models: Llama3, Llama4, and Mixtral
+* Hardware: NVIDIA H100 and AMD MI300x
+* _Note: Coverage may change based on new model releases and hardware availability_
+
+**Performance Validation Process:**
+
+**Step 1: Get Access**
+Request write access to the [pytorch/pytorch-integration-testing](https://github.com/pytorch/pytorch-integration-testing) repository to run the benchmark workflow.
+
+**Step 2: Review Benchmark Setup**
+Familiarize yourself with the benchmark configurations:
+
+* [CUDA setup](https://github.com/pytorch/pytorch-integration-testing/tree/main/vllm-benchmarks/benchmarks/cuda)
+* [ROCm setup](https://github.com/pytorch/pytorch-integration-testing/tree/main/vllm-benchmarks/benchmarks/rocm)
+
+**Step 3: Run the Benchmark**
+Navigate to the [vllm-benchmark workflow](https://github.com/pytorch/pytorch-integration-testing/actions/workflows/vllm-benchmark.yml) and configure:
+
+* **vLLM branch**: Set to the release branch (e.g., `releases/v0.9.2`)
+* **vLLM commit**: Set to the RC commit hash
+
+**Step 4: Review Results**
+Once the workflow completes, benchmark results will be available on the [vLLM benchmark dashboard](https://hud.pytorch.org/benchmark/llms?repoName=vllm-project%2Fvllm) under the corresponding branch and commit.
+
+**Step 5: Performance Comparison**
+Compare the current results against the previous release to verify no performance regressions have occurred. Here is an
+example of [v0.9.1 vs v0.9.2](https://hud.pytorch.org/benchmark/llms?startTime=Thu%2C%2017%20Apr%202025%2021%3A43%3A50%20GMT&stopTime=Wed%2C%2016%20Jul%202025%2021%3A43%3A50%20GMT&granularity=week&lBranch=releases/v0.9.1&lCommit=b6553be1bc75f046b00046a4ad7576364d03c835&rBranch=releases/v0.9.2&rCommit=a5dd03c1ebc5e4f56f3c9d3dc0436e9c582c978f&repoName=vllm-project%2Fvllm&benchmarkName=&modelName=All%20Models&backendName=All%20Backends&modeName=All%20Modes&dtypeName=All%20DType&deviceName=All%20Devices&archName=All%20Platforms).

SECURITY.md

@@ -1,13 +1,45 @@
 # Security Policy
 
-## Reporting a Vulnerability
+## Reporting security issues
 
-If you believe you have found a security vulnerability in vLLM, we encourage you to let us know right away. We will investigate all legitimate reports and do our best to quickly fix the problem.
+Please report security issues privately using [the vulnerability submission form](https://github.com/vllm-project/vllm/security/advisories/new).
 
-Please report security issues privately using [the vulnerability submission form](https://github.com/vllm-project/vllm/security/advisories/new). Reports will then be triaged by the [vulnerability management team](https://docs.vllm.ai/en/latest/contributing/vulnerability_management.html).
+## Issue triage
 
----
+Reports will then be triaged by the [vulnerability management team](https://docs.vllm.ai/en/latest/contributing/vulnerability_management.html).
+
+## Threat model
 
 Please see the [Security Guide in the vLLM documentation](https://docs.vllm.ai/en/latest/usage/security.html) for more information on vLLM's security assumptions and recommendations.
 
 Please see [PyTorch's Security Policy](https://github.com/pytorch/pytorch/blob/main/SECURITY.md) for more information and recommendations on how to securely interact with models.
+
+## Issue severity
+
+We will determine the risk of each issue, taking into account our experience dealing with past issues, versions affected, common defaults, and use cases. We use the following severity categories:
+
+### CRITICAL Severity
+
+Vulnerabilities that allow remote attackers to execute arbitrary code, take full control of the system, or significantly compromise confidentiality, integrity, or availability without any interaction or privileges needed, examples include remote code execution via network, deserialization issues that allow exploit chains. Generally those issues which are rated as CVSS  ≥ 9.0.
+
+### HIGH Severity
+
+Serious security flaws that allow elevated impact—like RCE in specific, limited contexts or significant data loss—but require advanced conditions or some trust, examples include RCE in advanced deployment modes (e.g. multi-node), or high impact issues where some sort of privileged network access is required. These issues typically have CVSS scores between 7.0 and 8.9
+
+### MODERATE Severity
+
+Vulnerabilities that cause denial of service or partial disruption, but do not allow arbitrary code execution or data breach and have limited impact. These issues have a CVSS rating between 4.0 and 6.9
+
+### LOW Severity
+
+Minor issues such as informational disclosures, logging errors, non-exploitable flaws, or weaknesses that require local or high-privilege access and offer negligible impact. Examples include side channel attacks or hash collisions. These issues often have CVSS scores less than 4.0
+
+## Prenotification policy
+
+For certain security issues of CRITICAL, HIGH, or MODERATE severity level, we may prenotify certain organizations or vendors that ship vLLM. The purpose of this prenotification is to allow for a coordinated release of fixes for severe issues.
+
+* This prenotification will be in the form of a private email notification. It may also include adding security contacts to the GitHub security advisory, typically a few days before release.
+
+* If you wish to be added to the prenotification group, please send an email copying all the members of the [vulnerability management team](https://docs.vllm.ai/en/latest/contributing/vulnerability_management.html). Each vendor contact will be analyzed on a case-by-case basis.
+
+* We may withdraw organizations from receiving future prenotifications if they release fixes or any other information about issues before they are public. Group membership may also change based on policy refinements for who may be included.

benchmarks/README.md

@@ -22,6 +22,17 @@ become available.
       <td style="text-align: center;">✅</td>
       <td><code>wget https://huggingface.co/datasets/anon8231489123/ShareGPT_Vicuna_unfiltered/resolve/main/ShareGPT_V3_unfiltered_cleaned_split.json</code></td>
     </tr>
+    <tr>
+      <td><strong>ShareGPT4V (Image)</strong></td>
+      <td style="text-align: center;">✅</td>
+      <td style="text-align: center;">✅</td>
+      <td>
+        <code>wget https://huggingface.co/datasets/Lin-Chen/ShareGPT4V/blob/main/sharegpt4v_instruct_gpt4-vision_cap100k.json</code>
+        <br>
+        <div>Note that the images need to be downloaded separately. For example, to download COCO's 2017 Train images:</div>
+        <code>wget http://images.cocodataset.org/zips/train2017.zip</code>
+      </td>
+    </tr>
     <tr>
       <td><strong>BurstGPT</strong></td>
       <td style="text-align: center;">✅</td>
@@ -29,7 +40,7 @@ become available.
       <td><code>wget https://github.com/HPMLL/BurstGPT/releases/download/v1.1/BurstGPT_without_fails_2.csv</code></td>
     </tr>
     <tr>
-      <td><strong>Sonnet</strong></td>
+      <td><strong>Sonnet (deprecated)</strong></td>
       <td style="text-align: center;">✅</td>
       <td style="text-align: center;">✅</td>
       <td>Local file: <code>benchmarks/sonnet.txt</code></td>
@@ -40,6 +51,12 @@ become available.
       <td style="text-align: center;">✅</td>
       <td><code>synthetic</code></td>
     </tr>
+    <tr>
+      <td><strong>Prefix Repetition</strong></td>
+      <td style="text-align: center;">✅</td>
+      <td style="text-align: center;">✅</td>
+      <td><code>synthetic</code></td>
+    </tr>
     <tr>
       <td><strong>HuggingFace-VisionArena</strong></td>
       <td style="text-align: center;">✅</td>
@@ -81,13 +98,17 @@ become available.
 
 **Note**: HuggingFace dataset's `dataset-name` should be set to `hf`
 
----
-## Example - Online Benchmark
+## 🚀 Example - Online Benchmark
+
+<details>
+<summary>Show more</summary>
+
+<br/>
 
 First start serving your model
 
 ```bash
-vllm serve NousResearch/Hermes-3-Llama-3.1-8B --disable-log-requests
+vllm serve NousResearch/Hermes-3-Llama-3.1-8B
 ```
 
 Then run the benchmarking script
@@ -95,7 +116,7 @@ Then run the benchmarking script
 ```bash
 # download dataset
 # wget https://huggingface.co/datasets/anon8231489123/ShareGPT_Vicuna_unfiltered/resolve/main/ShareGPT_V3_unfiltered_cleaned_split.json
-python3 vllm/benchmarks/benchmark_serving.py \
+vllm bench serve \
   --backend vllm \
   --model NousResearch/Hermes-3-Llama-3.1-8B \
   --endpoint /v1/completions \
@@ -106,47 +127,48 @@ python3 vllm/benchmarks/benchmark_serving.py \
 
 If successful, you will see the following output
 
-```
+```text
 ============ Serving Benchmark Result ============
-Successful requests:                     10        
-Benchmark duration (s):                  5.78      
-Total input tokens:                      1369      
-Total generated tokens:                  2212      
-Request throughput (req/s):              1.73      
-Output token throughput (tok/s):         382.89    
-Total Token throughput (tok/s):          619.85    
+Successful requests:                     10
+Benchmark duration (s):                  5.78
+Total input tokens:                      1369
+Total generated tokens:                  2212
+Request throughput (req/s):              1.73
+Output token throughput (tok/s):         382.89
+Total Token throughput (tok/s):          619.85
 ---------------Time to First Token----------------
-Mean TTFT (ms):                          71.54     
-Median TTFT (ms):                        73.88     
-P99 TTFT (ms):                           79.49     
+Mean TTFT (ms):                          71.54
+Median TTFT (ms):                        73.88
+P99 TTFT (ms):                           79.49
 -----Time per Output Token (excl. 1st token)------
-Mean TPOT (ms):                          7.91      
-Median TPOT (ms):                        7.96      
-P99 TPOT (ms):                           8.03      
+Mean TPOT (ms):                          7.91
+Median TPOT (ms):                        7.96
+P99 TPOT (ms):                           8.03
 ---------------Inter-token Latency----------------
-Mean ITL (ms):                           7.74      
-Median ITL (ms):                         7.70      
-P99 ITL (ms):                            8.39      
+Mean ITL (ms):                           7.74
+Median ITL (ms):                         7.70
+P99 ITL (ms):                            8.39
 ==================================================
 ```
 
 ### Custom Dataset
+
 If the dataset you want to benchmark is not supported yet in vLLM, even then you can benchmark on it using `CustomDataset`. Your data needs to be in `.jsonl` format and needs to have "prompt" field per entry, e.g., data.jsonl
 
-```
+```json
 {"prompt": "What is the capital of India?"}
 {"prompt": "What is the capital of Iran?"}
 {"prompt": "What is the capital of China?"}
-``` 
+```
 
 ```bash
 # start server
-VLLM_USE_V1=1 vllm serve meta-llama/Llama-3.1-8B-Instruct --disable-log-requests
+VLLM_USE_V1=1 vllm serve meta-llama/Llama-3.1-8B-Instruct
 ```
 
 ```bash
 # run benchmarking script
-python3 benchmarks/benchmark_serving.py --port 9001 --save-result --save-detailed \
+vllm bench serve --port 9001 --save-result --save-detailed \
   --backend vllm \
   --model meta-llama/Llama-3.1-8B-Instruct \
   --endpoint /v1/completions \
@@ -166,11 +188,11 @@ You can skip applying chat template if your data already has it by using `--cust
 
 ```bash
 # need a model with vision capability here
-vllm serve Qwen/Qwen2-VL-7B-Instruct --disable-log-requests
+vllm serve Qwen/Qwen2-VL-7B-Instruct
 ```
 
 ```bash
-python3 vllm/benchmarks/benchmark_serving.py \
+vllm bench serve \
   --backend openai-chat \
   --model Qwen/Qwen2-VL-7B-Instruct \
   --endpoint /v1/chat/completions \
@@ -190,7 +212,7 @@ VLLM_USE_V1=1 vllm serve meta-llama/Meta-Llama-3-8B-Instruct \
 ```
 
 ``` bash
-python3 benchmarks/benchmark_serving.py \
+vllm bench serve \
     --model meta-llama/Meta-Llama-3-8B-Instruct \
     --dataset-name hf \
     --dataset-path likaixin/InstructCoder \
@@ -200,13 +222,13 @@ python3 benchmarks/benchmark_serving.py \
 ### Other HuggingFaceDataset Examples
 
 ```bash
-vllm serve Qwen/Qwen2-VL-7B-Instruct --disable-log-requests
+vllm serve Qwen/Qwen2-VL-7B-Instruct
 ```
 
-**`lmms-lab/LLaVA-OneVision-Data`**
+`lmms-lab/LLaVA-OneVision-Data`:
 
 ```bash
-python3 vllm/benchmarks/benchmark_serving.py \
+vllm bench serve \
   --backend openai-chat \
   --model Qwen/Qwen2-VL-7B-Instruct \
   --endpoint /v1/chat/completions \
@@ -217,10 +239,10 @@ python3 vllm/benchmarks/benchmark_serving.py \
   --num-prompts 10
 ```
 
-**`Aeala/ShareGPT_Vicuna_unfiltered`**
+`Aeala/ShareGPT_Vicuna_unfiltered`:
 
 ```bash
-python3 vllm/benchmarks/benchmark_serving.py \
+vllm bench serve \
   --backend openai-chat \
   --model Qwen/Qwen2-VL-7B-Instruct \
   --endpoint /v1/chat/completions \
@@ -230,10 +252,10 @@ python3 vllm/benchmarks/benchmark_serving.py \
   --num-prompts 10
 ```
 
-**`AI-MO/aimo-validation-aime`**
+`AI-MO/aimo-validation-aime`:
 
 ``` bash
-python3 vllm/benchmarks/benchmark_serving.py \
+vllm bench serve \
     --model Qwen/QwQ-32B \
     --dataset-name hf \
     --dataset-path AI-MO/aimo-validation-aime \
@@ -241,10 +263,10 @@ python3 vllm/benchmarks/benchmark_serving.py \
     --seed 42
 ```
 
-**`philschmid/mt-bench`**
+`philschmid/mt-bench`:
 
 ``` bash
-python3 vllm/benchmarks/benchmark_serving.py \
+vllm bench serve \
     --model Qwen/QwQ-32B \
     --dataset-name hf \
     --dataset-path philschmid/mt-bench \
@@ -257,7 +279,7 @@ When using OpenAI-compatible backends such as `vllm`, optional sampling
 parameters can be specified. Example client command:
 
 ```bash
-python3 vllm/benchmarks/benchmark_serving.py \
+vllm bench serve \
   --backend vllm \
   --model NousResearch/Hermes-3-Llama-3.1-8B \
   --endpoint /v1/completions \
@@ -269,11 +291,34 @@ python3 vllm/benchmarks/benchmark_serving.py \
   --num-prompts 10
 ```
 
----
-## Example - Offline Throughput Benchmark
+### Running With Ramp-Up Request Rate
+
+The benchmark tool also supports ramping up the request rate over the
+duration of the benchmark run. This can be useful for stress testing the
+server or finding the maximum throughput that it can handle, given some latency budget.
+
+Two ramp-up strategies are supported:
+
+- `linear`: Increases the request rate linearly from a start value to an end value.
+- `exponential`: Increases the request rate exponentially.
+
+The following arguments can be used to control the ramp-up:
+
+- `--ramp-up-strategy`: The ramp-up strategy to use (`linear` or `exponential`).
+- `--ramp-up-start-rps`: The request rate at the beginning of the benchmark.
+- `--ramp-up-end-rps`: The request rate at the end of the benchmark.
+
+</details>
+
+## 📈 Example - Offline Throughput Benchmark
+
+<details>
+<summary>Show more</summary>
+
+<br/>
 
 ```bash
-python3 vllm/benchmarks/benchmark_throughput.py \
+vllm bench throughput \
   --model NousResearch/Hermes-3-Llama-3.1-8B \
   --dataset-name sonnet \
   --dataset-path vllm/benchmarks/sonnet.txt \
@@ -282,7 +327,7 @@ python3 vllm/benchmarks/benchmark_throughput.py \
 
 If successful, you will see the following output
 
-```
+```text
 Throughput: 7.15 requests/s, 4656.00 total tokens/s, 1072.15 output tokens/s
 Total num prompt tokens:  5014
 Total num output tokens:  1500
@@ -290,8 +335,8 @@ Total num output tokens:  1500
 
 ### VisionArena Benchmark for Vision Language Models
 
-``` bash
-python3 vllm/benchmarks/benchmark_throughput.py \
+```bash
+vllm bench throughput \
   --model Qwen/Qwen2-VL-7B-Instruct \
   --backend vllm-chat \
   --dataset-name hf \
@@ -302,7 +347,7 @@ python3 vllm/benchmarks/benchmark_throughput.py \
 
 The `num prompt tokens` now includes image token counts
 
-```
+```text
 Throughput: 2.55 requests/s, 4036.92 total tokens/s, 326.90 output tokens/s
 Total num prompt tokens:  14527
 Total num output tokens:  1280
@@ -313,7 +358,7 @@ Total num output tokens:  1280
 ``` bash
 VLLM_WORKER_MULTIPROC_METHOD=spawn \
 VLLM_USE_V1=1 \
-python3 vllm/benchmarks/benchmark_throughput.py \
+vllm bench throughput \
     --dataset-name=hf \
     --dataset-path=likaixin/InstructCoder \
     --model=meta-llama/Meta-Llama-3-8B-Instruct \
@@ -326,7 +371,7 @@ python3 vllm/benchmarks/benchmark_throughput.py \
     "prompt_lookup_min": 2}'
 ```
 
-```
+```text
 Throughput: 104.77 requests/s, 23836.22 total tokens/s, 10477.10 output tokens/s
 Total num prompt tokens:  261136
 Total num output tokens:  204800
@@ -334,10 +379,10 @@ Total num output tokens:  204800
 
 ### Other HuggingFaceDataset Examples
 
-**`lmms-lab/LLaVA-OneVision-Data`**
+`lmms-lab/LLaVA-OneVision-Data`:
 
 ```bash
-python3 vllm/benchmarks/benchmark_throughput.py \
+vllm bench throughput \
   --model Qwen/Qwen2-VL-7B-Instruct \
   --backend vllm-chat \
   --dataset-name hf \
@@ -347,10 +392,10 @@ python3 vllm/benchmarks/benchmark_throughput.py \
   --num-prompts 10
 ```
 
-**`Aeala/ShareGPT_Vicuna_unfiltered`**
+`Aeala/ShareGPT_Vicuna_unfiltered`:
 
 ```bash
-python3 vllm/benchmarks/benchmark_throughput.py \
+vllm bench throughput \
   --model Qwen/Qwen2-VL-7B-Instruct \
   --backend vllm-chat \
   --dataset-name hf \
@@ -359,10 +404,10 @@ python3 vllm/benchmarks/benchmark_throughput.py \
   --num-prompts 10
 ```
 
-**`AI-MO/aimo-validation-aime`**
+`AI-MO/aimo-validation-aime`:
 
 ```bash
-python3 benchmarks/benchmark_throughput.py \
+vllm bench throughput \
   --model Qwen/QwQ-32B \
   --backend vllm \
   --dataset-name hf \
@@ -371,12 +416,12 @@ python3 benchmarks/benchmark_throughput.py \
   --num-prompts 10
 ```
 
-### Benchmark with LoRA Adapters
+Benchmark with LoRA adapters:
 
 ``` bash
 # download dataset
 # wget https://huggingface.co/datasets/anon8231489123/ShareGPT_Vicuna_unfiltered/resolve/main/ShareGPT_V3_unfiltered_cleaned_split.json
-python3 vllm/benchmarks/benchmark_throughput.py \
+vllm bench throughput \
   --model meta-llama/Llama-2-7b-hf \
   --backend vllm \
   --dataset_path <your data path>/ShareGPT_V3_unfiltered_cleaned_split.json \
@@ -387,3 +432,256 @@ python3 vllm/benchmarks/benchmark_throughput.py \
   --enable-lora \
   --lora-path yard1/llama-2-7b-sql-lora-test
   ```
+
+</details>
+
+## 🛠️ Example - Structured Output Benchmark
+
+<details>
+<summary>Show more</summary>
+
+<br/>
+
+Benchmark the performance of structured output generation (JSON, grammar, regex).
+
+### Server Setup
+
+```bash
+vllm serve NousResearch/Hermes-3-Llama-3.1-8B
+```
+
+### JSON Schema Benchmark
+
+```bash
+python3 benchmarks/benchmark_serving_structured_output.py \
+  --backend vllm \
+  --model NousResearch/Hermes-3-Llama-3.1-8B \
+  --dataset json \
+  --structured-output-ratio 1.0 \
+  --request-rate 10 \
+  --num-prompts 1000
+```
+
+### Grammar-based Generation Benchmark
+
+```bash
+python3 benchmarks/benchmark_serving_structured_output.py \
+  --backend vllm \
+  --model NousResearch/Hermes-3-Llama-3.1-8B \
+  --dataset grammar \
+  --structure-type grammar \
+  --request-rate 10 \
+  --num-prompts 1000
+```
+
+### Regex-based Generation Benchmark
+
+```bash
+python3 benchmarks/benchmark_serving_structured_output.py \
+  --backend vllm \
+  --model NousResearch/Hermes-3-Llama-3.1-8B \
+  --dataset regex \
+  --request-rate 10 \
+  --num-prompts 1000
+```
+
+### Choice-based Generation Benchmark
+
+```bash
+python3 benchmarks/benchmark_serving_structured_output.py \
+  --backend vllm \
+  --model NousResearch/Hermes-3-Llama-3.1-8B \
+  --dataset choice \
+  --request-rate 10 \
+  --num-prompts 1000
+```
+
+### XGrammar Benchmark Dataset
+
+```bash
+python3 benchmarks/benchmark_serving_structured_output.py \
+  --backend vllm \
+  --model NousResearch/Hermes-3-Llama-3.1-8B \
+  --dataset xgrammar_bench \
+  --request-rate 10 \
+  --num-prompts 1000
+```
+
+</details>
+
+## 📚 Example - Long Document QA Benchmark
+
+<details>
+<summary>Show more</summary>
+
+<br/>
+
+Benchmark the performance of long document question-answering with prefix caching.
+
+### Basic Long Document QA Test
+
+```bash
+python3 benchmarks/benchmark_long_document_qa_throughput.py \
+  --model meta-llama/Llama-2-7b-chat-hf \
+  --enable-prefix-caching \
+  --num-documents 16 \
+  --document-length 2000 \
+  --output-len 50 \
+  --repeat-count 5
+```
+
+### Different Repeat Modes
+
+```bash
+# Random mode (default) - shuffle prompts randomly
+python3 benchmarks/benchmark_long_document_qa_throughput.py \
+  --model meta-llama/Llama-2-7b-chat-hf \
+  --enable-prefix-caching \
+  --num-documents 8 \
+  --document-length 3000 \
+  --repeat-count 3 \
+  --repeat-mode random
+
+# Tile mode - repeat entire prompt list in sequence
+python3 benchmarks/benchmark_long_document_qa_throughput.py \
+  --model meta-llama/Llama-2-7b-chat-hf \
+  --enable-prefix-caching \
+  --num-documents 8 \
+  --document-length 3000 \
+  --repeat-count 3 \
+  --repeat-mode tile
+
+# Interleave mode - repeat each prompt consecutively
+python3 benchmarks/benchmark_long_document_qa_throughput.py \
+  --model meta-llama/Llama-2-7b-chat-hf \
+  --enable-prefix-caching \
+  --num-documents 8 \
+  --document-length 3000 \
+  --repeat-count 3 \
+  --repeat-mode interleave
+```
+
+</details>
+
+## 🗂️ Example - Prefix Caching Benchmark
+
+<details>
+<summary>Show more</summary>
+
+<br/>
+
+Benchmark the efficiency of automatic prefix caching.
+
+### Fixed Prompt with Prefix Caching
+
+```bash
+python3 benchmarks/benchmark_prefix_caching.py \
+  --model meta-llama/Llama-2-7b-chat-hf \
+  --enable-prefix-caching \
+  --num-prompts 1 \
+  --repeat-count 100 \
+  --input-length-range 128:256
+```
+
+### ShareGPT Dataset with Prefix Caching
+
+```bash
+# download dataset
+# wget https://huggingface.co/datasets/anon8231489123/ShareGPT_Vicuna_unfiltered/resolve/main/ShareGPT_V3_unfiltered_cleaned_split.json
+
+python3 benchmarks/benchmark_prefix_caching.py \
+  --model meta-llama/Llama-2-7b-chat-hf \
+  --dataset-path /path/ShareGPT_V3_unfiltered_cleaned_split.json \
+  --enable-prefix-caching \
+  --num-prompts 20 \
+  --repeat-count 5 \
+  --input-length-range 128:256
+```
+
+### Prefix Repetition Dataset
+
+```bash
+vllm bench serve \
+  --backend openai \
+  --model meta-llama/Llama-2-7b-chat-hf \
+  --dataset-name prefix_repetition \
+  --num-prompts 100 \
+  --prefix-repetition-prefix-len 512 \
+  --prefix-repetition-suffix-len 128 \
+  --prefix-repetition-num-prefixes 5 \
+  --prefix-repetition-output-len 128 
+```
+
+</details>
+
+## ⚡ Example - Request Prioritization Benchmark
+
+<details>
+<summary>Show more</summary>
+
+<br/>
+
+Benchmark the performance of request prioritization in vLLM.
+
+### Basic Prioritization Test
+
+```bash
+python3 benchmarks/benchmark_prioritization.py \
+  --model meta-llama/Llama-2-7b-chat-hf \
+  --input-len 128 \
+  --output-len 64 \
+  --num-prompts 100 \
+  --scheduling-policy priority
+```
+
+### Multiple Sequences per Prompt
+
+```bash
+python3 benchmarks/benchmark_prioritization.py \
+  --model meta-llama/Llama-2-7b-chat-hf \
+  --input-len 128 \
+  --output-len 64 \
+  --num-prompts 100 \
+  --scheduling-policy priority \
+  --n 2
+```
+
+</details>
+
+## 👁️ Example - Multi-Modal Benchmark
+
+<details>
+<summary>Show more</summary>
+
+<br/>
+
+Benchmark the performance of multi-modal requests in vLLM.
+
+### Images (ShareGPT4V)
+
+Start vLLM:
+
+```bash
+python -m vllm.entrypoints.openai.api_server \
+  --model Qwen/Qwen2.5-VL-7B-Instruct \
+  --dtype bfloat16 \
+  --limit-mm-per-prompt '{"image": 1}' \
+  --allowed-local-media-path /path/to/sharegpt4v/images
+```
+
+Send requests with images:
+
+```bash
+python benchmarks/benchmark_serving.py \
+  --backend openai-chat \
+  --model Qwen/Qwen2.5-VL-7B-Instruct \
+  --dataset-name sharegpt \
+  --dataset-path /path/to/ShareGPT4V/sharegpt4v_instruct_gpt4-vision_cap100k.json \
+  --num-prompts 100 \
+  --save-result \
+  --result-dir ~/vllm_benchmark_results \
+  --save-detailed \
+  --endpoint /v1/chat/completion
+```
+
+</details>

benchmarks/auto_tune/README.md

@@ -0,0 +1,145 @@
+# Automated vLLM Server Parameter Tuning
+
+This script automates the process of finding the optimal server parameter combination (`max-num-seqs` and `max-num-batched-tokens`) to maximize throughput for a vLLM server. It also supports additional constraints such as E2E latency and prefix cache hit rate.
+
+## Table of Contents
+
+- [Prerequisites](#prerequisites)
+- [Configuration](#configuration)
+- [How to Run](#how-to-run)
+- [Example Use Cases](#example-use-cases)
+- [Output](#output)
+- [How It Works](#how-it-works)
+
+## Prerequisites
+
+Before running the script, please ensure the following steps are completed:
+
+1. **Clone vLLM & Set Up Branch**: Clone the vLLM repository and check out to your desired branch.
+
+```bash
+git clone https://github.com/vllm-project/vllm.git
+cd vllm
+# git checkout <your-branch>
+```
+
+1. **Install Environment**: Install or update the correct running environment. For TPU usage, activate your `conda` environment and install the corresponding `torch` and `torch_xla` versions.
+
+2. **Model Configuration**: If you are using a customized model, ensure its configuration files are correctly placed and accessible.
+
+## Configuration
+
+You must set the following variables at the top of the script before execution.
+
+| Variable | Description | Example Value |
+| --- | --- | --- |
+| `BASE` | **Required.** The absolute path to the parent directory of your vLLM repository directory. | `"$HOME"` |
+| `MODEL` | **Required.** The Hugging Face model identifier to be served by vllm. | `"meta-llama/Llama-3.1-8B-Instruct"` |
+| `SYSTEM`| **Required.** The hardware you are running on. Choices: `TPU` or `GPU`. (For other systems, it might not support saving profiles) | `"TPU"` |
+| `TP` | **Required.** The tensor-parallelism size. | `1` |
+| `DOWNLOAD_DIR` | **Required.** Directory to download and load model weights from. | `""` (default download path) |
+| `INPUT_LEN` | **Required.** Request input length. | `4000` |
+| `OUTPUT_LEN` | **Required.** Request output length. | `16` |
+| `MAX_MODEL_LEN` | **Required.** Max model length. | `4096` |
+| `MIN_CACHE_HIT_PCT` | Prefix cache hit rate in percentage (0-100). Set to `0` to disable. | `60` |
+| `MAX_LATENCY_ALLOWED_MS` | The maximum allowed P99 end-to-end latency in milliseconds. Set to a very large number (e.g., `100000000000`) to effectively ignore the latency constraint. | `500` |
+| `NUM_SEQS_LIST` | A space-separated string of `max-num-seqs` values to test. | `"128 256"` |
+| `NUM_BATCHED_TOKENS_LIST` | A space-separated string of `max-num-batched-tokens` values to test. | `"1024 2048 4096"` |
+
+**Note**: The default `NUM_SEQS_LIST` and `NUM_BATCHED_TOKENS_LIST` are set for medium-sized inputs/outputs. For very short contexts (e.g., 20 input, 20 output tokens), you may need to test larger values for `max-num-seqs`.
+
+## How to Run
+
+1. **Configure**: Edit the script and set the variables in the [Configuration](#configuration) section.
+2. **Execute**: Run the script. Since the process can take a long time, it is highly recommended to use a terminal multiplexer like `tmux` or `screen` to prevent the script from stopping if your connection is lost.
+
+```bash
+cd <FOLDER_OF_THIS_SCRIPT>
+bash auto_tune.sh
+```
+
+    Please note that the `bash auto_tune.sh` command cannot contain full or partial path with keyword `vllm`, otherwise `pkill -f vllm` command will also kill this script itself.
+
+## Example Use Cases
+
+Here are a few examples of how to configure the script for different goals:
+
+### 1. Maximize Throughput (No Latency Constraint)
+
+- **Goal**: Find the best `max-num-seqs` and `max-num-batched-tokens` to get the highest possible throughput for 1800 input tokens and 20 output tokens.
+- **Configuration**:
+
+```bash
+INPUT_LEN=1800
+OUTPUT_LEN=20
+MAX_MODEL_LEN=2048
+MIN_CACHE_HIT_PCT=0
+MAX_LATENCY_ALLOWED_MS=100000000000 # A very large number
+```
+
+#### 2. Maximize Throughput with a Latency Requirement
+
+- **Goal**: Find the best server parameters when P99 end-to-end latency must be below 500ms.
+- **Configuration**:
+
+```bash
+INPUT_LEN=1800
+OUTPUT_LEN=20
+MAX_MODEL_LEN=2048
+MIN_CACHE_HIT_PCT=0
+MAX_LATENCY_ALLOWED_MS=500
+```
+
+#### 3. Maximize Throughput with Prefix Caching and Latency Requirements
+
+- **Goal**: Find the best server parameters assuming a 60% prefix cache hit rate and a latency requirement of 500ms.
+- **Configuration**:
+
+```bash
+INPUT_LEN=1800
+OUTPUT_LEN=20
+MAX_MODEL_LEN=2048
+MIN_CACHE_HIT_PCT=60
+MAX_LATENCY_ALLOWED_MS=500
+```
+
+## Output
+
+After the script finishes, you will find the results in a new, timestamped directory created inside `$BASE/auto-benchmark/`.
+
+- **Log Files**: The directory (`$BASE/auto-benchmark/YYYY_MM_DD_HH_MM/`) contains detailed logs for each run:
+    - `vllm_log_...txt`: The log output from the vLLM server for each parameter combination.
+    - `bm_log_...txt`: The log output from the `vllm bench serve` command for each benchmark run.
+
+- **Final Result Summary**: A file named `result.txt` is created in the log directory. It contains a summary of each tested combination and concludes with the overall best parameters found.
+
+```text
+# Example result.txt content
+hash:a1b2c3d4...
+max_num_seqs: 128, max_num_batched_tokens: 2048, request_rate: 10.0, e2el: 450.5, throughput: 9.8, goodput: 9.8
+max_num_seqs: 128, max_num_batched_tokens: 4096 does not meet latency requirement 500
+...
+best_max_num_seqs: 256, best_num_batched_tokens: 2048, best_throughput: 12.5, profile saved in: /home/user/vllm/auto-benchmark/2024_08_01_10_30/profile
+```
+
+  If it cannot find the best parameters, the final row will be `best_max_num_seqs: 0, best_num_batched_tokens: 0, best_throughput: 0`. This can be due to either the server not starting properly, or the latency requirement being too strict.
+
+- **Profiler Trace**: A directory named `profile` is created inside the log directory. It contains the profiler trace file (e.g., `.xplane.pb` for TPU or a `.json` trace for GPU) from the single best-performing run.
+
+## How It Works
+
+The script follows a systematic process to find the optimal parameters:
+
+1. **Find Max GPU Memory Utilization**: The script first determines the highest safe `gpu-memory-utilization` (starting from 0.98 and decreasing) that does not cause an Out-Of-Memory (OOM) error when launching the server. This ensures the benchmark runs use the maximum available memory without crashing.
+
+2. **Iterate and Benchmark**: It then enters a nested loop, iterating through every combination of `max-num-seqs` and `max-num-batched-tokens` provided in the configuration lists.
+
+3. **Latency-Aware Throughput Search**: For each parameter combination:
+    - The vLLM server is started.
+    - A benchmark is first run with an infinite request rate (`--request-rate inf`).
+    - If the resulting P99 E2E latency is within the `MAX_LATENCY_ALLOWED_MS` limit, this throughput is considered the maximum for this configuration.
+    - If the latency is too high, the script performs a search by iteratively decreasing the request rate until the latency constraint is met. This finds the highest sustainable throughput for the given parameters and latency requirement.
+
+4. **Track Best Result**: Throughout the process, the script tracks the parameter combination that has yielded the highest valid throughput so far.
+
+5. **Profile Collection**: For the best-performing run, the script saves the vLLM profiler output, which can be used for deep-dive performance analysis with tools like TensorBoard.

benchmarks/auto_tune/auto_tune.sh

@@ -1,43 +1,18 @@
 #!/bin/bash
 
-# This script aims to tune the best server parameter combinations to maximize throughput for given requirement. 
-# The current server parameter combination is  max_num_seqs and max_num_batched_tokens
-# It also supports additional requirement: e2e latency and prefix cache. 
-
-# Pre-requisite:
-# 1. Checkout to your branch, install/ update the correct running env. For TPU, activate conda env and install the corresponding torch, xla version. 
-# 2. If the model is customized, replace the MODEL's config with the customized config.
-# 3. Set variables (ALL REQUIRED)
-#   BASE: your directory for vllm repo
-#   MODEL: the model served by vllm
-#   TP: ways of tensor parallelism
-#   DOWNLOAD_DIR: directory to download and load model weights.
-#   INPUT_LEN: request input len
-#   OUTPUT_LEN: request output len
-#   MIN_CACHE_HIT_PCT: prefix cache rate
-#   MAX_LATENCY_ALLOWED_MS: (e2e) latency requirement. If there's no latency requirement, set it to a large number like 1000000000
-#   NUM_SEQS_LIST: a list of `max-num-seqs` you want to loop with.
-#   NUM_BATCHED_TOKENS_LIST: a list of `max-num-batched-tokens` you want to loop with.
-#   Note that the default NUM_SEQS_LIST and NUM_BATCHED_TOKENS_LIST are set for medium size input/output len, for extra short context (such as 20:20), you might need to include larger numbers in NUM_SEQS_LIST.
-# 4. Run the script, it might take a long time, you can use tmux to avoid the script stop if disconnection happens.
-# 5. The final result will be saved in RESULT file. 
-
-
-# Example use cases 
-# 1. Given input_len=1800, output_len=20, what's the best max_num_seqs and max_num_batched_tokens to get highest throughput?
-# Use INPUT_LEN=1800,  OUTPUT_LEN=20, MIN_CACHE_HIT_PCT=0, MAX_LATENCY_ALLOWED_MS=100000000000
-# 2. If we have latency requirement to be lower than 500ms, what's the best server parameter?
-# Use INPUT_LEN=1800,  OUTPUT_LEN=20, MIN_CACHE_HIT_PCT=0, MAX_LATENCY_ALLOWED_MS=500
-# 3. If we want to reach 60% prefix cache, what's the best server parameter? 
-# Use INPUT_LEN=1800,  OUTPUT_LEN=20, MIN_CACHE_HIT_PCT=60, MAX_LATENCY_ALLOWED_MS=500
+# This script aims to tune the best server parameter combinations to maximize throughput for given requirement.
+# See details in README (benchmarks/auto_tune/README.md).
 
 TAG=$(date +"%Y_%m_%d_%H_%M")
-BASE=""
+SCRIPT_DIR=$( cd -- "$( dirname -- "${BASH_SOURCE[0]}" )" &> /dev/null && pwd )
+BASE="$SCRIPT_DIR/../../.."
 MODEL="meta-llama/Llama-3.1-8B-Instruct"
+SYSTEM="TPU"
 TP=1
 DOWNLOAD_DIR=""
 INPUT_LEN=4000
 OUTPUT_LEN=16
+MAX_MODEL_LEN=4096
 MIN_CACHE_HIT_PCT=0
 MAX_LATENCY_ALLOWED_MS=100000000000
 NUM_SEQS_LIST="128 256"
@@ -45,12 +20,15 @@ NUM_BATCHED_TOKENS_LIST="512 1024 2048 4096"
 
 LOG_FOLDER="$BASE/auto-benchmark/$TAG"
 RESULT="$LOG_FOLDER/result.txt"
+PROFILE_PATH="$LOG_FOLDER/profile"
 
 echo "result file: $RESULT"
 echo "model: $MODEL"
 
 rm -rf $LOG_FOLDER
+rm -rf $PROFILE_PATH
 mkdir -p $LOG_FOLDER
+mkdir -p $PROFILE_PATH
 
 cd "$BASE/vllm"
 
@@ -60,36 +38,61 @@ current_hash=$(git rev-parse HEAD)
 echo "hash:$current_hash" >> "$RESULT"
 echo "current_hash: $current_hash"
 
+TOTAL_LEN=$((INPUT_LEN + OUTPUT_LEN))
+RED='\033[0;31m'
+if (( TOTAL_LEN > MAX_MODEL_LEN )); then
+    echo -e "${RED}FAILED: INPUT_LEN($INPUT_LEN) + OUTPUT_LEN($OUTPUT_LEN) = $TOTAL_LEN, which is > MAX_MODEL_LEN = $MAX_MODEL_LEN.\033[0m" >&2
+    exit 1
+fi
+
 best_throughput=0
 best_max_num_seqs=0
 best_num_batched_tokens=0
 best_goodput=0
+best_request_rate=0
 
 start_server() {
     local gpu_memory_utilization=$1
     local max_num_seqs=$2
     local max_num_batched_tokens=$3
     local vllm_log=$4
-    
-    pkill -f vllm
+    local profile_dir=$5
 
-    VLLM_USE_V1=1 VLLM_SERVER_DEV_MODE=1 vllm serve $MODEL \
-        --disable-log-requests \
-        --port 8004 \
-        --gpu-memory-utilization $gpu_memory_utilization \
-        --max-num-seqs $max_num_seqs \
-        --max-num-batched-tokens $max_num_batched_tokens \
-        --tensor-parallel-size $TP \
-        --enable-prefix-caching \
-        --load-format dummy \
-        --download-dir "$DOWNLOAD_DIR" \
-        --max-model-len $(( INPUT_LEN+OUTPUT_LEN )) > "$vllm_log" 2>&1 &
+    pkill -if vllm
+
+    # Define the common arguments as a bash array.
+    # Each argument and its value are separate elements.
+    local common_args_array=(
+        "$MODEL"
+        "--disable-log-requests"
+        "--port" "8004"
+        "--gpu-memory-utilization" "$gpu_memory_utilization"
+        "--max-num-seqs" "$max_num_seqs"
+        "--max-num-batched-tokens" "$max_num_batched_tokens"
+        "--tensor-parallel-size" "$TP"
+        "--enable-prefix-caching"
+        "--load-format" "dummy"
+        "--download-dir" "$DOWNLOAD_DIR"
+        "--max-model-len" "$MAX_MODEL_LEN"
+    )
+
+    # Use the array expansion "${common_args_array[@]}"
+    # This correctly passes each element as a separate argument.
+    if [[ -n "$profile_dir" ]]; then
+        # Start server with profiling enabled
+        VLLM_USE_V1=1 VLLM_SERVER_DEV_MODE=1 VLLM_TORCH_PROFILER_DIR=$profile_dir \
+            vllm serve "${common_args_array[@]}" > "$vllm_log" 2>&1 &
+    else
+        # Start server without profiling
+        VLLM_USE_V1=1 VLLM_SERVER_DEV_MODE=1 \
+            vllm serve "${common_args_array[@]}" > "$vllm_log" 2>&1 &
+    fi
 
     # wait for 10 minutes...
     server_started=0
-    for i in {1..60}; do  
+    for i in {1..60}; do
         RESPONSE=$(curl -s -X GET "http://0.0.0.0:8004/health" -w "%{http_code}" -o /dev/stdout)
-        STATUS_CODE=$(echo "$RESPONSE" | tail -n 1) 
+        STATUS_CODE=$(echo "$RESPONSE" | tail -n 1)
         if [[ "$STATUS_CODE" -eq 200 ]]; then
             server_started=1
             break
@@ -97,6 +100,7 @@ start_server() {
             sleep 10
         fi
     done
+
     if (( ! server_started )); then
         echo "server did not start within 10 minutes. Please check server log at $vllm_log".
         return 1
@@ -114,10 +118,11 @@ run_benchmark() {
     echo "vllm_log: $vllm_log"
     echo
     rm -f $vllm_log
-    pkill -f vllm
+    pkill -if vllm
 
     echo "starting server..."
-    start_server $gpu_memory_utilization $max_num_seqs $max_num_batched_tokens $vllm_log
+    # Call start_server without a profile_dir to avoid profiling overhead
+    start_server $gpu_memory_utilization $max_num_seqs $max_num_batched_tokens $vllm_log ""
     result=$?
     if [[ "$result" -eq 1 ]]; then
         echo "server failed to start. gpu_memory_utilization:$gpu_memory_utilization, max_num_seqs:$max_num_seqs, max_num_batched_tokens: $max_num_batched_tokens"
@@ -125,17 +130,19 @@ run_benchmark() {
         echo "server started."
     fi
     echo
-    
+
     echo "run benchmark test..."
     meet_latency_requirement=0
     # get a basic qps by using request-rate inf
     bm_log="$LOG_FOLDER/bm_log_${max_num_seqs}_${max_num_batched_tokens}_requestrate_inf.txt"
     prefix_len=$(( INPUT_LEN * MIN_CACHE_HIT_PCT / 100 ))
-    python benchmarks/benchmark_serving.py \
+    adjusted_input_len=$(( INPUT_LEN - prefix_len ))
+    # --profile flag is removed from this call
+    vllm bench serve \
         --backend vllm \
         --model $MODEL  \
         --dataset-name random \
-        --random-input-len $INPUT_LEN \
+        --random-input-len $adjusted_input_len \
         --random-output-len $OUTPUT_LEN \
         --ignore-eos \
         --disable-tqdm \
@@ -162,11 +169,11 @@ run_benchmark() {
             curl -X POST http://0.0.0.0:8004/reset_prefix_cache
             sleep 5
             bm_log="$LOG_FOLDER/bm_log_${max_num_seqs}_${max_num_batched_tokens}_requestrate_${request_rate}.txt"
-            python benchmarks/benchmark_serving.py \
+            vllm bench serve \
                 --backend vllm \
                 --model $MODEL  \
                 --dataset-name random \
-                --random-input-len $INPUT_LEN \
+                --random-input-len $adjusted_input_len \
                 --random-output-len $OUTPUT_LEN \
                 --ignore-eos \
                 --disable-tqdm \
@@ -195,6 +202,7 @@ run_benchmark() {
             best_max_num_seqs=$max_num_seqs
             best_num_batched_tokens=$max_num_batched_tokens
             best_goodput=$goodput
+            best_request_rate=$request_rate
         fi
     else
         echo "max_num_seqs: $max_num_seqs, max_num_batched_tokens: $max_num_batched_tokens does not meet latency requirement ${MAX_LATENCY_ALLOWED_MS}"
@@ -203,7 +211,7 @@ run_benchmark() {
 
     echo "best_max_num_seqs: $best_max_num_seqs, best_num_batched_tokens: $best_num_batched_tokens, best_throughput: $best_throughput"
 
-    pkill vllm
+    pkill -if vllm
     sleep 10
     printf '=%.0s' $(seq 1 20)
     return 0
@@ -216,7 +224,8 @@ read -r -a num_batched_tokens_list <<< "$NUM_BATCHED_TOKENS_LIST"
 gpu_memory_utilization=0.98
 find_gpu_memory_utilization=0
 while (( $(echo "$gpu_memory_utilization >= 0.9" | bc -l) )); do
-    start_server $gpu_memory_utilization "${num_seqs_list[-1]}" "${num_batched_tokens_list[-1]}" "$LOG_FOLDER/vllm_log_gpu_memory_utilization_$gpu_memory_utilization.log"
+    # Pass empty string for profile_dir argument
+    start_server $gpu_memory_utilization "${num_seqs_list[-1]}" "${num_batched_tokens_list[-1]}" "$LOG_FOLDER/vllm_log_gpu_memory_utilization_$gpu_memory_utilization.log" ""
     result=$?
     if [[ "$result" -eq 0 ]]; then
         find_gpu_memory_utilization=1
@@ -239,6 +248,45 @@ for num_seqs in "${num_seqs_list[@]}"; do
     done
 done
 echo "finish permutations"
-echo "best_max_num_seqs: $best_max_num_seqs, best_num_batched_tokens: $best_num_batched_tokens, best_throughput: $best_throughput"
-echo "best_max_num_seqs: $best_max_num_seqs, best_num_batched_tokens: $best_num_batched_tokens, best_throughput: $best_throughput" >> "$RESULT"
 
+# =================================================================================
+# FINAL PROFILING RUN FOR THE BEST CONFIGURATION
+# =================================================================================
+if (( $(echo "$best_throughput > 0" | bc -l) )); then
+    echo
+    echo "Benchmark tuning finished. Now running profiling on the best configuration found..."
+    echo "Best config: max_num_seqs: $best_max_num_seqs, max_num_batched_tokens: $best_num_batched_tokens, throughput: $best_throughput"
+    echo
+
+    vllm_log="$LOG_FOLDER/vllm_log_BEST_PROFILE.txt"
+    bm_log="$LOG_FOLDER/bm_log_BEST_PROFILE.txt"
+
+    # Start server with the best params and profiling ENABLED
+    echo "Starting server for profiling..."
+    start_server $gpu_memory_utilization $best_max_num_seqs $best_num_batched_tokens "$vllm_log" "$PROFILE_PATH"
+
+    # Run benchmark with the best params and the --profile flag
+    echo "Running benchmark with profiling..."
+    prefix_len=$(( INPUT_LEN * MIN_CACHE_HIT_PCT / 100 ))
+    adjusted_input_len=$(( INPUT_LEN - prefix_len ))
+    vllm bench serve \
+        --backend vllm \
+        --model $MODEL \
+        --dataset-name random \
+        --random-input-len $adjusted_input_len \
+        --random-output-len $OUTPUT_LEN \
+        --ignore-eos \
+        --disable-tqdm \
+        --request-rate $best_request_rate \
+        --percentile-metrics ttft,tpot,itl,e2el \
+        --goodput e2el:$MAX_LATENCY_ALLOWED_MS \
+        --num-prompts 100 \
+        --random-prefix-len $prefix_len \
+        --port 8004 \
+        --profile &> "$bm_log"
+else
+    echo "No configuration met the latency requirements. Skipping final profiling run."
+fi
+pkill -if vllm
+echo "best_max_num_seqs: $best_max_num_seqs, best_num_batched_tokens: $best_num_batched_tokens, best_throughput: $best_throughput, profile saved in: $PROFILE_PATH"
+echo "best_max_num_seqs: $best_max_num_seqs, best_num_batched_tokens: $best_num_batched_tokens, best_throughput: $best_throughput, profile saved in: $PROFILE_PATH" >> "$RESULT"

benchmarks/backend_request_func.py

@@ -31,7 +31,7 @@ class RequestFuncInput:
     model_name: Optional[str] = None
     logprobs: Optional[int] = None
     extra_body: Optional[dict] = None
-    multi_modal_content: Optional[dict] = None
+    multi_modal_content: Optional[dict | list[dict]] = None
     ignore_eos: bool = False
     language: Optional[str] = None
 
@@ -364,7 +364,15 @@ async def async_request_openai_chat_completions(
     ) as session:
         content = [{"type": "text", "text": request_func_input.prompt}]
         if request_func_input.multi_modal_content:
-            content.append(request_func_input.multi_modal_content)
+            mm_content = request_func_input.multi_modal_content
+            if isinstance(mm_content, list):
+                content.extend(mm_content)
+            elif isinstance(mm_content, dict):
+                content.append(mm_content)
+            else:
+                raise TypeError(
+                    "multi_modal_content must be a dict or list[dict] for openai-chat"
+                )
         payload = {
             "model": request_func_input.model_name
             if request_func_input.model_name
@@ -404,8 +412,14 @@ async def async_request_openai_chat_completions(
                         chunk_bytes = chunk_bytes.strip()
                         if not chunk_bytes:
                             continue
+                        chunk_bytes = chunk_bytes.decode("utf-8")
+                        # NOTE: SSE comments (often used as pings) start with a colon.
+                        # These are not JSON data payload and should be skipped.
+                        if chunk_bytes.startswith(":"):
+                            continue
+
+                        chunk = chunk_bytes.removeprefix("data: ")
 
-                        chunk = chunk_bytes.decode("utf-8").removeprefix("data: ")
                         if chunk != "[DONE]":
                             timestamp = time.perf_counter()
                             data = json.loads(chunk)
@@ -485,7 +499,10 @@ async def async_request_openai_audio(
             buffer.seek(0)
             return buffer
 
-        with to_bytes(*request_func_input.multi_modal_content["audio"]) as f:
+        mm_audio = request_func_input.multi_modal_content
+        if not isinstance(mm_audio, dict) or "audio" not in mm_audio:
+            raise TypeError("multi_modal_content must be a dict containing 'audio'")
+        with to_bytes(*mm_audio["audio"]) as f:
             form = aiohttp.FormData()
             form.add_field("file", f, content_type="audio/wav")
             for key, value in payload.items():

benchmarks/benchmark_block_pool.py

@@ -0,0 +1,74 @@
+# SPDX-License-Identifier: Apache-2.0
+# SPDX-FileCopyrightText: Copyright contributors to the vLLM project
+import gc
+
+from tabulate import tabulate
+
+from benchmark_utils import TimeCollector
+from vllm.utils import FlexibleArgumentParser
+from vllm.v1.core.block_pool import BlockPool
+
+
+def main(args):
+    rows = []
+    for allocate_block in args.allocate_blocks:
+        # Enforce a GC collect ahead to minimize the impact among runs
+        gc.collect()
+        block_pool = BlockPool(num_gpu_blocks=args.num_gpu_blocks, enable_caching=True)
+
+        get_blocks_times = TimeCollector(TimeCollector.US)
+        free_blocks_times = TimeCollector(TimeCollector.US)
+        for _ in range(args.num_iteration):
+            with get_blocks_times:
+                blocks = block_pool.get_new_blocks(allocate_block)
+            with free_blocks_times:
+                block_pool.free_blocks(blocks)
+
+        rows.append(
+            [get_blocks_times.cnt, args.num_gpu_blocks, allocate_block]
+            + get_blocks_times.dump_avg_max()
+            + free_blocks_times.dump_avg_max()
+        )
+
+    print(
+        tabulate(
+            rows,
+            headers=[
+                "Iterations",
+                "Total\nBlocks",
+                "Allocated\nBlocks",
+                "Get Blocks\nAvg (us)",
+                "Get Blocks\nMax (us)",
+                "Free Blocks\nAvg (us)",
+                "Free Blocks\nMax (us)",
+            ],
+            tablefmt="grid",
+            floatfmt=".3f",
+        )
+    )
+
+
+def invoke_main() -> None:
+    parser = FlexibleArgumentParser(
+        description="Benchmark the performance of BlockPool for KV Cache."
+    )
+    parser.add_argument("--num-gpu-blocks", type=int, default=100000)
+    parser.add_argument(
+        "--num-iteration",
+        type=int,
+        default=1000,
+        help="Number of iterations to run to stablize final data readings",
+    )
+    parser.add_argument(
+        "--allocate-blocks",
+        type=int,
+        nargs="*",
+        default=[10, 50, 100, 500, 1000],
+        help="Number of blocks to allocate",
+    )
+    args = parser.parse_args()
+    main(args)
+
+
+if __name__ == "__main__":
+    invoke_main()  # pragma: no cover

benchmarks/benchmark_dataset.py

@@ -52,7 +52,7 @@ class SampleRequest:
     prompt: Union[str, Any]
     prompt_len: int
     expected_output_len: int
-    multi_modal_data: Optional[Union[MultiModalDataDict, dict]] = None
+    multi_modal_data: Optional[Union[MultiModalDataDict, dict, list[dict]]] = None
     lora_request: Optional[LoRARequest] = None
 
 
@@ -324,6 +324,9 @@ class RandomDataset(BenchmarkDataset):
         input_low = int(real_input_len * (1 - range_ratio))
         input_high = int(real_input_len * (1 + range_ratio))
         output_low = int(output_len * (1 - range_ratio))
+        # Ensure the lower bound for output length is at least 1 to prevent
+        # sampling 0 tokens, which can cause request failures.
+        output_low = max(output_low, 1)
         output_high = int(output_len * (1 + range_ratio))
 
         # Add logging for debugging
@@ -349,11 +352,12 @@ class RandomDataset(BenchmarkDataset):
             # [1650, 939, 486] -> ['Ġcall', 'sh', 'ere']
             # To avoid uncontrolled change of the prompt length,
             # the encoded sequence is truncated before being decode again.
+            total_input_len = prefix_len + int(input_lens[i])
             re_encoded_sequence = tokenizer.encode(prompt, add_special_tokens=False)[
-                : input_lens[i]
+                :total_input_len
             ]
             prompt = tokenizer.decode(re_encoded_sequence)
-            total_input_len = prefix_len + int(input_lens[i])
+            total_input_len = len(re_encoded_sequence)
             requests.append(
                 SampleRequest(
                     prompt=prompt,
@@ -426,14 +430,20 @@ class ShareGPTDataset(BenchmarkDataset):
                 skip_min_output_len_check=output_len is not None,
             ):
                 continue
+            # TODO: Also support ShareGPT4Video.
+            if image_path := entry.get("image"):
+                mm_content = process_image(image_path)
+            else:
+                mm_content = None
             if enable_multimodal_chat:
-                prompt = self.apply_multimodal_chat_transformation(prompt, None)
+                prompt = self.apply_multimodal_chat_transformation(prompt, mm_content)
             samples.append(
                 SampleRequest(
                     prompt=prompt,
                     prompt_len=prompt_len,
                     expected_output_len=new_output_len,
                     lora_request=lora_request,
+                    multi_modal_data=mm_content,
                 )
             )
         self.maybe_oversample_requests(samples, num_requests)
@@ -700,6 +710,7 @@ class HuggingFaceDataset(BenchmarkDataset):
         self,
         dataset_path: str,
         dataset_split: str,
+        no_stream: bool = False,
         dataset_subset: Optional[str] = None,
         **kwargs,
     ) -> None:
@@ -707,6 +718,7 @@ class HuggingFaceDataset(BenchmarkDataset):
 
         self.dataset_split = dataset_split
         self.dataset_subset = dataset_subset
+        self.load_stream = not no_stream
         self.load_data()
 
     def load_data(self) -> None:
@@ -715,7 +727,7 @@ class HuggingFaceDataset(BenchmarkDataset):
             self.dataset_path,
             name=self.dataset_subset,
             split=self.dataset_split,
-            streaming=True,
+            streaming=self.load_stream,
         )
         self.data = self.data.shuffle(seed=self.random_seed)
 

benchmarks/benchmark_latency.py

@@ -11,6 +11,7 @@ from typing import Any, Optional
 
 import numpy as np
 from tqdm import tqdm
+from typing_extensions import deprecated
 
 import vllm.envs as envs
 from benchmark_utils import convert_to_pytorch_benchmark_format, write_to_json
@@ -34,6 +35,10 @@ def save_to_pytorch_benchmark_format(
         write_to_json(pt_file, pt_records)
 
 
+@deprecated(
+    "benchmark_latency.py is deprecated and will be removed in a "
+    "future version. Please use 'vllm bench latency' instead.",
+)
 def main(args: argparse.Namespace):
     print(args)
 
@@ -123,7 +128,7 @@ def main(args: argparse.Namespace):
         save_to_pytorch_benchmark_format(args, results)
 
 
-if __name__ == "__main__":
+def create_argument_parser():
     parser = FlexibleArgumentParser(
         description="Benchmark the latency of processing a single batch of "
         "requests till completion."
@@ -171,6 +176,12 @@ if __name__ == "__main__":
     # V1 enables prefix caching by default which skews the latency
     # numbers. We need to disable prefix caching by default.
     parser.set_defaults(enable_prefix_caching=False)
+
+    return parser
+
+
+if __name__ == "__main__":
+    parser = create_argument_parser()
     args = parser.parse_args()
     if args.profile and not envs.VLLM_TORCH_PROFILER_DIR:
         raise OSError(

benchmarks/benchmark_long_document_qa_throughput.py

@@ -142,7 +142,7 @@ def main(args):
     )
 
 
-if __name__ == "__main__":
+def create_argument_parser():
     parser = FlexibleArgumentParser(
         description="Benchmark the performance with or "
         "without automatic prefix caching."
@@ -192,5 +192,11 @@ if __name__ == "__main__":
     )
 
     parser = EngineArgs.add_cli_args(parser)
+
+    return parser
+
+
+if __name__ == "__main__":
+    parser = create_argument_parser()
     args = parser.parse_args()
     main(args)

benchmarks/benchmark_ngram_proposer.py

@@ -0,0 +1,112 @@
+# SPDX-License-Identifier: Apache-2.0
+# SPDX-FileCopyrightText: Copyright contributors to the vLLM project
+import gc
+
+import numpy as np
+from tabulate import tabulate
+
+from benchmark_utils import TimeCollector
+from vllm.config import ModelConfig, SpeculativeConfig, VllmConfig
+from vllm.utils import FlexibleArgumentParser
+from vllm.v1.spec_decode.ngram_proposer import NgramProposer
+
+
+def main(args):
+    rows = []
+    for max_ngram in args.max_ngram:
+        collector = TimeCollector(TimeCollector.US)
+
+        model_config = ModelConfig(
+            model="facebook/opt-125m",
+            task="generate",
+            max_model_len=args.num_token + args.num_spec_token,
+            tokenizer="facebook/opt-125m",
+            tokenizer_mode="auto",
+            dtype="auto",
+            seed=None,
+            trust_remote_code=False,
+        )
+        proposer = NgramProposer(
+            vllm_config=VllmConfig(
+                model_config=model_config,
+                speculative_config=SpeculativeConfig(
+                    prompt_lookup_min=args.min_ngram,
+                    prompt_lookup_max=max_ngram,
+                    num_speculative_tokens=args.num_spec_token,
+                    method="ngram",
+                ),
+            )
+        )
+
+        # Warm up
+        proposer.propose(np.random.randint(0, 20, (args.num_token,)))
+
+        gc.collect()
+        for _ in range(args.num_iteration):
+            tokens = np.random.randint(0, 20, (args.num_req, args.num_token))
+            with collector:
+                for i in range(args.num_req):
+                    proposer.propose(tokens[i, :])
+        rows.append(
+            [args.num_req, args.num_token, args.min_ngram, max_ngram]
+            + collector.dump_avg_max()
+        )
+
+    print(
+        tabulate(
+            rows,
+            headers=[
+                "# Request",
+                "# Token",
+                "Min Ngram",
+                "Max Ngram",
+                "Avg (us)",
+                "Max (us)",
+            ],
+            tablefmt="grid",
+            floatfmt=".3f",
+        )
+    )
+
+
+def invoke_main() -> None:
+    parser = FlexibleArgumentParser(
+        description="Benchmark the performance of N-gram speculative decode drafting"
+    )
+    parser.add_argument(
+        "--num-iteration",
+        type=int,
+        default=100,
+        help="Number of iterations to run to stablize final data readings",
+    )
+    parser.add_argument(
+        "--num-req", type=int, default=128, help="Number of requests in the batch"
+    )
+    parser.add_argument(
+        "--num-token", type=int, default=1500, help="Number of tokens for each request"
+    )
+    parser.add_argument(
+        "--min-ngram",
+        type=int,
+        default=3,
+        help="Minimum n-gram to match",
+    )
+    parser.add_argument(
+        "--max-ngram",
+        type=int,
+        nargs="*",
+        default=[5, 7, 10, 15, 20],
+        help="Maximum n-gram to match",
+    )
+    parser.add_argument(
+        "--num-spec-token",
+        type=int,
+        default=3,
+        help="Number of speculative tokens to generate",
+    )
+    args = parser.parse_args()
+    main(args)
+
+
+if __name__ == "__main__":
+    invoke_main()  # pragma: no cover

benchmarks/benchmark_prefix_caching.py

@@ -218,7 +218,7 @@ def main(args):
     )
 
 
-if __name__ == "__main__":
+def create_argument_parser():
     parser = FlexibleArgumentParser(
         description="Benchmark the performance with or without "
         "automatic prefix caching."
@@ -268,5 +268,11 @@ if __name__ == "__main__":
     )
 
     parser = EngineArgs.add_cli_args(parser)
+
+    return parser
+
+
+if __name__ == "__main__":
+    parser = create_argument_parser()
     args = parser.parse_args()
     main(args)

benchmarks/benchmark_prioritization.py

@@ -161,7 +161,7 @@ def main(args: argparse.Namespace):
             json.dump(results, f, indent=4)
 
 
-if __name__ == "__main__":
+def create_argument_parser():
     parser = FlexibleArgumentParser(description="Benchmark the throughput.")
     parser.add_argument(
         "--backend", type=str, choices=["vllm", "hf", "mii"], default="vllm"
@@ -204,6 +204,12 @@ if __name__ == "__main__":
     )
 
     parser = EngineArgs.add_cli_args(parser)
+
+    return parser
+
+
+if __name__ == "__main__":
+    parser = create_argument_parser()
     args = parser.parse_args()
     if args.tokenizer is None:
         args.tokenizer = args.model

benchmarks/benchmark_serving.py

@@ -5,8 +5,7 @@ r"""Benchmark online serving throughput.
 On the server side, run one of the following commands:
     vLLM OpenAI API server
     vllm serve <your_model> \
-        --swap-space 16 \
-        --disable-log-requests
+        --swap-space 16
 
 On the client side, run:
     python benchmarks/benchmark_serving.py \
@@ -30,14 +29,15 @@ import os
 import random
 import time
 import warnings
-from collections.abc import AsyncGenerator, Iterable
+from collections.abc import Iterable
 from dataclasses import dataclass
 from datetime import datetime
-from typing import Any, Optional
+from typing import Any, Literal, Optional
 
 import numpy as np
 from tqdm.asyncio import tqdm
 from transformers import PreTrainedTokenizerBase
+from typing_extensions import deprecated
 
 from backend_request_func import (
     ASYNC_REQUEST_FUNCS,
@@ -73,6 +73,7 @@ from benchmark_dataset import (
     VisionArenaDataset,
 )
 from benchmark_utils import convert_to_pytorch_benchmark_format, write_to_json
+from vllm.benchmarks.serve import get_request
 
 MILLISECONDS_TO_SECONDS_CONVERSION = 1000
 
@@ -107,51 +108,6 @@ class BenchmarkMetrics:
     percentiles_e2el_ms: list[tuple[float, float]]
 
 
-async def get_request(
-    input_requests: list[SampleRequest],
-    request_rate: float,
-    burstiness: float = 1.0,
-) -> AsyncGenerator[SampleRequest, None]:
-    """
-    Asynchronously generates requests at a specified rate
-    with OPTIONAL burstiness.
-
-    Args:
-        input_requests:
-            A list of input requests, each represented as a SampleRequest.
-        request_rate:
-            The rate at which requests are generated (requests/s).
-        burstiness (optional):
-            The burstiness factor of the request generation.
-            Only takes effect when request_rate is not inf.
-            Default value is 1, which follows a Poisson process.
-            Otherwise, the request intervals follow a gamma distribution.
-            A lower burstiness value (0 < burstiness < 1) results
-            in more bursty requests, while a higher burstiness value
-            (burstiness > 1) results in a more uniform arrival of requests.
-    """
-    input_requests: Iterable[SampleRequest] = iter(input_requests)
-
-    # Calculate scale parameter theta to maintain the desired request_rate.
-    assert burstiness > 0, (
-        f"A positive burstiness factor is expected, but given {burstiness}."
-    )
-    theta = 1.0 / (request_rate * burstiness)
-
-    for request in input_requests:
-        yield request
-
-        if request_rate == float("inf"):
-            # If the request rate is infinity, then we don't need to wait.
-            continue
-
-        # Sample the request interval from the gamma distribution.
-        # If burstiness is 1, it follows exponential distribution.
-        interval = np.random.gamma(shape=burstiness, scale=theta)
-        # The next request will be sent after the interval.
-        await asyncio.sleep(interval)
-
-
 def calculate_metrics(
     input_requests: list[SampleRequest],
     outputs: list[RequestFuncOutput],
@@ -290,6 +246,9 @@ async def benchmark(
     max_concurrency: Optional[int],
     lora_modules: Optional[Iterable[str]],
     extra_body: Optional[dict],
+    ramp_up_strategy: Optional[Literal["linear", "exponential"]] = None,
+    ramp_up_start_rps: Optional[int] = None,
+    ramp_up_end_rps: Optional[int] = None,
 ):
     if backend in ASYNC_REQUEST_FUNCS:
         request_func = ASYNC_REQUEST_FUNCS[backend]
@@ -304,7 +263,14 @@ async def benchmark(
         input_requests[0].multi_modal_data,
     )
 
-    assert test_mm_content is None or isinstance(test_mm_content, dict)
+    assert (
+        test_mm_content is None
+        or isinstance(test_mm_content, dict)
+        or (
+            isinstance(test_mm_content, list)
+            and all(isinstance(item, dict) for item in test_mm_content)
+        )
+    ), "multi_modal_data must be a dict or list[dict]"
     test_input = RequestFuncInput(
         model=model_id,
         model_name=model_name,
@@ -353,7 +319,15 @@ async def benchmark(
 
     distribution = "Poisson process" if burstiness == 1.0 else "Gamma distribution"
 
-    print(f"Traffic request rate: {request_rate}")
+    if ramp_up_strategy is not None:
+        print(
+            f"Traffic ramp-up strategy: {ramp_up_strategy}. Will increase "
+            f"RPS from {ramp_up_start_rps} to {ramp_up_end_rps} RPS over "
+            "the duration of the benchmark."
+        )
+    else:
+        print(f"Traffic request rate: {request_rate} RPS.")
+
     print(f"Burstiness factor: {burstiness} ({distribution})")
     print(f"Maximum request concurrency: {max_concurrency}")
 
@@ -373,7 +347,34 @@ async def benchmark(
 
     benchmark_start_time = time.perf_counter()
     tasks: list[asyncio.Task] = []
-    async for request in get_request(input_requests, request_rate, burstiness):
+
+    rps_change_events = []
+    last_int_rps = -1
+    if ramp_up_strategy is not None and ramp_up_start_rps is not None:
+        last_int_rps = ramp_up_start_rps
+        rps_change_events.append(
+            {
+                "rps": last_int_rps,
+                "timestamp": datetime.now().isoformat(),
+            }
+        )
+
+    async for request, current_request_rate in get_request(
+        input_requests,
+        request_rate,
+        burstiness,
+        ramp_up_strategy,
+        ramp_up_start_rps,
+        ramp_up_end_rps,
+    ):
+        if ramp_up_strategy is not None:
+            current_int_rps = int(current_request_rate)
+            if current_int_rps > last_int_rps:
+                timestamp = datetime.now().isoformat()
+                for rps_val in range(last_int_rps + 1, current_int_rps + 1):
+                    rps_change_events.append({"rps": rps_val, "timestamp": timestamp})
+                last_int_rps = current_int_rps
+
         prompt, prompt_len, output_len, mm_content = (
             request.prompt,
             request.prompt_len,
@@ -397,27 +398,10 @@ async def benchmark(
             ignore_eos=ignore_eos,
             extra_body=extra_body,
         )
-        tasks.append(
-            asyncio.create_task(
-                limited_request_func(request_func_input=request_func_input, pbar=pbar)
-            )
-        )
+        task = limited_request_func(request_func_input=request_func_input, pbar=pbar)
+        tasks.append(asyncio.create_task(task))
     outputs: list[RequestFuncOutput] = await asyncio.gather(*tasks)
 
-    if profile:
-        print("Stopping profiler...")
-        profile_input = RequestFuncInput(
-            model=model_id,
-            prompt=test_prompt,
-            api_url=base_url + "/stop_profile",
-            prompt_len=test_prompt_len,
-            output_len=test_output_len,
-            logprobs=logprobs,
-        )
-        profile_output = await request_func(request_func_input=profile_input)
-        if profile_output.success:
-            print("Profiler stopped")
-
     if pbar is not None:
         pbar.close()
 
@@ -435,6 +419,10 @@ async def benchmark(
 
     print("{s:{c}^{n}}".format(s=" Serving Benchmark Result ", n=50, c="="))
     print("{:<40} {:<10}".format("Successful requests:", metrics.completed))
+    if max_concurrency is not None:
+        print("{:<40} {:<10}".format("Maximum request concurrency:", max_concurrency))
+    if request_rate != float("inf"):
+        print("{:<40} {:<10.2f}".format("Request rate configured (RPS):", request_rate))
     print("{:<40} {:<10.2f}".format("Benchmark duration (s):", benchmark_duration))
     print("{:<40} {:<10}".format("Total input tokens:", metrics.total_input))
     print("{:<40} {:<10}".format("Total generated tokens:", metrics.total_output))
@@ -466,7 +454,7 @@ async def benchmark(
         "total_input_tokens": metrics.total_input,
         "total_output_tokens": metrics.total_output,
         "request_throughput": metrics.request_throughput,
-        "request_goodput:": metrics.request_goodput if goodput_config_dict else None,
+        "request_goodput": metrics.request_goodput if goodput_config_dict else None,
         "output_throughput": metrics.output_throughput,
         "total_token_throughput": metrics.total_token_throughput,
         "input_lens": [output.prompt_len for output in outputs],
@@ -477,6 +465,9 @@ async def benchmark(
         "errors": [output.error for output in outputs],
     }
 
+    if rps_change_events:
+        result["rps_change_events"] = rps_change_events
+
     def process_one_metric(
         # E.g., "ttft"
         metric_attribute_name: str,
@@ -523,6 +514,20 @@ async def benchmark(
 
     print("=" * 50)
 
+    if profile:
+        print("Stopping profiler...")
+        profile_input = RequestFuncInput(
+            model=model_id,
+            prompt=test_prompt,
+            api_url=base_url + "/stop_profile",
+            prompt_len=test_prompt_len,
+            output_len=test_output_len,
+            logprobs=logprobs,
+        )
+        profile_output = await request_func(request_func_input=profile_input)
+        if profile_output.success:
+            print("Profiler stopped")
+
     return result
 
 
@@ -599,6 +604,10 @@ def save_to_pytorch_benchmark_format(
         write_to_json(pt_file, pt_records)
 
 
+@deprecated(
+    "benchmark_serving.py is deprecated and will be removed in a future "
+    "version. Please use 'vllm bench serve' instead.",
+)
 def main(args: argparse.Namespace):
     print(args)
     random.seed(args.seed)
@@ -610,6 +619,26 @@ def main(args: argparse.Namespace):
     tokenizer_id = args.tokenizer if args.tokenizer is not None else args.model
     tokenizer_mode = args.tokenizer_mode
 
+    # Validate ramp-up arguments
+    if args.ramp_up_strategy is not None:
+        if args.request_rate != float("inf"):
+            raise ValueError(
+                "When using ramp-up, do not specify --request-rate. "
+                "The request rate will be controlled by ramp-up parameters. "
+                "Please remove the --request-rate argument."
+            )
+        if args.ramp_up_start_rps is None or args.ramp_up_end_rps is None:
+            raise ValueError(
+                "When using --ramp-up-strategy, both --ramp-up-start-rps and "
+                "--ramp-up-end-rps must be specified"
+            )
+        if args.ramp_up_start_rps < 0 or args.ramp_up_end_rps < 0:
+            raise ValueError("Ramp-up start and end RPS must be non-negative")
+        if args.ramp_up_start_rps > args.ramp_up_end_rps:
+            raise ValueError("Ramp-up start RPS must be less than end RPS")
+        if args.ramp_up_strategy == "exponential" and args.ramp_up_start_rps == 0:
+            raise ValueError("For exponential ramp-up, the start RPS cannot be 0.")
+
     if args.base_url is not None:
         api_url = f"{args.base_url}{args.endpoint}"
         base_url = f"{args.base_url}"
@@ -717,6 +746,7 @@ def main(args: argparse.Namespace):
             dataset_subset=args.hf_subset,
             dataset_split=args.hf_split,
             random_seed=args.seed,
+            no_stream=args.no_stream,
         ).sample(
             num_requests=args.num_prompts,
             tokenizer=tokenizer,
@@ -802,6 +832,9 @@ def main(args: argparse.Namespace):
             max_concurrency=args.max_concurrency,
             lora_modules=args.lora_modules,
             extra_body=sampling_params,
+            ramp_up_strategy=args.ramp_up_strategy,
+            ramp_up_start_rps=args.ramp_up_start_rps,
+            ramp_up_end_rps=args.ramp_up_end_rps,
         )
     )
 
@@ -834,6 +867,11 @@ def main(args: argparse.Namespace):
         result_json["burstiness"] = args.burstiness
         result_json["max_concurrency"] = args.max_concurrency
 
+        if args.ramp_up_strategy is not None:
+            result_json["ramp_up_strategy"] = args.ramp_up_strategy
+            result_json["ramp_up_start_rps"] = args.ramp_up_start_rps
+            result_json["ramp_up_end_rps"] = args.ramp_up_end_rps
+
         # Merge with benchmark result
         result_json = {**result_json, **benchmark_result}
 
@@ -859,7 +897,10 @@ def main(args: argparse.Namespace):
             if args.max_concurrency is not None
             else ""
         )
-        file_name = f"{backend}-{args.request_rate}qps{max_concurrency_str}-{base_model_id}-{current_dt}.json"  # noqa
+        if args.ramp_up_strategy is not None:
+            file_name = f"{backend}-ramp-up-{args.ramp_up_strategy}-{args.ramp_up_start_rps}qps-{args.ramp_up_end_rps}qps{max_concurrency_str}-{base_model_id}-{current_dt}.json"  # noqa
+        else:
+            file_name = f"{backend}-{args.request_rate}qps{max_concurrency_str}-{base_model_id}-{current_dt}.json"  # noqa
         if args.result_filename:
             file_name = args.result_filename
         if args.result_dir:
@@ -875,7 +916,7 @@ def main(args: argparse.Namespace):
         save_to_pytorch_benchmark_format(args, result_json, file_name)
 
 
-if __name__ == "__main__":
+def create_argument_parser():
     parser = FlexibleArgumentParser(
         description="Benchmark the online serving throughput."
     )
@@ -914,6 +955,11 @@ if __name__ == "__main__":
         help="Path to the sharegpt/sonnet dataset. "
         "Or the huggingface dataset ID if using HF dataset.",
     )
+    parser.add_argument(
+        "--no-stream",
+        action="store_true",
+        help="Do not load the dataset in streaming mode.",
+    )
     parser.add_argument(
         "--max-concurrency",
         type=int,
@@ -1225,6 +1271,35 @@ if __name__ == "__main__":
         "script chooses a LoRA module at random.",
     )
 
-    args = parser.parse_args()
+    parser.add_argument(
+        "--ramp-up-strategy",
+        type=str,
+        default=None,
+        choices=["linear", "exponential"],
+        help="The ramp-up strategy. This would be used to "
+        "ramp up the request rate from initial RPS to final "
+        "RPS rate (specified by --ramp-up-start-rps and --ramp-up-end-rps). "
+        "over the duration of the benchmark.",
+    )
+    parser.add_argument(
+        "--ramp-up-start-rps",
+        type=int,
+        default=None,
+        help="The starting request rate for ramp-up (RPS). "
+        "Needs to be specified when --ramp-up-strategy is used.",
+    )
+    parser.add_argument(
+        "--ramp-up-end-rps",
+        type=int,
+        default=None,
+        help="The ending request rate for ramp-up (RPS). "
+        "Needs to be specified when --ramp-up-strategy is used.",
+    )
 
+    return parser
+
+
+if __name__ == "__main__":
+    parser = create_argument_parser()
+    args = parser.parse_args()
     main(args)

benchmarks/benchmark_serving_structured_output.py

@@ -4,7 +4,7 @@ r"""Benchmark online serving throughput with structured outputs.
 
 On the server side, run one of the following commands:
     (vLLM OpenAI API server)
-    vllm serve <your_model> --disable-log-requests
+    vllm serve <your_model>
 
 On the client side, run:
     python benchmarks/benchmark_serving_structured_output.py \
@@ -538,20 +538,6 @@ async def benchmark(
         )
     outputs: list[RequestFuncOutput] = await asyncio.gather(*tasks)
 
-    if profile:
-        print("Stopping profiler...")
-        profile_input = RequestFuncInput(
-            model=model_id,
-            prompt=test_request.prompt,
-            api_url=base_url + "/stop_profile",
-            prompt_len=test_request.prompt_len,
-            output_len=test_request.expected_output_len,
-            extra_body={test_request.structure_type: test_request.schema},
-        )
-        profile_output = await request_func(request_func_input=profile_input)
-        if profile_output.success:
-            print("Profiler stopped")
-
     if pbar is not None:
         pbar.close()
 
@@ -569,6 +555,10 @@ async def benchmark(
 
     print("{s:{c}^{n}}".format(s=" Serving Benchmark Result ", n=50, c="="))
     print("{:<40} {:<10}".format("Successful requests:", metrics.completed))
+    if max_concurrency is not None:
+        print("{:<40} {:<10}".format("Maximum request concurrency:", max_concurrency))
+    if request_rate != float("inf"):
+        print("{:<40} {:<10.2f}".format("Request rate configured (RPS):", request_rate))
     print("{:<40} {:<10.2f}".format("Benchmark duration (s):", benchmark_duration))
     print("{:<40} {:<10}".format("Total input tokens:", metrics.total_input))
     print("{:<40} {:<10}".format("Total generated tokens:", metrics.total_output))
@@ -666,6 +656,20 @@ async def benchmark(
 
     print("=" * 50)
 
+    if profile:
+        print("Stopping profiler...")
+        profile_input = RequestFuncInput(
+            model=model_id,
+            prompt=test_request.prompt,
+            api_url=base_url + "/stop_profile",
+            prompt_len=test_request.prompt_len,
+            output_len=test_request.expected_output_len,
+            extra_body={test_request.structure_type: test_request.schema},
+        )
+        profile_output = await request_func(request_func_input=profile_input)
+        if profile_output.success:
+            print("Profiler stopped")
+
     return result, ret
 
 
@@ -850,7 +854,7 @@ def main(args: argparse.Namespace):
             json.dump(results, outfile, indent=4)
 
 
-if __name__ == "__main__":
+def create_argument_parser():
     parser = FlexibleArgumentParser(
         description="Benchmark the online serving throughput."
     )
@@ -1034,5 +1038,10 @@ if __name__ == "__main__":
         help="Ratio of Structured Outputs requests",
     )
 
+    return parser
+
+
+if __name__ == "__main__":
+    parser = create_argument_parser()
     args = parser.parse_args()
     main(args)

benchmarks/benchmark_throughput.py

@@ -15,6 +15,7 @@ import torch
 import uvloop
 from tqdm import tqdm
 from transformers import AutoModelForCausalLM, AutoTokenizer, PreTrainedTokenizerBase
+from typing_extensions import deprecated
 
 from benchmark_dataset import (
     AIMODataset,
@@ -97,7 +98,7 @@ def run_vllm(
         assert lora_requests is None, "BeamSearch API does not support LoRA"
         prompts = [request.prompt for request in requests]
         # output_len should be the same for all requests.
-        output_len = requests[0][2]
+        output_len = requests[0].expected_output_len
         for request in requests:
             assert request.expected_output_len == output_len
         start = time.perf_counter()
@@ -167,7 +168,8 @@ async def run_vllm_async(
     from vllm import SamplingParams
 
     async with build_async_engine_client_from_engine_args(
-        engine_args, disable_frontend_multiprocessing
+        engine_args,
+        disable_frontend_multiprocessing=disable_frontend_multiprocessing,
     ) as llm:
         model_config = await llm.get_model_config()
         assert all(
@@ -356,6 +358,7 @@ def get_requests(args, tokenizer):
     elif args.dataset_name == "burstgpt":
         dataset_cls = BurstGPTDataset
     elif args.dataset_name == "hf":
+        common_kwargs["no_stream"] = args.no_stream
         if args.dataset_path in VisionArenaDataset.SUPPORTED_DATASET_PATHS:
             dataset_cls = VisionArenaDataset
             common_kwargs["dataset_subset"] = None
@@ -380,6 +383,10 @@ def get_requests(args, tokenizer):
     return dataset_cls(**common_kwargs).sample(**sample_kwargs)
 
 
+@deprecated(
+    "benchmark_throughput.py is deprecated and will be removed in a "
+    "future version. Please use 'vllm bench throughput' instead.",
+)
 def main(args: argparse.Namespace):
     if args.seed is None:
         args.seed = 0
@@ -595,7 +602,7 @@ def validate_args(args):
         )
 
 
-if __name__ == "__main__":
+def create_argument_parser():
     parser = FlexibleArgumentParser(description="Benchmark the throughput.")
     parser.add_argument(
         "--backend",
@@ -610,6 +617,11 @@ if __name__ == "__main__":
         help="Name of the dataset to benchmark on.",
         default="sharegpt",
     )
+    parser.add_argument(
+        "--no-stream",
+        action="store_true",
+        help="Do not load the dataset in streaming mode.",
+    )
     parser.add_argument(
         "--dataset",
         type=str,
@@ -717,6 +729,12 @@ if __name__ == "__main__":
     )
 
     parser = AsyncEngineArgs.add_cli_args(parser)
+
+    return parser
+
+
+if __name__ == "__main__":
+    parser = create_argument_parser()
     args = parser.parse_args()
     if args.tokenizer is None:
         args.tokenizer = args.model

benchmarks/benchmark_utils.py

@@ -1,11 +1,12 @@
 # SPDX-License-Identifier: Apache-2.0
 # SPDX-FileCopyrightText: Copyright contributors to the vLLM project
-
 import argparse
 import json
 import math
 import os
-from typing import Any
+import time
+from types import TracebackType
+from typing import Any, Optional, Union
 
 
 def convert_to_pytorch_benchmark_format(
@@ -72,3 +73,53 @@ def write_to_json(filename: str, records: list) -> None:
             cls=InfEncoder,
             default=lambda o: f"<{type(o).__name__} object is not JSON serializable>",
         )
+
+
+# Collect time and generate time metrics
+#
+# Example Usage:
+#   collector = TimeCollector(TimeCollector.US)
+#   for _ in range(total_iteration):
+#      with collector:
+#          ...
+#   collector.dump_avg_max()
+class TimeCollector:
+    NS: int = 1
+    US: int = NS * 1000
+    MS: int = US * 1000
+    S: int = MS * 1000
+
+    def __init__(self, scale: int) -> None:
+        self.cnt: int = 0
+        self._sum: int = 0
+        self._max: Optional[int] = None
+        self.scale = scale
+        self.start_time: int = time.monotonic_ns()
+
+    def collect(self, v: int) -> None:
+        self.cnt += 1
+        self._sum += v
+        if self._max is None:
+            self._max = v
+        else:
+            self._max = max(self._max, v)
+
+    def avg(self) -> Union[float, str]:
+        return self._sum * 1.0 / self.cnt / self.scale if self.cnt > 0 else "N/A"
+
+    def max(self) -> Union[float, str]:
+        return self._max / self.scale if self._max else "N/A"
+
+    def dump_avg_max(self) -> list[Union[float, str]]:
+        return [self.avg(), self.max()]
+
+    def __enter__(self) -> None:
+        self.start_time = time.monotonic_ns()
+
+    def __exit__(
+        self,
+        exc_type: Optional[type[BaseException]],
+        exc_value: Optional[BaseException],
+        exc_traceback: Optional[TracebackType],
+    ) -> None:
+        self.collect(time.monotonic_ns() - self.start_time)

benchmarks/cutlass_benchmarks/w8a8_benchmarks.py

@@ -19,7 +19,7 @@ from vllm import _custom_ops as ops
 from vllm.model_executor.layers.quantization.utils.fp8_utils import (
     w8a8_block_fp8_matmul,
 )
-from vllm.utils import FlexibleArgumentParser
+from vllm.utils import FlexibleArgumentParser, cdiv
 
 DEFAULT_MODELS = list(WEIGHT_SHAPES.keys())
 DEFAULT_BATCH_SIZES = [1, 16, 32, 64, 128, 256, 512]
@@ -117,14 +117,9 @@ def bench_fp8(
     scale_a = torch.tensor(1.0, device="cuda", dtype=torch.float32)
     scale_b = torch.tensor(1.0, device="cuda", dtype=torch.float32)
 
-    def ceil_div(x: int, y: int) -> int:
-        return (x + y - 1) // y
-
-    block_scale_a = torch.rand(
-        (m, ceil_div(k, 128)), device="cuda", dtype=torch.float32
-    )
+    block_scale_a = torch.rand((m, cdiv(k, 128)), device="cuda", dtype=torch.float32)
     block_scale_b = torch.rand(
-        ceil_div(k, 128), ceil_div(n, 128), device="cuda", dtype=torch.float32
+        cdiv(k, 128), cdiv(n, 128), device="cuda", dtype=torch.float32
     )
     block_scale_a_M_major = block_scale_a.t().contiguous().t()
     block_scale_b_K_major = block_scale_b.t().contiguous().t()

benchmarks/disagg_benchmarks/disagg_overhead_benchmark.sh

@@ -3,7 +3,7 @@
 # benchmark the overhead of disaggregated prefill.
 # methodology:
 # - send all request to prefill vLLM instance. It will buffer KV cache.
-# - then send all request to decode instance. 
+# - then send all request to decode instance.
 # - The TTFT of decode instance is the overhead.
 
 set -ex
@@ -12,6 +12,8 @@ kill_gpu_processes() {
   # kill all processes on GPU.
   pgrep pt_main_thread | xargs -r kill -9
   pgrep python3 | xargs -r kill -9
+  # vLLM now names the process with VLLM prefix after https://github.com/vllm-project/vllm/pull/21445
+  pgrep VLLM | xargs -r kill -9
   sleep 10
 
   # remove vllm config file
@@ -61,7 +63,7 @@ benchmark() {
     --gpu-memory-utilization 0.6 \
     --kv-transfer-config \
     '{"kv_connector":"PyNcclConnector","kv_role":"kv_producer","kv_rank":0,"kv_parallel_size":2,"kv_buffer_size":5e9}' &
-    
+
 
   CUDA_VISIBLE_DEVICES=1 python3 \
     -m vllm.entrypoints.openai.api_server \
@@ -76,38 +78,38 @@ benchmark() {
   wait_for_server 8200
 
   # let the prefill instance finish prefill
-  python3 ../benchmark_serving.py \
-          --backend vllm \
-          --model $model \
-          --dataset-name $dataset_name \
-          --dataset-path $dataset_path \
-          --sonnet-input-len $input_len \
-          --sonnet-output-len "$output_len" \
-          --sonnet-prefix-len $prefix_len \
-          --num-prompts $num_prompts \
-          --port 8100 \
-          --save-result \
-          --result-dir $results_folder \
-          --result-filename disagg_prefill_tp1.json \
-          --request-rate "inf"
+  vllm bench serve \
+    --backend vllm \
+    --model $model \
+    --dataset-name $dataset_name \
+    --dataset-path $dataset_path \
+    --sonnet-input-len $input_len \
+    --sonnet-output-len "$output_len" \
+    --sonnet-prefix-len $prefix_len \
+    --num-prompts $num_prompts \
+    --port 8100 \
+    --save-result \
+    --result-dir $results_folder \
+    --result-filename disagg_prefill_tp1.json \
+    --request-rate "inf"
 
 
   # send the request to decode.
   # The TTFT of this command will be the overhead of disagg prefill impl.
-  python3 ../benchmark_serving.py \
-          --backend vllm \
-          --model $model \
-          --dataset-name $dataset_name \
-          --dataset-path $dataset_path \
-          --sonnet-input-len $input_len \
-          --sonnet-output-len "$output_len" \
-          --sonnet-prefix-len $prefix_len \
-          --num-prompts $num_prompts \
-          --port 8200 \
-          --save-result \
-          --result-dir $results_folder \
-          --result-filename disagg_prefill_tp1_overhead.json \
-          --request-rate "$qps"
+  vllm bench serve \
+    --backend vllm \
+    --model $model \
+    --dataset-name $dataset_name \
+    --dataset-path $dataset_path \
+    --sonnet-input-len $input_len \
+    --sonnet-output-len "$output_len" \
+    --sonnet-prefix-len $prefix_len \
+    --num-prompts $num_prompts \
+    --port 8200 \
+    --save-result \
+    --result-dir $results_folder \
+    --result-filename disagg_prefill_tp1_overhead.json \
+    --request-rate "$qps"
   kill_gpu_processes
 
 }

benchmarks/disagg_benchmarks/disagg_performance_benchmark.sh

@@ -18,6 +18,8 @@ kill_gpu_processes() {
   # kill all processes on GPU.
   pgrep pt_main_thread | xargs -r kill -9
   pgrep python3 | xargs -r kill -9
+  # vLLM now names the process with VLLM prefix after https://github.com/vllm-project/vllm/pull/21445
+  pgrep VLLM | xargs -r kill -9
   for port in 8000 8100 8200; do lsof -t -i:$port | xargs -r kill -9; done
   sleep 1
 }
@@ -58,7 +60,7 @@ launch_chunked_prefill() {
 
 
 launch_disagg_prefill() {
-  model="meta-llama/Meta-Llama-3.1-8B-Instruct" 
+  model="meta-llama/Meta-Llama-3.1-8B-Instruct"
   # disagg prefill
   CUDA_VISIBLE_DEVICES=0 python3 \
     -m vllm.entrypoints.openai.api_server \
@@ -97,20 +99,20 @@ benchmark() {
   output_len=$2
   tag=$3
 
-  python3 ../benchmark_serving.py \
-          --backend vllm \
-          --model $model \
-          --dataset-name $dataset_name \
-          --dataset-path $dataset_path \
-          --sonnet-input-len $input_len \
-          --sonnet-output-len "$output_len" \
-          --sonnet-prefix-len $prefix_len \
-          --num-prompts $num_prompts \
-          --port 8000 \
-          --save-result \
-          --result-dir $results_folder \
-          --result-filename "$tag"-qps-"$qps".json \
-          --request-rate "$qps"
+  vllm bench serve \
+    --backend vllm \
+    --model $model \
+    --dataset-name $dataset_name \
+    --dataset-path $dataset_path \
+    --sonnet-input-len $input_len \
+    --sonnet-output-len "$output_len" \
+    --sonnet-prefix-len $prefix_len \
+    --num-prompts $num_prompts \
+    --port 8000 \
+    --save-result \
+    --result-dir $results_folder \
+    --result-filename "$tag"-qps-"$qps".json \
+    --request-rate "$qps"
 
   sleep 2
 }

benchmarks/disagg_benchmarks/disagg_prefill_proxy_server.py

@@ -1,63 +1,199 @@
 # SPDX-License-Identifier: Apache-2.0
 # SPDX-FileCopyrightText: Copyright contributors to the vLLM project
 
+import argparse
+import asyncio
+import logging
 import os
 
 import aiohttp
-from quart import Quart, make_response, request
+from quart import Quart, Response, make_response, request
+from rate_limiter import RateLimiter
+from request_queue import RequestQueue
 
-AIOHTTP_TIMEOUT = aiohttp.ClientTimeout(total=6 * 60 * 60)
-
-app = Quart(__name__)
+# Configure logging
+logging.basicConfig(level=logging.INFO)
+logger = logging.getLogger(__name__)
 
 
-async def forward_request(url, data):
-    async with aiohttp.ClientSession(timeout=AIOHTTP_TIMEOUT) as session:
+def parse_args():
+    """parse command line arguments"""
+    parser = argparse.ArgumentParser(description="vLLM P/D disaggregation proxy server")
+
+    # Add args
+    parser.add_argument(
+        "--timeout",
+        type=float,
+        default=300,
+        help="Timeout for backend service requests in seconds (default: 300)",
+    )
+    parser.add_argument(
+        "--max-concurrent",
+        type=int,
+        default=100,
+        help="Maximum concurrent requests to backend services (default: 100)",
+    )
+    parser.add_argument(
+        "--queue-size",
+        type=int,
+        default=500,
+        help="Maximum number of requests in the queue (default: 500)",
+    )
+    parser.add_argument(
+        "--rate-limit",
+        type=int,
+        default=40,
+        help="Maximum requests per second (default: 40)",
+    )
+    parser.add_argument(
+        "--port",
+        type=int,
+        default=8000,
+        help="Port to run the server on (default: 8000)",
+    )
+    parser.add_argument(
+        "--prefill-url",
+        type=str,
+        default="http://localhost:8100/v1/completions",
+        help="Prefill service endpoint URL",
+    )
+    parser.add_argument(
+        "--decode-url",
+        type=str,
+        default="http://localhost:8200/v1/completions",
+        help="Decode service endpoint URL",
+    )
+
+    return parser.parse_args()
+
+
+def main():
+    """parse command line arguments"""
+    args = parse_args()
+
+    # Initialize configuration using command line parameters
+    AIOHTTP_TIMEOUT = aiohttp.ClientTimeout(total=args.timeout)
+    MAX_CONCURRENT_REQUESTS = args.max_concurrent
+    REQUEST_QUEUE_SIZE = args.queue_size
+    RATE_LIMIT = args.rate_limit
+    PREFILL_SERVICE_URL = args.prefill_url
+    DECODE_SERVICE_URL = args.decode_url
+    PORT = args.port
+
+    app = Quart(__name__)
+
+    # Initialize the rate limiter and request queue
+    rate_limiter = RateLimiter(RATE_LIMIT)
+    request_queue = RequestQueue(MAX_CONCURRENT_REQUESTS, REQUEST_QUEUE_SIZE)
+
+    # Attach the configuration object to the application instance
+    app.config.update(
+        {
+            "AIOHTTP_TIMEOUT": AIOHTTP_TIMEOUT,
+            "rate_limiter": rate_limiter,
+            "request_queue": request_queue,
+            "PREFILL_SERVICE_URL": PREFILL_SERVICE_URL,
+            "DECODE_SERVICE_URL": DECODE_SERVICE_URL,
+        }
+    )
+
+    # Start queue processing on app startup
+    @app.before_serving
+    async def startup():
+        """Start request processing task when app starts serving"""
+        asyncio.create_task(request_queue.process())
+
+    async def forward_request(url, data):
+        """Forward request to backend service with rate limiting and error handling"""
         headers = {"Authorization": f"Bearer {os.environ.get('OPENAI_API_KEY')}"}
-        async with session.post(url=url, json=data, headers=headers) as response:
-            if response.status == 200:
-                # if response.headers.get('Transfer-Encoding') == 'chunked':
-                if True:
-                    async for chunk_bytes in response.content.iter_chunked(1024):
-                        yield chunk_bytes
-                else:
-                    content = await response.read()
-                    yield content
 
-
-@app.route("/v1/completions", methods=["POST"])
-async def handle_request():
-    try:
-        original_request_data = await request.get_json()
-
-        prefill_request = original_request_data.copy()
-        # change max_tokens = 1 to let it only do prefill
-        prefill_request["max_tokens"] = 1
-
-        # finish prefill
-        async for _ in forward_request(
-            "http://localhost:8100/v1/completions", prefill_request
+        # Use rate limiter as context manager
+        async with (
+            rate_limiter,
+            aiohttp.ClientSession(timeout=AIOHTTP_TIMEOUT) as session,
         ):
-            continue
+            try:
+                async with session.post(
+                    url=url, json=data, headers=headers
+                ) as response:
+                    if response.status == 200:
+                        # Stream response chunks
+                        async for chunk_bytes in response.content.iter_chunked(1024):
+                            yield chunk_bytes
+                    else:
+                        # Handle backend service errors
+                        error_text = await response.text()
+                        logger.error(
+                            "Backend service error: %s - %s",
+                            response.status,
+                            error_text,
+                        )
+                        yield b'{"error": "Backend service error"}'
+            except aiohttp.ClientError as e:
+                # Handle connection errors
+                logger.error("Connection error to %s: %s", url, str(e))
+                yield b'{"error": "Service unavailable"}'
+            except asyncio.TimeoutError:
+                # Handle timeout errors
+                logger.error("Timeout connecting to %s", url)
+                yield b'{"error": "Service timeout"}'
 
-        # return decode
-        generator = forward_request(
-            "http://localhost:8200/v1/completions", original_request_data
-        )
-        response = await make_response(generator)
-        response.timeout = None
+    async def process_request():
+        """Process a single request through prefill and decode stages"""
+        try:
+            original_request_data = await request.get_json()
 
-        return response
+            # Create prefill request (max_tokens=1)
+            prefill_request = original_request_data.copy()
+            prefill_request["max_tokens"] = 1
 
-    except Exception as e:
-        import sys
-        import traceback
+            # Execute prefill stage
+            async for _ in forward_request(PREFILL_SERVICE_URL, prefill_request):
+                continue
 
-        exc_info = sys.exc_info()
-        print("Error occurred in disagg prefill proxy server")
-        print(e)
-        print("".join(traceback.format_exception(*exc_info)))
+            # Execute decode stage and stream response
+            generator = forward_request(DECODE_SERVICE_URL, original_request_data)
+            response = await make_response(generator)
+            response.timeout = None  # Disable timeout for streaming response
+            return response
+
+        except Exception:
+            logger.exception("Error processing request")
+            return Response(
+                response=b'{"error": "Internal server error"}',
+                status=500,
+                content_type="application/json",
+            )
+
+    @app.route("/v1/completions", methods=["POST"])
+    async def handle_request():
+        """Handle incoming API requests with concurrency and rate limiting"""
+        # Create task for request processing
+        task = asyncio.create_task(process_request())
+
+        # Enqueue request or reject if queue is full
+        if not await request_queue.enqueue(task):
+            return Response(
+                response=b'{"error": "Server busy, try again later"}',
+                status=503,
+                content_type="application/json",
+            )
+
+        try:
+            # Return the response from the processing task
+            return await task
+        except asyncio.CancelledError:
+            # Handle task cancellation (timeout or queue full)
+            logger.warning("Request cancelled due to timeout or queue full")
+            return Response(
+                response=b'{"error": "Request cancelled"}',
+                status=503,
+                content_type="application/json",
+            )
+
+    # Start the Quart server with host can be set to 0.0.0.0
+    app.run(port=PORT)
 
 
 if __name__ == "__main__":
-    app.run(port=8000)
+    main()

benchmarks/disagg_benchmarks/rate_limiter.py

@@ -0,0 +1,45 @@
+# SPDX-License-Identifier: Apache-2.0
+# SPDX-FileCopyrightText: Copyright contributors to the vLLM project
+
+import asyncio
+import time
+
+
+class RateLimiter:
+    """Token bucket rate limiter implementation"""
+
+    def __init__(self, rate_limit):
+        self.rate_limit = rate_limit  # Requests per second
+        self.num_available_tokens = rate_limit  # Available tokens
+        self.last_refill = time.monotonic()  # Last token refill time
+        self.lock = asyncio.Lock()  # Synchronization lock
+
+    async def acquire(self):
+        """Acquire a token from the rate limiter"""
+        while True:
+            async with self.lock:
+                current_time = time.monotonic()
+                elapsed = current_time - self.last_refill
+
+                # Refill num_available_tokens if more than 1 second has passed
+                if elapsed > 1.0:
+                    self.num_available_tokens = self.rate_limit
+                    self.last_refill = current_time
+
+                # Check if num_available_tokens are available
+                if self.num_available_tokens > 0:
+                    self.num_available_tokens -= 1
+                    return True
+
+                # Calculate wait time if no num_available_tokens available
+                wait_time = 1.0 - elapsed
+            await asyncio.sleep(wait_time)
+
+    async def __aenter__(self):
+        """Enter async context manager - acquire token"""
+        await self.acquire()
+        return self
+
+    async def __aexit__(self, exc_type, exc_value, traceback):
+        """Exit async context manager - no cleanup needed"""
+        pass

benchmarks/disagg_benchmarks/request_queue.py

@@ -0,0 +1,39 @@
+# SPDX-License-Identifier: Apache-2.0
+# SPDX-FileCopyrightText: Copyright contributors to the vLLM project
+
+import asyncio
+from collections import deque
+
+
+class RequestQueue:
+    """Request queue manager with concurrency control"""
+
+    def __init__(self, max_concurrent, max_queue_size):
+        # Maximum concurrent requests
+        self.max_concurrent = max_concurrent
+        self.max_queue_size = max_queue_size  # Maximum queue size
+        # Concurrency control
+        self.semaphore = asyncio.Semaphore(max_concurrent)
+        self.queue = deque()  # Request queue
+        self.queue_size = 0  # Current queue size
+        self.lock = asyncio.Lock()  # Sync queue Lock
+
+    async def enqueue(self, task):
+        """Add a request task to the queue"""
+        async with self.lock:
+            if self.queue_size >= self.max_queue_size:
+                return False
+
+            self.queue.append(task)
+            self.queue_size += 1
+            return True
+
+    async def process(self):
+        """Process queued requests using semaphore for concurrency control"""
+        while True:
+            if self.queue:
+                async with self.semaphore, self.lock:
+                    task = self.queue.popleft()
+                    self.queue_size -= 1
+                    await task
+            await asyncio.sleep(0.01)  # Yield control to event loop

benchmarks/kernels/bench_fp8_gemm.py

@@ -11,6 +11,80 @@ from vllm._custom_ops import cutlass_scaled_mm as vllm_scaled_mm
 from vllm._custom_ops import scaled_fp8_quant as vllm_scaled_fp8_quant
 from vllm.triton_utils import triton
 
+PROVIDER_CFGS = {
+    "torch-bf16": dict(enabled=True),
+    "fp8-tensor-w-token-a": dict(
+        w="tensor", a="token", no_a_quant=False, enabled=False
+    ),
+    "fp8-tensor-w-tensor-a": dict(
+        w="tensor", a="tensor", no_a_quant=False, enabled=True
+    ),
+    "fp8-channel-w-token-a": dict(
+        w="channel", a="token", no_a_quant=False, enabled=True
+    ),
+    "fp8-channel-w-tensor-a": dict(
+        w="channel", a="tensor", no_a_quant=False, enabled=False
+    ),
+    "fp8-tensor-w-token-a-noquant": dict(
+        w="tensor", a="token", no_a_quant=True, enabled=False
+    ),
+    "fp8-tensor-w-tensor-a-noquant": dict(
+        w="tensor", a="tensor", no_a_quant=True, enabled=True
+    ),
+    "fp8-channel-w-token-a-noquant": dict(
+        w="channel", a="token", no_a_quant=True, enabled=True
+    ),
+    "fp8-channel-w-tensor-a-noquant": dict(
+        w="channel", a="tensor", no_a_quant=True, enabled=False
+    ),
+}
+
+_enabled = [k for k, v in PROVIDER_CFGS.items() if v["enabled"]]
+
+
+def _quant_weight_fp8(b: torch.Tensor, w_type: str, device: str):
+    if w_type == "tensor":
+        scale_b = torch.ones(1, device=device, dtype=torch.float32)
+        b_fp8, scale_b_fp8 = vllm_scaled_fp8_quant(b, scale_b)
+    else:
+        b_fp8, scale_b_fp8 = vllm_scaled_fp8_quant(b, use_per_token_if_dynamic=True)
+    return b_fp8.t(), scale_b_fp8
+
+
+def build_fp8_runner(cfg, a, b, dtype, device):
+    b_fp8, scale_b_fp8 = _quant_weight_fp8(b, cfg["w"], device)
+
+    scale_a_const = (
+        torch.ones(1, device=device, dtype=torch.float32)
+        if cfg["a"] == "tensor"
+        else None
+    )
+
+    if cfg["no_a_quant"]:
+        if cfg["a"] == "tensor":
+            a_fp8, scale_a_fp8 = vllm_scaled_fp8_quant(a, scale_a_const)
+        else:
+            a_fp8, scale_a_fp8 = vllm_scaled_fp8_quant(a, use_per_token_if_dynamic=True)
+
+        def run():
+            return vllm_scaled_mm(a_fp8, b_fp8, scale_a_fp8, scale_b_fp8, dtype)
+
+        return run
+
+    if cfg["a"] == "tensor":
+
+        def run():
+            a_fp8, scale_a_fp8 = vllm_scaled_fp8_quant(a, scale_a_const)
+            return vllm_scaled_mm(a_fp8, b_fp8, scale_a_fp8, scale_b_fp8, dtype)
+
+    else:
+
+        def run():
+            a_fp8, scale_a_fp8 = vllm_scaled_fp8_quant(a, use_per_token_if_dynamic=True)
+            return vllm_scaled_mm(a_fp8, b_fp8, scale_a_fp8, scale_b_fp8, dtype)
+
+    return run
+
 
 @triton.testing.perf_report(
     triton.testing.Benchmark(
@@ -18,28 +92,8 @@ from vllm.triton_utils import triton
         x_vals=[1, 16, 64, 128, 256, 512, 1024, 2048, 4096, 8192, 16384],
         x_log=False,
         line_arg="provider",
-        line_vals=[
-            "torch-bf16",
-            # "fp8-tensor-w-token-a",
-            "fp8-tensor-w-tensor-a",
-            "fp8-channel-w-token-a",
-            # "fp8-channel-w-tensor-a",
-            # "fp8-tensor-w-token-a-noquant",
-            "fp8-tensor-w-tensor-a-noquant",
-            "fp8-channel-w-token-a-noquant",
-            # "fp8-channel-w-tensor-a-noquant",
-        ],
-        line_names=[
-            "torch-bf16",
-            # "fp8-tensor-w-token-a",
-            "fp8-tensor-w-tensor-a",
-            "fp8-channel-w-token-a",
-            # "fp8-channel-w-tensor-a",
-            # "fp8-tensor-w-token-a-noquant",
-            "fp8-tensor-w-tensor-a-noquant",
-            "fp8-channel-w-token-a-noquant",
-            # "fp8-channel-w-tensor-a-noquant",
-        ],
+        line_vals=_enabled,
+        line_names=_enabled,
         ylabel="TFLOP/s (larger is better)",
         plot_name="BF16 vs FP8 GEMMs",
         args={},
@@ -50,144 +104,34 @@ def benchmark(batch_size, provider, N, K):
     device = "cuda"
     dtype = torch.bfloat16
 
-    # Create input tensors
     a = torch.randn((M, K), device=device, dtype=dtype)
     b = torch.randn((N, K), device=device, dtype=dtype)
 
     quantiles = [0.5, 0.2, 0.8]
 
-    if "torch-bf16" in provider:
+    if provider == "torch-bf16":
         ms, min_ms, max_ms = triton.testing.do_bench_cudagraph(
             lambda: torch.nn.functional.linear(a, b), quantiles=quantiles
         )
-
-    elif "fp8" in provider:
-        # Weights are always quantized ahead of time
-        if "noquant" in provider:
-            # For no quantization, we just measure the GEMM
-            if "tensor-w-token-a" in provider:
-                # Dynamic per-token quant for A, per-tensor quant for B
-                b_fp8, scale_b_fp8 = vllm_scaled_fp8_quant(b)
-                assert scale_b_fp8.numel() == 1
-                a_fp8, scale_a_fp8 = vllm_scaled_fp8_quant(
-                    a, use_per_token_if_dynamic=True
-                )
-
-                def run_quant():
-                    return vllm_scaled_mm(a_fp8, b_fp8, scale_a_fp8, scale_b_fp8, dtype)
-
-            elif "tensor-w-tensor-a" in provider:
-                # Static per-tensor quantization with fixed scales
-                # for both A and B
-                scale_a = torch.tensor([1.0], device=device, dtype=torch.float32)
-                scale_b = torch.tensor([1.0], device=device, dtype=torch.float32)
-                b_fp8, scale_b_fp8 = vllm_scaled_fp8_quant(b, scale_b)
-                assert scale_b_fp8.numel() == 1
-                a_fp8, scale_a_fp8 = vllm_scaled_fp8_quant(a, scale_a)
-
-                def run_quant():
-                    return vllm_scaled_mm(a_fp8, b_fp8, scale_a_fp8, scale_b_fp8, dtype)
-
-            elif "channel-w-token-a" in provider:
-                # Static per-channel quantization for weights, per-token
-                # quant for A
-                scale_b = torch.tensor((N,), device=device, dtype=torch.float32)
-                b_fp8, scale_b_fp8 = vllm_scaled_fp8_quant(b, scale_b)
-                scale_b_fp8 = scale_b_fp8.expand(N).contiguous()
-                assert scale_b_fp8.numel() == N
-                a_fp8, scale_a_fp8 = vllm_scaled_fp8_quant(
-                    a, use_per_token_if_dynamic=True
-                )
-
-                def run_quant():
-                    return vllm_scaled_mm(a_fp8, b_fp8, scale_a_fp8, scale_b_fp8, dtype)
-
-            elif "channel-w-tensor-a" in provider:
-                # Static per-channel quantization for weights, per-tensor
-                # quant for A
-                scale_a = torch.tensor([1.0], device=device, dtype=torch.float32)
-                scale_b = torch.tensor((N,), device=device, dtype=torch.float32)
-                b_fp8, scale_b_fp8 = vllm_scaled_fp8_quant(b, scale_b)
-                scale_b_fp8 = scale_b_fp8.expand(N).contiguous()
-                assert scale_b_fp8.numel() == N
-                a_fp8, scale_a_fp8 = vllm_scaled_fp8_quant(a, scale_a)
-
-                def run_quant():
-                    return vllm_scaled_mm(a_fp8, b_fp8, scale_a_fp8, scale_b_fp8, dtype)
-
-        else:
-            # In these cases, we quantize the activations during the GEMM call
-            if "tensor-w-token-a" in provider:
-                # Dynamic per-token quant for A, per-tensor quant for B
-                b_fp8, scale_b_fp8 = vllm_scaled_fp8_quant(b)
-                assert scale_b_fp8.numel() == 1
-
-                def run_quant():
-                    a_fp8, scale_a_fp8 = vllm_scaled_fp8_quant(
-                        a, use_per_token_if_dynamic=True
-                    )
-                    return vllm_scaled_mm(a_fp8, b_fp8, scale_a_fp8, scale_b_fp8, dtype)
-
-            elif "tensor-w-tensor-a" in provider:
-                # Static per-tensor quantization with fixed scales
-                # for both A and B
-                scale_a = torch.tensor([1.0], device=device, dtype=torch.float32)
-                scale_b = torch.tensor([1.0], device=device, dtype=torch.float32)
-                b_fp8, scale_b_fp8 = vllm_scaled_fp8_quant(b, scale_b)
-                assert scale_b_fp8.numel() == 1
-
-                def run_quant():
-                    a_fp8, scale_a_fp8 = vllm_scaled_fp8_quant(a, scale_a)
-                    return vllm_scaled_mm(a_fp8, b_fp8, scale_a_fp8, scale_b_fp8, dtype)
-
-            elif "channel-w-token-a" in provider:
-                # Static per-channel quantization for weights, per-token
-                # quant for A
-                scale_b = torch.tensor((N,), device=device, dtype=torch.float32)
-                b_fp8, scale_b_fp8 = vllm_scaled_fp8_quant(b, scale_b)
-                scale_b_fp8 = scale_b_fp8.expand(N).contiguous()
-                assert scale_b_fp8.numel() == N
-
-                def run_quant():
-                    a_fp8, scale_a_fp8 = vllm_scaled_fp8_quant(
-                        a, use_per_token_if_dynamic=True
-                    )
-                    return vllm_scaled_mm(a_fp8, b_fp8, scale_a_fp8, scale_b_fp8, dtype)
-
-            elif "channel-w-tensor-a" in provider:
-                # Static per-channel quantization for weights, per-tensor
-                # quant for A
-                scale_a = torch.tensor([1.0], device=device, dtype=torch.float32)
-                scale_b = torch.tensor((N,), device=device, dtype=torch.float32)
-                b_fp8, scale_b_fp8 = vllm_scaled_fp8_quant(b, scale_b)
-                scale_b_fp8 = scale_b_fp8.expand(N).contiguous()
-                assert scale_b_fp8.numel() == N
-
-                def run_quant():
-                    a_fp8, scale_a_fp8 = vllm_scaled_fp8_quant(a, scale_a)
-                    return vllm_scaled_mm(a_fp8, b_fp8, scale_a_fp8, scale_b_fp8, dtype)
-
-        b_fp8 = b_fp8.t()
-
+    else:
+        cfg = PROVIDER_CFGS[provider]
+        run_quant = build_fp8_runner(cfg, a, b, dtype, device)
         ms, min_ms, max_ms = triton.testing.do_bench_cudagraph(
             lambda: run_quant(), quantiles=quantiles
         )
 
-    # Calculate TFLOP/s, two flops per multiply-add
-    tflops = lambda ms: (2 * M * N * K) * 1e-12 / (ms * 1e-3)
-    return tflops(ms), tflops(max_ms), tflops(min_ms)
+    to_tflops = lambda t_ms: (2 * M * N * K) * 1e-12 / (t_ms * 1e-3)
+    return to_tflops(ms), to_tflops(max_ms), to_tflops(min_ms)
 
 
 def prepare_shapes(args):
-    KN_model_names = []
-    models_tps = list(itertools.product(args.models, args.tp_sizes))
-    for model, tp_size in models_tps:
-        assert model in WEIGHT_SHAPES
-        for KN, tp_split_dim in copy.deepcopy(WEIGHT_SHAPES[model]):
-            KN[tp_split_dim] = KN[tp_split_dim] // tp_size
+    out = []
+    for model, tp_size in itertools.product(args.models, args.tp_sizes):
+        for KN, tp_dim in copy.deepcopy(WEIGHT_SHAPES[model]):
+            KN[tp_dim] //= tp_size
             KN.append(model)
-            KN_model_names.append(KN)
-    return KN_model_names
+            out.append(KN)
+    return out
 
 
 if __name__ == "__main__":
@@ -197,21 +141,13 @@ if __name__ == "__main__":
         nargs="+",
         type=str,
         default=["meta-llama/Llama-3.1-8B-Instruct"],
-        choices=[*WEIGHT_SHAPES.keys()],
-        help="List of models to benchmark",
-    )
-    parser.add_argument(
-        "--tp-sizes",
-        nargs="+",
-        type=int,
-        default=[1],
-        help="List of tensor parallel sizes",
+        choices=list(WEIGHT_SHAPES.keys()),
     )
+    parser.add_argument("--tp-sizes", nargs="+", type=int, default=[1])
     args = parser.parse_args()
 
-    KN_model_names = prepare_shapes(args)
-    for K, N, model_name in KN_model_names:
-        print(f"{model_name}, N={N} K={K}, BF16 vs FP8 GEMMs TFLOP/s:")
+    for K, N, model in prepare_shapes(args):
+        print(f"{model}, N={N} K={K}, BF16 vs FP8 GEMMs TFLOP/s:")
         benchmark.run(
             print_data=True,
             show_plots=True,

benchmarks/kernels/bench_int8_gemm.py

@@ -0,0 +1,169 @@
+# SPDX-License-Identifier: Apache-2.0
+# SPDX-FileCopyrightText: Copyright contributors to the vLLM project
+import argparse
+import copy
+import itertools
+
+import torch
+from weight_shapes import WEIGHT_SHAPES
+
+from vllm._custom_ops import cutlass_scaled_mm as vllm_scaled_mm
+from vllm._custom_ops import scaled_int8_quant as vllm_scaled_int8_quant
+from vllm.triton_utils import triton
+
+PROVIDER_CFGS = {
+    "torch-bf16": dict(enabled=True),
+    "int8-tensor-w-token-a": dict(
+        w="tensor", a="token", no_a_quant=False, enabled=False
+    ),
+    "int8-tensor-w-tensor-a": dict(
+        w="tensor", a="tensor", no_a_quant=False, enabled=True
+    ),
+    "int8-channel-w-token-a": dict(
+        w="channel", a="token", no_a_quant=False, enabled=True
+    ),
+    "int8-channel-w-tensor-a": dict(
+        w="channel", a="tensor", no_a_quant=False, enabled=False
+    ),
+    "int8-tensor-w-token-a-noquant": dict(
+        w="tensor", a="token", no_a_quant=True, enabled=False
+    ),
+    "int8-tensor-w-tensor-a-noquant": dict(
+        w="tensor", a="tensor", no_a_quant=True, enabled=True
+    ),
+    "int8-channel-w-token-a-noquant": dict(
+        w="channel", a="token", no_a_quant=True, enabled=True
+    ),
+    "int8-channel-w-tensor-a-noquant": dict(
+        w="channel", a="tensor", no_a_quant=True, enabled=False
+    ),
+}
+
+
+def _quant_weight(b, w_type, device):
+    if w_type == "tensor":
+        scale_b = torch.ones(1, device=device, dtype=torch.float32)
+        b_int8, scale_b_int8, _ = vllm_scaled_int8_quant(b, scale_b)
+        assert scale_b_int8.numel() == 1
+    else:  # channel
+        b_int8, scale_b_int8, _ = vllm_scaled_int8_quant(b)
+        assert scale_b_int8.numel() == b.shape[0]
+    return b_int8.t(), scale_b_int8
+
+
+def build_int8_runner(cfg, a, b, dtype, device):
+    # quant before running the kernel
+    b_int8, scale_b_int8 = _quant_weight(b, cfg["w"], device)
+
+    scale_a_const = None
+    if cfg["a"] == "tensor":
+        scale_a_const = torch.ones(1, device=device, dtype=torch.float32)
+
+    # no quant, create activation ahead
+    if cfg["no_a_quant"]:
+        if cfg["a"] == "tensor":
+            a_int8, scale_a_int8, _ = vllm_scaled_int8_quant(a, scale_a_const)
+        else:  # token
+            a_int8, scale_a_int8, _ = vllm_scaled_int8_quant(a)
+
+        def run_quant():
+            return vllm_scaled_mm(a_int8, b_int8, scale_a_int8, scale_b_int8, dtype)
+
+        return run_quant
+
+    # dynamic quant, create activation inside
+    if cfg["a"] == "tensor":
+
+        def run_quant():
+            a_int8, scale_a_int8, _ = vllm_scaled_int8_quant(a, scale_a_const)
+            return vllm_scaled_mm(a_int8, b_int8, scale_a_int8, scale_b_int8, dtype)
+
+    else:  # token
+
+        def run_quant():
+            a_int8, scale_a_int8, _ = vllm_scaled_int8_quant(a)
+            return vllm_scaled_mm(a_int8, b_int8, scale_a_int8, scale_b_int8, dtype)
+
+    return run_quant
+
+
+_enabled = [k for k, v in PROVIDER_CFGS.items() if v.get("enabled")]
+
+
+@triton.testing.perf_report(
+    triton.testing.Benchmark(
+        x_names=["batch_size"],
+        x_vals=[1, 16, 64, 128, 256, 512, 1024, 2048, 4096, 8192, 16384],
+        x_log=False,
+        line_arg="provider",
+        line_vals=_enabled,
+        line_names=[k for k in _enabled],
+        ylabel="TFLOP/s (larger is better)",
+        plot_name="BF16 vs INT8 GEMMs",
+        args={},
+    )
+)
+def benchmark(batch_size, provider, N, K):
+    M = batch_size
+    device = "cuda"
+    dtype = torch.bfloat16
+    a = torch.randn((M, K), device=device, dtype=dtype)
+    b = torch.randn((N, K), device=device, dtype=dtype)
+
+    quantiles = [0.5, 0.2, 0.8]
+
+    if provider == "torch-bf16":
+        ms, min_ms, max_ms = triton.testing.do_bench_cudagraph(
+            lambda: torch.nn.functional.linear(a, b), quantiles=quantiles
+        )
+    else:
+        cfg = PROVIDER_CFGS[provider]
+        run_quant = build_int8_runner(cfg, a, b, dtype, device)
+        ms, min_ms, max_ms = triton.testing.do_bench_cudagraph(
+            lambda: run_quant(), quantiles=quantiles
+        )
+
+    to_tflops = lambda t_ms: (2 * M * N * K) * 1e-12 / (t_ms * 1e-3)
+    return to_tflops(ms), to_tflops(max_ms), to_tflops(min_ms)
+
+
+def prepare_shapes(args):
+    KN_model_names = []
+    for model, tp_size in itertools.product(args.models, args.tp_sizes):
+        for KN, tp_dim in copy.deepcopy(WEIGHT_SHAPES[model]):
+            KN[tp_dim] //= tp_size
+            KN.append(model)
+            KN_model_names.append(KN)
+    return KN_model_names
+
+
+if __name__ == "__main__":
+    parser = argparse.ArgumentParser()
+    parser.add_argument(
+        "--models",
+        nargs="+",
+        type=str,
+        default=["meta-llama/Llama-3.1-8B-Instruct"],
+        choices=list(WEIGHT_SHAPES.keys()),
+        help="List of models to benchmark",
+    )
+    parser.add_argument(
+        "--tp-sizes",
+        nargs="+",
+        type=int,
+        default=[1],
+        help="List of tensor parallel sizes",
+    )
+    args = parser.parse_args()
+
+    for K, N, model in prepare_shapes(args):
+        print(f"{model}, N={N} K={K}, BF16 vs INT8 GEMMs TFLOP/s:")
+        benchmark.run(
+            print_data=True,
+            show_plots=True,
+            save_path=f"bench_int8_res_n{N}_k{K}",
+            N=N,
+            K=K,
+        )
+
+    print("Benchmark finished!")

benchmarks/kernels/bench_nvfp4_gemm.py

@@ -0,0 +1,141 @@
+# SPDX-License-Identifier: Apache-2.0
+# SPDX-FileCopyrightText: Copyright contributors to the vLLM project
+import argparse
+import copy
+import itertools
+
+import torch
+from weight_shapes import WEIGHT_SHAPES
+
+from vllm import _custom_ops as ops
+from vllm.platforms import current_platform
+from vllm.scalar_type import scalar_types
+from vllm.triton_utils import triton
+
+if not current_platform.has_device_capability(100):
+    raise RuntimeError("NVFP4 requires compute capability of 10.0 (Blackwell)")
+
+
+FLOAT4_E2M1_MAX = scalar_types.float4_e2m1f.max()
+FLOAT8_E4M3_MAX = torch.finfo(torch.float8_e4m3fn).max
+
+PROVIDER_CFGS = {
+    "torch-bf16": dict(enabled=True),
+    "nvfp4": dict(no_a_quant=False, enabled=True),
+    "nvfp4-noquant": dict(no_a_quant=True, enabled=True),
+}
+
+_enabled = [k for k, v in PROVIDER_CFGS.items() if v["enabled"]]
+
+
+def _quant_weight_nvfp4(b: torch.Tensor, device: str):
+    # Compute global scale for weight
+    b_amax = torch.abs(b).max().to(torch.float32)
+    b_global_scale = FLOAT8_E4M3_MAX * FLOAT4_E2M1_MAX / b_amax
+    b_fp4, scale_b_fp4 = ops.scaled_fp4_quant(b, b_global_scale)
+    return b_fp4, scale_b_fp4, b_global_scale
+
+
+def build_nvfp4_runner(cfg, a, b, dtype, device):
+    b_fp4, scale_b_fp4, b_global_scale = _quant_weight_nvfp4(b, device)
+
+    # Compute global scale for activation
+    # NOTE: This is generally provided ahead-of-time by the model checkpoint.
+    a_amax = torch.abs(a).max().to(torch.float32)
+    a_global_scale = FLOAT8_E4M3_MAX * FLOAT4_E2M1_MAX / a_amax
+
+    # Alpha for the GEMM operation
+    alpha = 1.0 / (a_global_scale * b_global_scale)
+
+    if cfg["no_a_quant"]:
+        # Pre-quantize activation
+        a_fp4, scale_a_fp4 = ops.scaled_fp4_quant(a, a_global_scale)
+
+        def run():
+            return ops.cutlass_scaled_fp4_mm(
+                a_fp4, b_fp4, scale_a_fp4, scale_b_fp4, alpha, dtype
+            )
+
+        return run
+
+    # Quantize activation on-the-fly
+    def run():
+        a_fp4, scale_a_fp4 = ops.scaled_fp4_quant(a, a_global_scale)
+        return ops.cutlass_scaled_fp4_mm(
+            a_fp4, b_fp4, scale_a_fp4, scale_b_fp4, alpha, dtype
+        )
+
+    return run
+
+
+@triton.testing.perf_report(
+    triton.testing.Benchmark(
+        x_names=["batch_size"],
+        x_vals=[1, 16, 64, 128, 256, 512, 1024, 2048, 4096, 8192, 16384],
+        x_log=False,
+        line_arg="provider",
+        line_vals=_enabled,
+        line_names=_enabled,
+        ylabel="TFLOP/s (larger is better)",
+        plot_name="BF16 vs NVFP4 GEMMs",
+        args={},
+    )
+)
+def benchmark(batch_size, provider, N, K):
+    M = batch_size
+    device = "cuda"
+    dtype = torch.bfloat16
+
+    a = torch.randn((M, K), device=device, dtype=dtype)
+    b = torch.randn((N, K), device=device, dtype=dtype)
+
+    quantiles = [0.5, 0.2, 0.8]
+
+    if provider == "torch-bf16":
+        ms, min_ms, max_ms = triton.testing.do_bench_cudagraph(
+            lambda: torch.nn.functional.linear(a, b), quantiles=quantiles
+        )
+    else:
+        cfg = PROVIDER_CFGS[provider]
+        run_quant = build_nvfp4_runner(cfg, a, b, dtype, device)
+        ms, min_ms, max_ms = triton.testing.do_bench_cudagraph(
+            lambda: run_quant(), quantiles=quantiles
+        )
+
+    to_tflops = lambda t_ms: (2 * M * N * K) * 1e-12 / (t_ms * 1e-3)
+    return to_tflops(ms), to_tflops(max_ms), to_tflops(min_ms)
+
+
+def prepare_shapes(args):
+    out = []
+    for model, tp_size in itertools.product(args.models, args.tp_sizes):
+        for KN, tp_dim in copy.deepcopy(WEIGHT_SHAPES[model]):
+            KN[tp_dim] //= tp_size
+            KN.append(model)
+            out.append(KN)
+    return out
+
+
+if __name__ == "__main__":
+    parser = argparse.ArgumentParser()
+    parser.add_argument(
+        "--models",
+        nargs="+",
+        type=str,
+        default=["meta-llama/Llama-3.1-8B-Instruct"],
+        choices=list(WEIGHT_SHAPES.keys()),
+    )
+    parser.add_argument("--tp-sizes", nargs="+", type=int, default=[1])
+    args = parser.parse_args()
+
+    for K, N, model in prepare_shapes(args):
+        print(f"{model}, N={N} K={K}, BF16 vs NVFP4 GEMMs TFLOP/s:")
+        benchmark.run(
+            print_data=True,
+            show_plots=True,
+            save_path=f"bench_nvfp4_res_n{N}_k{K}",
+            N=N,
+            K=K,
+        )
+
+    print("Benchmark finished!")

benchmarks/kernels/bench_per_token_quant_fp8.py

@@ -0,0 +1,98 @@
+# SPDX-License-Identifier: Apache-2.0
+# SPDX-FileCopyrightText: Copyright contributors to the vLLM project
+import itertools
+from typing import Callable
+
+import torch
+
+from vllm import _custom_ops as ops
+from vllm.config import CompilationConfig, VllmConfig, set_current_vllm_config
+from vllm.model_executor.layers.quantization.input_quant_fp8 import QuantFP8
+from vllm.model_executor.layers.quantization.utils.quant_utils import GroupShape
+from vllm.triton_utils import triton
+
+
+# TODO(luka): use standalone_compile utility
+def with_dyn_arg(fn: Callable, arg_index: int, dim_index: int):
+    def inner(*args):
+        torch._dynamo.mark_dynamic(args[arg_index], dim_index)
+        return fn(*args)
+
+    return inner
+
+
+torch._dynamo.config.recompile_limit = 8888
+compilation_config = CompilationConfig(custom_ops=["none"])
+with set_current_vllm_config(VllmConfig(compilation_config=compilation_config)):
+    torch_per_token_quant_fp8 = torch.compile(
+        QuantFP8(False, GroupShape.PER_TOKEN),
+        fullgraph=True,
+        dynamic=False,  # recompile for different shapes
+    )
+
+    # First dim is explicitly dynamic to simulate vLLM usage
+    torch_per_token_quant_fp8 = with_dyn_arg(torch_per_token_quant_fp8, 0, 0)
+
+
+def cuda_per_token_quant_fp8(
+    input: torch.Tensor,
+) -> tuple[torch.Tensor, torch.Tensor]:
+    return ops.scaled_fp8_quant(input)
+
+
+def calculate_diff(batch_size: int, seq_len: int):
+    """Calculate difference between Triton and CUDA implementations."""
+    device = torch.device("cuda")
+    x = torch.rand((batch_size * seq_len, 4096), dtype=torch.float16, device=device)
+
+    torch_out, torch_scale = torch_per_token_quant_fp8(x)
+    cuda_out, cuda_scale = cuda_per_token_quant_fp8(x)
+
+    if torch.allclose(
+        cuda_out.to(torch.float32), torch_out.to(torch.float32), rtol=1e-3, atol=1e-5
+    ) and torch.allclose(cuda_scale, torch_scale, rtol=1e-3, atol=1e-5):
+        print("✅ All implementations match")
+    else:
+        print("❌ Implementations differ")
+
+
+batch_size_range = [1, 16, 32, 64, 128]
+seq_len_range = [1, 16, 64, 128, 256, 512, 1024, 2048, 4096]
+
+configs = list(itertools.product(batch_size_range, seq_len_range))
+
+
+@triton.testing.perf_report(
+    triton.testing.Benchmark(
+        x_names=["batch_size", "seq_len"],
+        x_vals=configs,
+        line_arg="provider",
+        line_vals=["torch", "cuda"],
+        line_names=["Torch", "CUDA"],
+        styles=[("blue", "-"), ("green", "-")],
+        ylabel="us",
+        plot_name="per-token-dynamic-quant-fp8-performance",
+        args={},
+    )
+)
+def benchmark_quantization(batch_size, seq_len, provider):
+    dtype = torch.float16
+    device = torch.device("cuda")
+
+    x = torch.randn(batch_size * seq_len, 4096, device=device, dtype=dtype)
+
+    quantiles = [0.5, 0.2, 0.8]
+
+    if provider == "torch":
+        fn = lambda: torch_per_token_quant_fp8(x.clone())
+    elif provider == "cuda":
+        fn = lambda: cuda_per_token_quant_fp8(x.clone())
+
+    ms, min_ms, max_ms = triton.testing.do_bench_cudagraph(fn, quantiles=quantiles)
+
+    return 1000 * ms, 1000 * max_ms, 1000 * min_ms
+
+
+if __name__ == "__main__":
+    calculate_diff(batch_size=4, seq_len=4096)
+    benchmark_quantization.run(print_data=True)

benchmarks/kernels/benchmark_aqlm.py

@@ -1,345 +0,0 @@
-# SPDX-License-Identifier: Apache-2.0
-# SPDX-FileCopyrightText: Copyright contributors to the vLLM project
-
-import os
-import sys
-from typing import Optional
-
-import torch
-import torch.nn.functional as F
-
-from vllm import _custom_ops as ops
-from vllm.model_executor.layers.quantization.aqlm import (
-    dequantize_weight,
-    generic_dequantize_gemm,
-    get_int_dtype,
-    optimized_dequantize_gemm,
-)
-from vllm.utils import FlexibleArgumentParser
-
-os.environ["CUDA_VISIBLE_DEVICES"] = "0"
-
-
-def torch_mult(
-    # [..., in_features]
-    input: torch.Tensor,
-    weights: torch.Tensor,
-    # [num_out_groups, 1, 1, 1]
-    scales: torch.Tensor,
-) -> torch.Tensor:
-    output = F.linear(input, weights)
-    return output
-
-
-def dequant_out_scale(
-    # [..., in_features]
-    input: torch.Tensor,
-    # [num_out_groups, num_in_groups, num_codebooks]
-    codes: torch.IntTensor,
-    # [num_codebooks, codebook_size, out_group_size, in_group_size]
-    codebooks: torch.Tensor,
-    # [num_out_groups, 1, 1, 1]
-    scales: torch.Tensor,
-    output_partition_sizes: torch.IntTensor,
-    bias: Optional[torch.Tensor],
-) -> torch.Tensor:
-    weights = ops.aqlm_dequant(codes, codebooks, output_partition_sizes)
-
-    if bias is None:
-        output = F.linear(input, weights, bias)
-        orig_shape = output.shape
-        flattened_output = output.view(-1, output.size(-1))
-        f_scales = scales.view(-1, scales.shape[0])
-        b_scales = f_scales.expand(flattened_output.shape[0], -1)
-        flattened_output *= b_scales
-        return flattened_output.view(orig_shape)
-    else:
-        b_scales = scales.view(scales.shape[:-3] + (-1,)).expand(-1, weights.shape[1])
-        weights *= b_scales
-        return F.linear(input, weights, bias)
-
-
-def dequant_weight_scale(
-    # [..., in_features]
-    input: torch.Tensor,
-    # [num_out_groups, num_in_groups, num_codebooks]
-    codes: torch.IntTensor,
-    # [num_codebooks, codebook_size, out_group_size, in_group_size]
-    codebooks: torch.Tensor,
-    # [num_out_groups, 1, 1, 1]
-    scales: torch.Tensor,
-    output_partition_sizes: torch.IntTensor,
-    bias: Optional[torch.Tensor],
-) -> torch.Tensor:
-    weights = ops.aqlm_dequant(codes, codebooks, output_partition_sizes)
-
-    b_scales = scales.view(scales.shape[:-3] + (-1,)).expand(-1, weights.shape[1])
-    weights *= b_scales
-    return F.linear(input, weights, bias)
-
-
-def dequant_no_scale(
-    # [..., in_features]
-    input: torch.Tensor,
-    # [num_out_groups, num_in_groups, num_codebooks]
-    codes: torch.IntTensor,
-    # [num_codebooks, codebook_size, out_group_size, in_group_size]
-    codebooks: torch.Tensor,
-    # [num_out_groups, 1, 1, 1]
-    scales: torch.Tensor,
-    output_partition_sizes: torch.IntTensor,
-    bias: Optional[torch.Tensor],
-) -> torch.Tensor:
-    weights = ops.aqlm_dequant(codes, codebooks, output_partition_sizes)
-
-    return F.linear(input, weights, bias)
-
-
-# Compare the optimized 1x16 and 2x8 cuda decompression/dequant kernels against
-# the generic pytorch version.
-# Just visual comparison.
-def dequant_test(k: int, parts: torch.Tensor, nbooks: int, bits: int) -> None:
-    n = int(parts.sum().item())
-
-    device = torch.device("cuda:0")
-
-    code_range = (1 << bits) // 2
-    ingroups = 8
-
-    codes = torch.randint(
-        -code_range,
-        code_range,
-        size=(n, k // ingroups, nbooks),
-        dtype=get_int_dtype(bits),
-        device=device,
-    )
-
-    codebooks = torch.randn(
-        size=(parts.shape[0] * nbooks, 1 << bits, 1, 8),
-        dtype=torch.float16,
-        device=device,
-    )
-
-    count = 0
-    for index in range(16):
-        for i in range(8):
-            for book in range(nbooks):
-                codebooks[book, index, 0, i] = count * (10**book)
-            count += 1
-
-    print("codes shape", codes.shape)
-
-    for i in range(16):
-        for book in range(nbooks):
-            codes[0, i, book] = i
-            codes[0, -i, book] = i
-
-    weights = dequantize_weight(codes, codebooks, None)
-    weights2 = ops.aqlm_dequant(codes, codebooks, parts)
-
-    print("weights shape:", weights.shape)
-    print("weights2 shape:", weights2.shape)
-
-    print("weights are:", weights)
-    print("weights2 are:", weights2)
-
-    print("first 128 weights are", weights[0, 0:128].to(torch.int32))
-    print("first 128 weights2 are:", weights2[0, 0:128].to(torch.int32))
-
-    print("last 128 weights are", weights[0, -128:])
-    print("last 128 weights2 are:", weights2[0, -128:])
-
-
-def main():
-    parser = FlexibleArgumentParser(description="Benchmark aqlm performance.")
-
-    # Add arguments
-    parser.add_argument(
-        "--nbooks", type=int, default=1, help="Number of codebooks (default: 1)"
-    )
-    parser.add_argument(
-        "--bits",
-        type=int,
-        default=16,
-        help="Number of bits per code element (default: 16)",
-    )
-    parser.add_argument(
-        "--test",
-        type=bool,
-        default=False,
-        help="Run the decompression/dequant tester rather than benchmarking "
-        "(default: False)",
-    )
-
-    # Parse the arguments
-    args = parser.parse_args()
-
-    # Extract values
-    nbooks = args.nbooks
-    bits = args.bits
-
-    if args.test:
-        dequant_test(4096, torch.tensor((4096,)), nbooks, bits)
-        return
-
-    # Otherwise, benchmark.
-    methods = [
-        ops.aqlm_gemm,
-        dequant_out_scale,
-        generic_dequantize_gemm,
-        optimized_dequantize_gemm,
-        dequant_weight_scale,
-        torch_mult,
-        dequant_no_scale,
-    ]
-
-    filename = f"./aqlm_benchmark_{nbooks}x{bits}.csv"
-    print(f"writing benchmarks to file {filename}")
-    with open(filename, "w") as f:
-        sys.stdout = f
-
-        print("m | k | n | n parts", end="")
-        for method in methods:
-            print(f" | {method.__name__.replace('_', ' ')} (µs)", end="")
-        print("")
-
-        # These are reasonable prefill sizes.
-        ksandpartions = (
-            (4096, (4096, 4096, 4096)),
-            (4096, (4096,)),
-            (4096, (11008, 11008)),
-            (11008, (4096,)),
-        )
-
-        # reasonable ranges for m.
-        for m in [
-            1,
-            2,
-            4,
-            8,
-            10,
-            12,
-            14,
-            16,
-            24,
-            32,
-            48,
-            52,
-            56,
-            64,
-            96,
-            112,
-            128,
-            256,
-            512,
-            1024,
-            1536,
-            2048,
-            3072,
-            4096,
-        ]:
-            print(f"{m}", file=sys.__stdout__)
-            for ksp in ksandpartions:
-                run_grid(m, ksp[0], torch.tensor(ksp[1]), nbooks, bits, methods)
-
-        sys.stdout = sys.__stdout__
-
-
-def run_grid(m: int, k: int, parts: torch.Tensor, nbooks: int, bits: int, methods):
-    # I didn't see visible improvements from increasing these, but feel free :)
-    num_warmup_trials = 1
-    num_trials = 1
-
-    num_calls = 100
-
-    # warmup.
-    for method in methods:
-        for _ in range(num_warmup_trials):
-            run_timing(
-                num_calls=num_calls,
-                m=m,
-                k=k,
-                parts=parts,
-                nbooks=nbooks,
-                bits=bits,
-                method=method,
-            )
-
-    n = parts.sum().item()
-    print(f"{m} | {k} | {n} | {parts.tolist()}", end="")
-
-    for method in methods:
-        best_time_us = 1e20
-        for _ in range(num_trials):
-            kernel_dur_ms = run_timing(
-                num_calls=num_calls,
-                m=m,
-                k=k,
-                parts=parts,
-                nbooks=nbooks,
-                bits=bits,
-                method=method,
-            )
-
-            kernel_dur_us = 1000 * kernel_dur_ms
-
-            if kernel_dur_us < best_time_us:
-                best_time_us = kernel_dur_us
-
-        print(f" | {kernel_dur_us:.0f}", end="")
-
-    print("")
-
-
-def run_timing(
-    num_calls: int, m: int, k: int, parts: torch.Tensor, nbooks: int, bits: int, method
-) -> float:
-    n = int(parts.sum().item())
-
-    device = torch.device("cuda:0")
-
-    input = torch.randn((1, m, k), dtype=torch.float16, device=device)
-
-    code_range = (1 << bits) // 2
-    ingroups = 8
-
-    codes = torch.randint(
-        -code_range,
-        code_range,
-        size=(n, k // ingroups, nbooks),
-        dtype=get_int_dtype(bits),
-        device=device,
-    )
-
-    codebooks = torch.randn(
-        size=(parts.shape[0] * nbooks, 1 << bits, 1, 8),
-        dtype=torch.float16,
-        device=device,
-    )
-
-    scales = torch.randn(size=(n, 1, 1, 1), dtype=torch.float16, device=device)
-
-    # for comparison to just a pytorch mult.
-    weights = torch.randn((n, k), dtype=torch.float16, device=device)
-
-    start_event = torch.cuda.Event(enable_timing=True)
-    end_event = torch.cuda.Event(enable_timing=True)
-
-    start_event.record()
-
-    if method is torch_mult:
-        for i in range(num_calls):
-            torch_mult(input, weights, scales)
-    else:
-        for i in range(num_calls):
-            method(input, codes, codebooks, scales, parts, None)
-
-    end_event.record()
-    end_event.synchronize()
-
-    dur_ms = start_event.elapsed_time(end_event) / num_calls
-    return dur_ms
-
-
-if __name__ == "__main__":
-    sys.exit(main())

benchmarks/kernels/benchmark_bitblas.py

@@ -3,6 +3,8 @@
 # Copyright (c) Microsoft Corporation.
 # Licensed under the MIT License.
 
+from packaging import version
+
 from vllm.model_executor.layers.quantization.utils.bitblas_utils import (
     MINIMUM_BITBLAS_VERSION,
 )
@@ -10,7 +12,7 @@ from vllm.model_executor.layers.quantization.utils.bitblas_utils import (
 try:
     import bitblas
 
-    if bitblas.__version__ < MINIMUM_BITBLAS_VERSION:
+    if version.parse(bitblas.__version__) < version.parse(MINIMUM_BITBLAS_VERSION):
         raise ImportError(
             "bitblas version is wrong. Please "
             f"install bitblas>={MINIMUM_BITBLAS_VERSION}"

benchmarks/kernels/benchmark_grouped_gemm_cutlass.py

@@ -113,6 +113,7 @@ def bench_run(
         w2_scale: torch.Tensor,
         topk_weights: torch.Tensor,
         topk_ids: torch.Tensor,
+        per_act_token: bool,
         num_repeats: int,
     ):
         for _ in range(num_repeats):
@@ -124,7 +125,8 @@ def bench_run(
                 topk_ids,
                 w1_scale,
                 w2_scale,
-                a1_scale=a_scale,
+                per_act_token,
+                a1_scale=None,
             )
 
     def run_cutlass_from_graph(
@@ -148,7 +150,8 @@ def bench_run(
                 topk_ids,
                 w1_scale,
                 w2_scale,
-                a1_scale=a_scale,
+                per_act_token,
+                a1_scale=None,
             )
 
     def run_triton_from_graph(
@@ -227,6 +230,7 @@ def bench_run(
         "w2_q": w2_q,
         "w1_scale": w1_scale,
         "w2_scale": w2_scale,
+        "per_act_token": per_act_token,
         # cuda graph params
         "cutlass_graph": cutlass_graph,
         "triton_graph": triton_graph,
@@ -287,12 +291,13 @@ def bench_run(
         w2_scale,
         topk_weights,
         topk_ids,
+        per_act_token,
         num_warmup,
     )
 
     results.append(
         benchmark.Timer(
-            stmt="run_cutlass_moe(a, a_scale, w1_q, w2_q, w1_scale, w2_scale, topk_weights, topk_ids, num_runs)",  # noqa: E501
+            stmt="run_cutlass_moe(a, a_scale, w1_q, w2_q, w1_scale, w2_scale, topk_weights, topk_ids, per_act_token, num_runs)",  # noqa: E501
             globals=globals,
             label=label,
             sub_label=sub_label,

benchmarks/kernels/benchmark_machete.py

@@ -234,8 +234,11 @@ def marlin_create_bench_fn(bt: BenchmarkTensors) -> Callable:
 
         fn = lambda: ops.gptq_marlin_gemm(
             a=bt.a,
+            c=None,
             b_q_weight=w_q,
+            b_bias=None,
             b_scales=w_s,
+            global_scale=None,
             b_zeros=w_zp,
             g_idx=g_idx,
             perm=sort_indices,

benchmarks/kernels/benchmark_marlin.py

@@ -22,8 +22,16 @@ from vllm.model_executor.layers.quantization.utils.marlin_utils import (
     MARLIN_SUPPORTED_GROUP_SIZES,
     query_marlin_supported_quant_types,
 )
+from vllm.model_executor.layers.quantization.utils.marlin_utils_fp4 import (
+    FP4_MARLIN_SUPPORTED_GROUP_SIZES,
+    rand_marlin_weight_fp4_like,
+)
+from vllm.model_executor.layers.quantization.utils.marlin_utils_fp8 import (
+    marlin_quant_fp8_torch,
+)
 from vllm.model_executor.layers.quantization.utils.marlin_utils_test import (
     MarlinWorkspace,
+    awq_marlin_quantize,
     marlin_quantize,
 )
 from vllm.model_executor.layers.quantization.utils.marlin_utils_test_24 import (
@@ -35,7 +43,7 @@ from vllm.model_executor.layers.quantization.utils.quant_utils import (
     quantize_weights,
     sort_weights,
 )
-from vllm.scalar_type import ScalarType
+from vllm.scalar_type import ScalarType, scalar_types
 from vllm.utils import FlexibleArgumentParser
 
 DEFAULT_MODELS = ["meta-llama/Llama-2-7b-hf/TP1"]
@@ -57,80 +65,144 @@ def bench_run(
     size_n: int,
 ):
     label = "Quant Matmul"
-
     sub_label = "{}, act={} k_full={}, q={}, g={}, MKN=({}x{}x{})".format(
         model, act_order, is_k_full, str(quant_type), group_size, size_m, size_k, size_n
     )
-
     print(f"Testing: {sub_label}")
 
     a = torch.randn(size_m, size_k).to(torch.half).cuda()
     b = torch.rand(size_k, size_n).to(torch.half).cuda()
+    has_zp = quant_type in [scalar_types.uint4, scalar_types.uint8]
+    if act_order and (group_size == -1 or group_size == size_k or has_zp):
+        return
+    if size_k % group_size != 0:
+        return
 
-    a_tmp = torch.zeros(size_m, size_k).to(torch.half).cuda()
-
-    # Marlin quant
-    (
-        marlin_w_ref,
-        marlin_q_w,
-        marlin_s,
-        marlin_g_idx,
-        marlin_sort_indices,
-        marlin_rand_perm,
-    ) = marlin_quantize(b, quant_type, group_size, act_order)
-
-    # Marlin_24 quant
-    (marlin_24_w_ref, marlin_24_q_w_comp, marlin_24_meta, marlin_24_s) = (
-        marlin_24_quantize(b, quant_type, group_size)
+    marlin_24_supported = (
+        quant_type in GPTQ_MARLIN_24_SUPPORTED_QUANT_TYPES
+        and group_size in GPTQ_MARLIN_24_SUPPORTED_GROUP_SIZES
     )
-
-    marlin_zp = torch.empty(0, dtype=torch.int, device=b.device)
-
-    # GPTQ quant
-    (w_ref, q_w, s, g_idx, rand_perm) = gptq_quantize_weights(
-        b, quant_type, group_size, act_order
+    repack_supported = (
+        quant_type in GPTQ_MARLIN_24_SUPPORTED_QUANT_TYPES
+        and group_size in MARLIN_SUPPORTED_GROUP_SIZES
     )
-    q_w_gptq = gptq_pack(q_w, quant_type.size_bits, size_k, size_n)
-
-    # For act_order, sort the "weights" and "g_idx"
-    # so that group ids are increasing
-    repack_sort_indices = torch.empty(0, dtype=torch.int, device=b.device)
-    if act_order:
-        (q_w, g_idx, repack_sort_indices) = sort_weights(q_w, g_idx)
-
-    # Prepare
-    marlin_workspace = MarlinWorkspace(
-        size_n, GPTQ_MARLIN_MIN_THREAD_N, GPTQ_MARLIN_MAX_PARALLEL
-    )
-
-    marlin_24_workspace = MarlinWorkspace(
-        size_n, GPTQ_MARLIN_24_MIN_THREAD_N, GPTQ_MARLIN_24_MAX_PARALLEL
-    )
-    marlin_zp = torch.zeros_like(marlin_s, dtype=torch.int)
-
-    # AllSpark W8A16 quant
-    as_supported_case = (
+    allspark_supported = (
         quant_type in ALLSPARK_SUPPORTED_QUANT_TYPES
         and group_size == -1
         and not act_order
         and is_k_full
     )
-    if as_supported_case:
-        properties = torch.cuda.get_device_properties(b.device.index)
-        sm_count = properties.multi_processor_count
-        sm_version = properties.major * 10 + properties.minor
 
-        supported_arch = sm_version >= 80 and sm_version < 90
-        as_supported_case = as_supported_case and supported_arch
-        if supported_arch:
-            has_zp = False
-            w_ref, qw, s, zp = quantize_weights(b, quant_type, group_size, has_zp)
-            qw = qw.to(torch.uint8)
-
-            qw_reorder, s_reorder, zp_reorder = ops.allspark_repack_weight(
-                qw, s, zp, has_zp
+    def gen_marlin_params():
+        # Marlin quant
+        marlin_g_idx = marlin_sort_indices = marlin_zp = marlin_s2 = None
+        if quant_type == scalar_types.float4_e2m1f:
+            if group_size != 16 or act_order:
+                return
+            marlin_w_ref, marlin_q_w, marlin_s, marlin_s2 = rand_marlin_weight_fp4_like(
+                b.T, group_size
             )
-            CUBLAS_M_THRESHOLD = ALLSPARK_AMPERE_M_CUBLAS_THRESHOLD
+        elif quant_type == scalar_types.float8_e4m3fn:
+            if group_size not in [-1, 128] or act_order:
+                return
+            marlin_w_ref, marlin_q_w, marlin_s = marlin_quant_fp8_torch(b.T, group_size)
+        elif group_size == 16:
+            return
+        elif has_zp:
+            marlin_w_ref, marlin_q_w, marlin_s, marlin_zp = awq_marlin_quantize(
+                b, quant_type, group_size
+            )
+        else:
+            marlin_w_ref, marlin_q_w, marlin_s, marlin_g_idx, marlin_sort_indices, _ = (
+                marlin_quantize(b, quant_type, group_size, act_order)
+            )
+        return (
+            marlin_w_ref,
+            marlin_q_w,
+            marlin_s,
+            marlin_s2,
+            marlin_zp,
+            marlin_g_idx,
+            marlin_sort_indices,
+        )
+
+    def gen_marlin_24_params():
+        marlin_24_w_ref = marlin_24_q_w_comp = marlin_24_meta = marlin_24_s = None
+        if marlin_24_supported:
+            (marlin_24_w_ref, marlin_24_q_w_comp, marlin_24_meta, marlin_24_s) = (
+                marlin_24_quantize(b, quant_type, group_size)
+            )
+        return (marlin_24_w_ref, marlin_24_q_w_comp, marlin_24_meta, marlin_24_s)
+
+    def gen_repack_params():
+        q_w_gptq = None
+        repack_sort_indices = None
+        if repack_supported:
+            (w_ref, q_w, s, g_idx, rand_perm) = gptq_quantize_weights(
+                b, quant_type, group_size, act_order
+            )
+            q_w_gptq = gptq_pack(q_w, quant_type.size_bits, size_k, size_n)
+
+            # For act_order, sort the "weights" and "g_idx"
+            # so that group ids are increasing
+            repack_sort_indices = torch.empty(0, dtype=torch.int, device=b.device)
+            if act_order:
+                (q_w, g_idx, repack_sort_indices) = sort_weights(q_w, g_idx)
+        return q_w_gptq, repack_sort_indices
+
+    def gen_allspark_params():
+        qw_reorder = s_reorder = zp_reorder = sm_count = sm_version = (
+            CUBLAS_M_THRESHOLD
+        ) = None
+        nonlocal allspark_supported
+        if allspark_supported:
+            properties = torch.cuda.get_device_properties(b.device.index)
+            sm_count = properties.multi_processor_count
+            sm_version = properties.major * 10 + properties.minor
+
+            supported_arch = sm_version >= 80 and sm_version < 90
+            allspark_supported = allspark_supported and supported_arch
+            if supported_arch:
+                w_ref, qw, s, zp = quantize_weights(b, quant_type, group_size, has_zp)
+                qw = qw.to(torch.uint8)
+
+                qw_reorder, s_reorder, zp_reorder = ops.allspark_repack_weight(
+                    qw, s, zp, has_zp
+                )
+                CUBLAS_M_THRESHOLD = ALLSPARK_AMPERE_M_CUBLAS_THRESHOLD
+        return (
+            qw_reorder,
+            s_reorder,
+            zp_reorder,
+            sm_count,
+            sm_version,
+            CUBLAS_M_THRESHOLD,
+        )
+
+    (
+        marlin_w_ref,
+        marlin_q_w,
+        marlin_s,
+        marlin_s2,
+        marlin_zp,
+        marlin_g_idx,
+        marlin_sort_indices,
+    ) = gen_marlin_params()
+    marlin_24_w_ref, marlin_24_q_w_comp, marlin_24_meta, marlin_24_s = (
+        gen_marlin_24_params()
+    )
+    q_w_gptq, repack_sort_indices = gen_repack_params()
+    qw_reorder, s_reorder, zp_reorder, sm_count, sm_version, CUBLAS_M_THRESHOLD = (
+        gen_allspark_params()
+    )
+
+    # Prepare
+    marlin_workspace = MarlinWorkspace(
+        size_n, GPTQ_MARLIN_MIN_THREAD_N, GPTQ_MARLIN_MAX_PARALLEL
+    )
+    marlin_24_workspace = MarlinWorkspace(
+        size_n, GPTQ_MARLIN_24_MIN_THREAD_N, GPTQ_MARLIN_24_MAX_PARALLEL
+    )
 
     globals = {
         # Gen params
@@ -140,15 +212,14 @@ def bench_run(
         "size_n": size_n,
         "size_k": size_k,
         "a": a,
-        "a_tmp": a_tmp,
         # Marlin params
         "marlin_w_ref": marlin_w_ref,
         "marlin_q_w": marlin_q_w,
         "marlin_s": marlin_s,
+        "marlin_s2": marlin_s2,
         "marlin_zp": marlin_zp,
         "marlin_g_idx": marlin_g_idx,
         "marlin_sort_indices": marlin_sort_indices,
-        "marlin_rand_perm": marlin_rand_perm,
         "marlin_workspace": marlin_workspace,
         "is_k_full": is_k_full,
         # Marlin_24 params
@@ -161,12 +232,12 @@ def bench_run(
         "q_w_gptq": q_w_gptq,
         "repack_sort_indices": repack_sort_indices,
         # AllSpark W8A16 params
-        "qw_reorder": qw_reorder if as_supported_case else None,
-        "s_reorder": s_reorder if as_supported_case else None,
-        "zp_reorder": zp_reorder if as_supported_case else None,
-        "sm_count": sm_count if as_supported_case else None,
-        "sm_version": sm_version if as_supported_case else None,
-        "CUBLAS_M_THRESHOLD": CUBLAS_M_THRESHOLD if as_supported_case else None,
+        "qw_reorder": qw_reorder,
+        "s_reorder": s_reorder,
+        "zp_reorder": zp_reorder,
+        "sm_count": sm_count,
+        "sm_version": sm_version,
+        "CUBLAS_M_THRESHOLD": CUBLAS_M_THRESHOLD,
         # Kernels
         "gptq_marlin_gemm": ops.gptq_marlin_gemm,
         "gptq_marlin_24_gemm": ops.gptq_marlin_24_gemm,
@@ -177,7 +248,7 @@ def bench_run(
     min_run_time = 1
 
     # Warmup pytorch
-    for i in range(5):
+    for _ in range(5):
         torch.matmul(a, marlin_w_ref)
 
     results.append(
@@ -192,17 +263,17 @@ def bench_run(
 
     results.append(
         benchmark.Timer(
-            stmt="output = gptq_marlin_gemm(a, marlin_q_w, marlin_s, marlin_zp, marlin_g_idx, marlin_sort_indices, marlin_workspace.scratch, quant_type, size_m, size_n, size_k, is_k_full, False, False, False)",  # noqa: E501
+            stmt="output = gptq_marlin_gemm(a, None, marlin_q_w, marlin_s, marlin_s2, marlin_zp, marlin_g_idx, marlin_sort_indices, marlin_workspace.scratch, quant_type, size_m, size_n, size_k, is_k_full, False, False, False)",  # noqa: E501
             globals=globals,
             label=label,
             sub_label=sub_label,
-            description="gptq_marlin_gemm_fp16",
+            description="gptq_marlin_gemm",
         ).blocked_autorange(min_run_time=min_run_time)
     )
 
     results.append(
         benchmark.Timer(
-            stmt="output = gptq_marlin_gemm(a, marlin_q_w, marlin_s, marlin_zp, marlin_g_idx, marlin_sort_indices, marlin_workspace.scratch, quant_type, size_m, size_n, size_k, is_k_full, False, True, False)",  # noqa: E501
+            stmt="output = gptq_marlin_gemm(a, None, marlin_q_w, marlin_s, marlin_s2, marlin_zp, marlin_g_idx, marlin_sort_indices, marlin_workspace.scratch, quant_type, size_m, size_n, size_k, is_k_full, False, True, False)",  # noqa: E501
             globals=globals,
             label=label,
             sub_label=sub_label,
@@ -210,10 +281,7 @@ def bench_run(
         ).blocked_autorange(min_run_time=min_run_time)
     )
 
-    if (
-        quant_type in GPTQ_MARLIN_24_SUPPORTED_QUANT_TYPES
-        and group_size in GPTQ_MARLIN_24_SUPPORTED_GROUP_SIZES
-    ):
+    if marlin_24_supported:
         results.append(
             benchmark.Timer(
                 stmt="output = gptq_marlin_24_gemm(a, marlin_24_q_w_comp, marlin_24_meta, marlin_24_s, marlin_24_workspace.scratch, quant_type, size_m, size_n, size_k)",  # noqa: E501
@@ -224,17 +292,18 @@ def bench_run(
             ).blocked_autorange(min_run_time=min_run_time)
         )
 
-    results.append(
-        benchmark.Timer(
-            stmt="q_res = gptq_marlin_repack(q_w_gptq, repack_sort_indices, size_k, size_n, quant_type.size_bits)",  # noqa: E501
-            globals=globals,
-            label=label,
-            sub_label=sub_label,
-            description="gptq_marlin_repack",
-        ).blocked_autorange(min_run_time=min_run_time)
-    )
+    if repack_supported:
+        results.append(
+            benchmark.Timer(
+                stmt="q_res = gptq_marlin_repack(q_w_gptq, repack_sort_indices, size_k, size_n, quant_type.size_bits)",  # noqa: E501
+                globals=globals,
+                label=label,
+                sub_label=sub_label,
+                description="gptq_marlin_repack",
+            ).blocked_autorange(min_run_time=min_run_time)
+        )
 
-    if as_supported_case:
+    if allspark_supported:
         results.append(
             benchmark.Timer(
                 stmt="output = allspark_w8a16_gemm(a, qw_reorder, s_reorder, zp_reorder, size_n, group_size, sm_count, sm_version, CUBLAS_M_THRESHOLD, False, True)",  # noqa: E501
@@ -250,7 +319,6 @@ def main(args):
     print("Benchmarking models:")
     for i, model in enumerate(args.models):
         print(f"[{i}]  {model}")
-
     results: list[benchmark.Measurement] = []
 
     for model in args.models:
@@ -278,14 +346,17 @@ def main(args):
                     ):
                         continue
 
-                    for quant_type in query_marlin_supported_quant_types(False):
+                    for quant_type in query_marlin_supported_quant_types():
                         if (
                             len(args.limit_num_bits) > 0
                             and quant_type.size_bits not in args.limit_num_bits
                         ):
                             continue
 
-                        for group_size in MARLIN_SUPPORTED_GROUP_SIZES:
+                        for group_size in (
+                            MARLIN_SUPPORTED_GROUP_SIZES
+                            + FP4_MARLIN_SUPPORTED_GROUP_SIZES
+                        ):
                             if (
                                 len(args.limit_group_size) > 0
                                 and group_size not in args.limit_group_size

benchmarks/kernels/benchmark_moe.py

@@ -3,6 +3,7 @@
 
 import argparse
 import json
+import os
 import time
 from contextlib import nullcontext
 from datetime import datetime
@@ -22,6 +23,13 @@ from vllm.utils import FlexibleArgumentParser
 FP8_DTYPE = current_platform.fp8_dtype()
 
 
+def ensure_divisibility(numerator, denominator, text):
+    """Ensure that numerator is divisible by the denominator."""
+    assert numerator % denominator == 0, "{} {} is not divisible by tp {}.".format(
+        text, numerator, denominator
+    )
+
+
 class BenchmarkConfig(TypedDict):
     BLOCK_SIZE_M: int
     BLOCK_SIZE_N: int
@@ -86,6 +94,9 @@ def benchmark_config(
             (num_experts, 2 * shard_intermediate_size), dtype=torch.float32
         )
         w2_scale = torch.randn((hidden_size, num_experts), dtype=torch.float32)
+    if use_deep_gemm:
+        # we use the default block shape for deepgemm
+        block_quant_shape = [128, 128]
     if use_fp8_w8a8:
         if block_quant_shape:
             block_n, block_k = block_quant_shape[0], block_quant_shape[1]
@@ -532,6 +543,7 @@ def save_configs(
     use_fp8_w8a8: bool,
     use_int8_w8a16: bool,
     block_quant_shape: list[int],
+    save_dir: str,
 ) -> None:
     dtype_str = get_config_dtype_str(
         dtype, use_int8_w8a16=use_int8_w8a16, use_fp8_w8a8=use_fp8_w8a8
@@ -542,7 +554,8 @@ def save_configs(
     filename = get_config_file_name(
         num_experts, shard_intermediate_size // 2, dtype_str, block_quant_shape
     )
-
+    os.makedirs(save_dir, exist_ok=True)
+    filename = os.path.join(save_dir, filename)
     print(f"Writing best config to {filename}...")
     with open(filename, "w") as f:
         json.dump(configs, f, indent=4)
@@ -567,22 +580,26 @@ def main(args: argparse.Namespace):
         E = config.ffn_config.moe_num_experts
         topk = config.ffn_config.moe_top_k
         intermediate_size = config.ffn_config.ffn_hidden_size
-        shard_intermediate_size = 2 * intermediate_size // args.tp_size
     elif config.architectures[0] == "JambaForCausalLM":
         E = config.num_experts
         topk = config.num_experts_per_tok
         intermediate_size = config.intermediate_size
-        shard_intermediate_size = 2 * intermediate_size // args.tp_size
-    elif config.architectures[0] in ("DeepseekV3ForCausalLM", "DeepseekV2ForCausalLM"):
+    elif config.architectures[0] in (
+        "DeepseekV3ForCausalLM",
+        "DeepseekV2ForCausalLM",
+        "Glm4MoeForCausalLM",
+    ):
         E = config.n_routed_experts
         topk = config.num_experts_per_tok
         intermediate_size = config.moe_intermediate_size
-        shard_intermediate_size = 2 * intermediate_size // args.tp_size
     elif config.architectures[0] in ("Qwen2MoeForCausalLM", "Qwen3MoeForCausalLM"):
         E = config.num_experts
         topk = config.num_experts_per_tok
         intermediate_size = config.moe_intermediate_size
-        shard_intermediate_size = 2 * intermediate_size // args.tp_size
+    elif config.architectures[0] in ("HunYuanMoEV1ForCausalLM"):
+        E = config.num_experts
+        topk = config.moe_topk[0]
+        intermediate_size = config.moe_intermediate_size[0]
     else:
         # Support for llama4
         config = config.get_text_config()
@@ -590,8 +607,14 @@ def main(args: argparse.Namespace):
         E = config.num_local_experts
         topk = config.num_experts_per_tok
         intermediate_size = config.intermediate_size
+    enable_ep = bool(args.enable_expert_parallel)
+    if enable_ep:
+        ensure_divisibility(E, args.tp_size, "Number of experts")
+        E = E // args.tp_size
+        shard_intermediate_size = 2 * intermediate_size
+    else:
+        ensure_divisibility(intermediate_size, args.tp_size, "intermediate_size")
         shard_intermediate_size = 2 * intermediate_size // args.tp_size
-
     hidden_size = config.hidden_size
     dtype = torch.float16 if current_platform.is_rocm() else config.torch_dtype
     use_fp8_w8a8 = args.dtype == "fp8_w8a8"
@@ -620,7 +643,7 @@ def main(args: argparse.Namespace):
             4096,
         ]
     else:
-        batch_sizes = [args.batch_size]
+        batch_sizes = args.batch_size
 
     use_deep_gemm = bool(args.use_deep_gemm)
 
@@ -687,6 +710,7 @@ def main(args: argparse.Namespace):
             use_fp8_w8a8,
             use_int8_w8a16,
             block_quant_shape,
+            args.save_dir,
         )
         end = time.time()
         print(f"Tuning took {end - start:.2f} seconds")
@@ -723,12 +747,16 @@ if __name__ == "__main__":
     parser.add_argument(
         "--tp-size", "-tp", "--tensor-parallel-size", type=int, default=2
     )
+    parser.add_argument("--enable-expert-parallel", "-enable-ep", action="store_true")
     parser.add_argument(
         "--dtype", type=str, choices=["auto", "fp8_w8a8", "int8_w8a16"], default="auto"
     )
     parser.add_argument("--use-deep-gemm", action="store_true")
+    parser.add_argument(
+        "--save-dir", type=str, default="./", help="Directory to save tuned results"
+    )
     parser.add_argument("--seed", type=int, default=0)
-    parser.add_argument("--batch-size", type=int, required=False)
+    parser.add_argument("--batch-size", type=int, nargs="+", required=False)
     parser.add_argument("--tune", action="store_true")
     parser.add_argument("--trust-remote-code", action="store_true")
     parser.add_argument("--model-prefix", type=str, required=False)

benchmarks/kernels/benchmark_moe_align_block_size.py

@@ -0,0 +1,74 @@
+# SPDX-License-Identifier: Apache-2.0
+# SPDX-FileCopyrightText: Copyright contributors to the vLLM project
+import argparse
+import itertools
+
+import torch
+
+from vllm.model_executor.layers.fused_moe.moe_align_block_size import (
+    moe_align_block_size,
+)
+from vllm.triton_utils import triton
+
+
+def get_topk_ids(num_tokens: int, num_experts: int, topk: int) -> torch.Tensor:
+    return torch.stack(
+        [
+            torch.randperm(num_experts, dtype=torch.int32, device="cuda")[:topk]
+            for _ in range(num_tokens)
+        ]
+    )
+
+
+# test configurations
+num_tokens_range = [1, 16, 256, 4096]
+num_experts_range = [16, 64, 224, 256, 280, 512]
+topk_range = [1, 2, 8]
+configs = list(itertools.product(num_tokens_range, num_experts_range, topk_range))
+
+
+@triton.testing.perf_report(
+    triton.testing.Benchmark(
+        x_names=["num_tokens", "num_experts", "topk"],
+        x_vals=configs,
+        line_arg="provider",
+        line_vals=["vllm"],
+        line_names=["vLLM"],
+        plot_name="moe-align-block-size-performance",
+        args={},
+    )
+)
+def benchmark(num_tokens, num_experts, topk, provider):
+    """Benchmark function for Triton."""
+    block_size = 256
+    topk_ids = get_topk_ids(num_tokens, num_experts, topk)
+
+    quantiles = [0.5, 0.2, 0.8]
+
+    if provider == "vllm":
+        ms, min_ms, max_ms = triton.testing.do_bench(
+            lambda: moe_align_block_size(topk_ids, block_size, num_experts),
+            quantiles=quantiles,
+        )
+
+    return 1000 * ms, 1000 * max_ms, 1000 * min_ms
+
+
+if __name__ == "__main__":
+    parser = argparse.ArgumentParser()
+    parser.add_argument(
+        "--num_experts",
+        type=int,
+        default=64,
+        choices=[8, 16, 32, 64, 128, 256],
+    )
+    parser.add_argument(
+        "--topk",
+        type=int,
+        default=8,
+        choices=[2, 4, 8],
+        help="Top-k value for correctness check.",
+    )
+    args = parser.parse_args()
+
+    benchmark.run(print_data=True, show_plots=True)

benchmarks/kernels/benchmark_moe_permute_unpermute.py

@@ -8,12 +8,13 @@ import ray
 import torch
 from transformers import AutoConfig
 
-from vllm.model_executor.layers.fused_moe.deep_gemm_moe import (
+from vllm.model_executor.layers.fused_moe.fused_moe import *
+from vllm.model_executor.layers.fused_moe.moe_permute_unpermute import (
     _moe_permute,
     _moe_unpermute_and_reduce,
+    moe_permute,
+    moe_unpermute,
 )
-from vllm.model_executor.layers.fused_moe.fused_moe import *
-from vllm.model_executor.layers.fused_moe.moe_permute_unpermute import *
 from vllm.model_executor.layers.fused_moe.utils import _fp8_quantize
 from vllm.platforms import current_platform
 from vllm.utils import FlexibleArgumentParser
@@ -63,18 +64,19 @@ def benchmark_permute(
 
     def run():
         if use_customized_permute:
-            (permuted_hidden_states, first_token_off, inv_perm_idx, m_indices) = (
-                moe_permute(
-                    qhidden_states,
-                    topk_weights=topk_weights,
-                    topk_ids=topk_ids,
-                    token_expert_indices=token_expert_indices,
-                    topk=topk,
-                    n_expert=num_experts,
-                    n_local_expert=num_experts,
-                    expert_map=None,
-                    align_block_size=align_block_size,
-                )
+            (
+                permuted_hidden_states,
+                a1q_scale,
+                first_token_off,
+                inv_perm_idx,
+                m_indices,
+            ) = moe_permute(
+                qhidden_states,
+                a1q_scale=None,
+                topk_ids=topk_ids,
+                n_expert=num_experts,
+                expert_map=None,
+                align_block_size=align_block_size,
             )
         else:
             (
@@ -150,18 +152,19 @@ def benchmark_unpermute(
 
     def prepare():
         if use_customized_permute:
-            (permuted_hidden_states, first_token_off, inv_perm_idx, m_indices) = (
-                moe_permute(
-                    qhidden_states,
-                    topk_weights=topk_weights,
-                    topk_ids=topk_ids,
-                    token_expert_indices=token_expert_indices,
-                    topk=topk,
-                    n_expert=num_experts,
-                    n_local_expert=num_experts,
-                    expert_map=None,
-                    align_block_size=align_block_size,
-                )
+            (
+                permuted_hidden_states,
+                a1q_scale,
+                first_token_off,
+                inv_perm_idx,
+                m_indices,
+            ) = moe_permute(
+                qhidden_states,
+                a1q_scale=None,
+                topk_ids=topk_ids,
+                n_expert=num_experts,
+                expert_map=None,
+                align_block_size=align_block_size,
             )
             # convert to fp16/bf16 as gemm output
             return (
@@ -191,16 +194,19 @@ def benchmark_unpermute(
 
     def run(input: tuple):
         if use_customized_permute:
-            (permuted_hidden_states, first_token_off, inv_perm_idx, m_indices) = input
+            (
+                permuted_hidden_states,
+                first_token_off,
+                inv_perm_idx,
+                m_indices,
+            ) = input
+            output = torch.empty_like(hidden_states)
             moe_unpermute(
+                output,
                 permuted_hidden_states,
                 topk_weights,
-                topk_ids,
                 inv_perm_idx,
                 first_token_off,
-                topk,
-                num_experts,
-                num_experts,
             )
         else:
             (
@@ -211,7 +217,11 @@ def benchmark_unpermute(
                 inv_perm,
             ) = input
             _moe_unpermute_and_reduce(
-                output_hidden_states, permuted_hidden_states, inv_perm, topk_weights
+                output_hidden_states,
+                permuted_hidden_states,
+                inv_perm,
+                topk_weights,
+                True,
             )
 
     # JIT compilation & warmup
@@ -318,6 +328,7 @@ def main(args: argparse.Namespace):
     elif (
         config.architectures[0] == "DeepseekV3ForCausalLM"
         or config.architectures[0] == "DeepseekV2ForCausalLM"
+        or config.architectures[0] == "Glm4MoeForCausalLM"
     ):
         E = config.n_routed_experts
         topk = config.num_experts_per_tok

benchmarks/kernels/benchmark_mrope.py

@@ -0,0 +1,328 @@
+# SPDX-License-Identifier: Apache-2.0
+# SPDX-FileCopyrightText: Copyright contributors to the vLLM project
+
+# This script benchmarks the mrope kernel (mainly for Qwen2VL and Qwen2.5VL models).
+# It generates test data, runs benchmarks, and saves results to a CSV file.
+#
+# The CSV file (named with current date/time) contains these columns:
+# model_name, tp_size, num_tokens, num_heads, num_kv_heads, head_dim, max_position,
+# rope_theta, is_neox_style, rope_scaling, dtype, torch_mean, torch_median, torch_p99,
+# torch_min, torch_max, triton_mean, triton_median, triton_p99, triton_min, triton_max,
+# speedup
+#
+# == Usage Examples ==
+#
+# Single model benchmark:
+# python3 benchmark_mrope.py --model-name Qwen/Qwen2-VL-7B-Instruct --tp-size 1 \
+#   --warmup-iter 10 --benchmark-iter 100 --dtype bfloat16 --seed 0 --num-tokens 1024
+#
+# All models benchmark:
+# python3 benchmark_mrope.py --model-name "" --tp-size 1 --warmup-iter 10 \
+#   --benchmark-iter 100 --dtype bfloat16 --seed 0 --num-tokens 1024
+#
+# All models with different TP sizes:
+# python3 benchmark_mrope.py --model-name "" --tp-size 1 2 4 8 --warmup-iter 10 \
+#   --benchmark-iter 100 --dtype bfloat16 --seed 0 --num-tokens 1024
+#
+# All models with different token counts:
+# python3 benchmark_mrope.py --model-name "" --tp-size 1 --warmup-iter 10 \
+#   --benchmark-iter 100 --dtype bfloat16 --seed 0 --num-tokens 1024 4096 16384
+import csv
+import os
+import time
+from datetime import datetime
+from typing import Any
+
+import numpy as np
+import torch
+
+from vllm.model_executor.layers.rotary_embedding import get_rope
+from vllm.platforms import current_platform
+from vllm.transformers_utils.config import get_config
+from vllm.utils import FlexibleArgumentParser
+
+device = torch.device("cuda" if torch.cuda.is_available() else "cpu")
+
+
+def generate_test_data(
+    num_tokens: int,
+    num_q_heads: int,
+    num_kv_heads: int,
+    head_size: int,
+    max_position_embeddings: int,
+    dtype: torch.dtype,
+    device: torch.device,
+):
+    """Generate test data for given configuration."""
+    # Create 2D positions (3, num_tokens) for multimodal case
+    positions = torch.randint(
+        0, max_position_embeddings // 4, (3, num_tokens), device=device
+    )
+
+    # Create query and key tensors
+    query = torch.randn(num_tokens, num_q_heads * head_size, dtype=dtype, device=device)
+    key = torch.randn(num_tokens, num_kv_heads * head_size, dtype=dtype, device=device)
+
+    return positions, query, key
+
+
+def calculate_stats(times: list[float]) -> dict[str, float]:
+    """Calculate statistics from a list of times."""
+    times_array = np.array(times)
+    return {
+        "mean": np.mean(times_array),
+        "median": np.median(times_array),
+        "p99": np.percentile(times_array, 99),
+        "min": np.min(times_array),
+        "max": np.max(times_array),
+    }
+
+
+def benchmark_mrope(
+    model_name: str,
+    num_tokens: int,
+    head_dim: int,
+    tp_size: int,
+    num_heads: int,
+    num_kv_heads: int,
+    max_position: int = 8192,
+    rope_theta: float = 10000,
+    is_neox_style: bool = True,
+    rope_scaling: dict[str, Any] = None,
+    dtype: torch.dtype = torch.bfloat16,
+    seed: int = 0,
+    warmup_iter: int = 10,
+    benchmark_iter: int = 100,
+    csv_writer=None,
+):
+    current_platform.seed_everything(seed)
+    torch.set_default_device(device)
+    # the parameters to compute the q k v size based on tp_size
+    mrope_helper_class = get_rope(
+        head_size=head_dim,
+        rotary_dim=head_dim,
+        max_position=max_position,
+        base=rope_theta,
+        is_neox_style=is_neox_style,
+        rope_scaling=rope_scaling,
+        dtype=dtype,
+    ).to(device=device)
+
+    print(80 * "=")
+    print(
+        f"Evaluating model: {model_name} "
+        f"with tp_size: {tp_size} "
+        f"and num_tokens: {num_tokens}, "
+        f"dtype: {dtype}"
+    )
+
+    # create q k v input tensors
+    # create rotary pos emb input tensors
+    positions, query, key = generate_test_data(
+        num_tokens, num_heads, num_kv_heads, head_dim, max_position, dtype, device
+    )
+
+    # Warm up
+    for _ in range(warmup_iter):
+        mrope_helper_class.forward_native(
+            positions,
+            query.clone(),
+            key.clone(),
+        )
+
+        mrope_helper_class.forward_cuda(
+            positions,
+            query.clone(),
+            key.clone(),
+        )
+
+    torch.cuda.synchronize()
+
+    # Time reference implementation
+    torch_times = []
+    for _ in range(benchmark_iter):
+        query_clone = query.clone()
+        key_clone = key.clone()
+        torch.cuda.synchronize()
+        start_time = time.time()
+
+        mrope_helper_class.forward_native(
+            positions,
+            query_clone,
+            key_clone,
+        )
+
+        torch.cuda.synchronize()
+        torch_times.append(time.time() - start_time)
+
+    # Time triton kernel implementation
+    triton_times = []
+    for _ in range(benchmark_iter):
+        query_clone = query.clone()
+        key_clone = key.clone()
+        torch.cuda.synchronize()
+        start_time = time.time()
+        mrope_helper_class.forward_cuda(
+            positions,
+            query_clone,
+            key_clone,
+        )
+        torch.cuda.synchronize()
+        triton_times.append(time.time() - start_time)
+
+    # Calculate statistics
+    torch_stats = calculate_stats(torch_times)
+    triton_stats = calculate_stats(triton_times)
+    print(f"\nPerformance for config ({num_tokens}, {num_heads}, {num_kv_heads}):")
+
+    print(
+        f"Torch implementation: "
+        f"mean={torch_stats['mean']:.8f}s, "
+        f"median={torch_stats['median']:.8f}s, "
+        f"p99={torch_stats['p99']:.8f}s"
+    )
+
+    print(
+        f"Triton implementation: "
+        f"mean={triton_stats['mean']:.8f}s, "
+        f"median={triton_stats['median']:.8f}s, "
+        f"p99={triton_stats['p99']:.8f}s"
+    )
+
+    print(
+        f"Triton Speedup over Torch: {torch_stats['mean'] / triton_stats['mean']:.8f}x"
+    )
+
+    # Write to CSV
+    if csv_writer:
+        row = [
+            model_name,
+            tp_size,
+            num_tokens,
+            num_heads,
+            num_kv_heads,
+            head_dim,
+            max_position,
+            rope_theta,
+            is_neox_style,
+            str(rope_scaling),
+            str(dtype).split(".")[-1],
+            torch_stats["mean"],
+            torch_stats["median"],
+            torch_stats["p99"],
+            torch_stats["min"],
+            torch_stats["max"],
+            triton_stats["mean"],
+            triton_stats["median"],
+            triton_stats["p99"],
+            triton_stats["min"],
+            triton_stats["max"],
+            torch_stats["mean"] / triton_stats["mean"],  # speedup
+        ]
+        csv_writer.writerow(row)
+
+    return torch_stats, triton_stats
+
+
+if __name__ == "__main__":
+    parser = FlexibleArgumentParser(
+        description="Benchmark the rotary embedding kernels."
+    )
+    parser.add_argument("--model-name", type=str, default="")
+    parser.add_argument("--tp-size", type=int, default=1)
+    parser.add_argument("--warmup-iter", type=int, default=10)
+    parser.add_argument("--benchmark-iter", type=int, default=100)
+    parser.add_argument("--dtype", type=str, choices=["bfloat16"], default="bfloat16")
+    parser.add_argument("--seed", type=int, default=0)
+    parser.add_argument("--num-tokens", type=int, nargs="+", required=False)
+    parser.add_argument("--trust-remote-code", action="store_true")
+    parser.add_argument("--output-csv", type=str, default="mrope_benchmark_results.csv")
+    args = parser.parse_args()
+    print(args)
+
+    # Create CSV file for results
+    timestamp = datetime.now().strftime("%Y%m%d_%H%M%S")
+    csv_filename = f"{os.path.splitext(args.output_csv)[0]}_{timestamp}.csv"
+
+    with open(csv_filename, "w", newline="") as csvfile:
+        csv_writer = csv.writer(csvfile)
+        # Write header
+        header = [
+            "model_name",
+            "tp_size",
+            "num_tokens",
+            "num_heads",
+            "num_kv_heads",
+            "head_dim",
+            "max_position",
+            "rope_theta",
+            "is_neox_style",
+            "rope_scaling",
+            "dtype",
+            "torch_mean",
+            "torch_median",
+            "torch_p99",
+            "torch_min",
+            "torch_max",
+            "triton_mean",
+            "triton_median",
+            "triton_p99",
+            "triton_min",
+            "triton_max",
+            "speedup",
+        ]
+        csv_writer.writerow(header)
+
+        model_tp_dict = {}
+        if args.model_name == "":
+            model_tp_dict = {
+                "Qwen/Qwen2-VL-2B-Instruct": [1],
+                "Qwen/Qwen2-VL-7B-Instruct": [1],
+                "Qwen/Qwen2-VL-72B-Instruct": [2, 4, 8],
+                "Qwen/Qwen2.5-VL-3B-Instruct": [1, 2, 4, 8],
+                "Qwen/Qwen2.5-VL-7B-Instruct": [1, 2, 4, 8],
+                "Qwen/Qwen2.5-VL-72B-Instruct": [2, 4, 8],
+            }
+        else:
+            model_tp_dict[args.model_name] = [args.tp_size]
+
+        if args.num_tokens is None:
+            num_tokens_list = [2**i for i in range(0, 18)]
+        else:
+            num_tokens_list = args.num_tokens
+
+        for model_name, tp_list in model_tp_dict.items():
+            config = get_config(model_name, trust_remote_code=args.trust_remote_code)
+            for tp_size in tp_list:
+                # get the model config
+                total_num_kv_heads = config.num_key_value_heads
+                total_num_heads = config.num_attention_heads
+                num_heads = total_num_heads // tp_size
+                num_kv_heads = max(1, total_num_kv_heads // tp_size)
+                head_dim = config.hidden_size // total_num_heads
+                q_size = num_heads * head_dim
+                kv_size = num_kv_heads * head_dim
+                is_neox_style = True
+                rope_theta = config.rope_theta
+                max_position = config.max_position_embeddings
+
+                for num_tokens in num_tokens_list:
+                    benchmark_mrope(
+                        model_name=model_name,
+                        num_tokens=num_tokens,
+                        head_dim=head_dim,
+                        tp_size=tp_size,
+                        num_heads=num_heads,
+                        num_kv_heads=num_kv_heads,
+                        max_position=max_position,
+                        rope_theta=rope_theta,
+                        is_neox_style=is_neox_style,
+                        rope_scaling=config.rope_scaling,
+                        dtype=getattr(torch, args.dtype),
+                        seed=args.seed,
+                        warmup_iter=args.warmup_iter,
+                        benchmark_iter=args.benchmark_iter,
+                        csv_writer=csv_writer,
+                    )
+
+    print(f"Benchmark results saved to {csv_filename}")

benchmarks/kernels/benchmark_per_token_group_quant.py

@@ -0,0 +1,159 @@
+# SPDX-License-Identifier: Apache-2.0
+# SPDX-FileCopyrightText: Copyright contributors to the vLLM project
+
+import argparse
+import math
+from contextlib import contextmanager
+from typing import Callable
+from unittest.mock import patch
+
+import torch
+
+from vllm.model_executor.layers.quantization.utils import fp8_utils, int8_utils
+from vllm.platforms import current_platform
+
+
+@contextmanager
+def _triton_mode():
+    """Temporarily force the Triton fallback path"""
+    with patch("vllm.platforms.current_platform.is_cuda", return_value=False):
+        yield
+
+
+def _time_cuda(
+    fn: Callable[[], tuple[torch.Tensor, torch.Tensor]],
+    warmup_iters: int,
+    bench_iters: int,
+) -> float:
+    # warmup
+    for _ in range(warmup_iters):
+        fn()
+    torch.cuda.synchronize()
+
+    start = torch.cuda.Event(enable_timing=True)
+    end = torch.cuda.Event(enable_timing=True)
+
+    start.record()
+    for _ in range(bench_iters):
+        fn()
+    end.record()
+    torch.cuda.synchronize()
+
+    return start.elapsed_time(end) / bench_iters  # ms/iter
+
+
+def _run_single(
+    shape: tuple[int, int],
+    group_size: int,
+    dtype: str,
+    *,
+    column_major: bool = False,
+    scale_ue8m0: bool = False,
+    warmup_iters: int,
+    bench_iters: int,
+) -> None:
+    num_tokens, hidden_dim = shape
+
+    device = torch.device("cuda")
+    torch.manual_seed(42)
+    x = torch.randn(num_tokens, hidden_dim, device=device, dtype=torch.bfloat16) * 8
+
+    if dtype == "fp8":
+
+        def cuda_impl():
+            return fp8_utils.per_token_group_quant_fp8(
+                x,
+                group_size,
+                column_major_scales=column_major,
+                use_ue8m0=scale_ue8m0,
+            )
+
+        def triton_impl():
+            with _triton_mode():
+                return fp8_utils.per_token_group_quant_fp8(
+                    x,
+                    group_size,
+                    column_major_scales=column_major,
+                    use_ue8m0=scale_ue8m0,
+                )
+    elif dtype == "int8":
+
+        def cuda_impl():
+            return int8_utils.per_token_group_quant_int8(x, group_size)
+
+        def triton_impl():
+            with _triton_mode():
+                return int8_utils.per_token_group_quant_int8(x, group_size)
+    else:
+        raise ValueError("dtype must be 'fp8' or 'int8'")
+
+    cuda_ms = _time_cuda(cuda_impl, warmup_iters, bench_iters)
+    triton_ms = _time_cuda(triton_impl, warmup_iters, bench_iters)
+
+    speedup = triton_ms / cuda_ms if cuda_ms else math.inf
+
+    cfg_desc = (
+        f"shape={shape}  gs={group_size:<3}  col_major={column_major:<5}  "
+        f"ue8m0={scale_ue8m0:<5}  dtype={dtype}"
+    )
+    print(
+        f"{cfg_desc:55} | CUDA {cuda_ms:7.3f} ms  | Triton {triton_ms:7.3f} ms  | "
+        f"speed-up ×{speedup:5.2f}"
+    )
+
+
+def parse_args():
+    parser = argparse.ArgumentParser()
+    parser.add_argument("--warmup-iters", type=int, default=10)
+    parser.add_argument("--bench-iters", type=int, default=100)
+    parser.add_argument("--dtype", choices=["fp8", "int8", "both"], default="both")
+    return parser.parse_args()
+
+
+if __name__ == "__main__":
+    if not current_platform.is_cuda():
+        raise RuntimeError("CUDA device is required to run this benchmark.")
+
+    args = parse_args()
+    warmup_iters, bench_iters = args.warmup_iters, args.bench_iters
+
+    shapes = [(32, 128), (64, 256), (16, 512)]
+    group_sizes = [64, 128]
+
+    dtypes = ["fp8", "int8"] if args.dtype == "both" else [args.dtype]
+
+    header = (
+        "Configuration".ljust(55)
+        + " | "
+        + "CUDA (ms)".center(12)
+        + " | "
+        + "Triton (ms)".center(13)
+        + " | "
+        + "Speed-up"
+    )
+    print(header)
+    print("-" * len(header))
+
+    for dtype in dtypes:
+        for shape in shapes:
+            for gs in group_sizes:
+                if dtype == "fp8":
+                    for col_major in (False, True):
+                        for ue8m0 in (False, True):
+                            _run_single(
+                                shape,
+                                gs,
+                                dtype,
+                                column_major=col_major,
+                                scale_ue8m0=ue8m0,
+                                warmup_iters=warmup_iters,
+                                bench_iters=bench_iters,
+                            )
+                else:  # INT8 has no col-major / ue8m0 switches
+                    _run_single(
+                        shape,
+                        gs,
+                        dtype,
+                        warmup_iters=warmup_iters,
+                        bench_iters=bench_iters,
+                    )

benchmarks/kernels/benchmark_reshape_and_cache_flash.py

@@ -0,0 +1,156 @@
+# SPDX-License-Identifier: Apache-2.0
+# SPDX-FileCopyrightText: Copyright contributors to the vLLM project
+from __future__ import annotations
+
+import random
+import time
+
+import torch
+from tabulate import tabulate
+
+from vllm import _custom_ops as ops
+from vllm.logger import init_logger
+from vllm.platforms import current_platform
+from vllm.utils import (
+    STR_DTYPE_TO_TORCH_DTYPE,
+    FlexibleArgumentParser,
+    create_kv_caches_with_random_flash,
+)
+
+logger = init_logger(__name__)
+
+
+@torch.inference_mode()
+def run_benchmark(
+    num_tokens: int,
+    num_heads: int,
+    head_size: int,
+    block_size: int,
+    num_blocks: int,
+    dtype: torch.dtype,
+    kv_cache_dtype: str,
+    kv_cache_layout: str,
+    num_iters: int,
+    device: str = "cuda",
+) -> float:
+    """Return latency (seconds) for given num_tokens."""
+
+    if kv_cache_dtype == "fp8" and head_size % 16:
+        raise ValueError("fp8 kv-cache requires head_size to be a multiple of 16.")
+
+    current_platform.seed_everything(42)
+    torch.set_default_device(device)
+
+    # create random key / value tensors [T, H, D].
+    key = torch.randn(num_tokens, num_heads, head_size, dtype=dtype, device=device)
+    value = torch.randn_like(key)
+
+    # prepare the slot mapping.
+    # each token is assigned a unique slot in the KV-cache.
+    num_slots = block_size * num_blocks
+    if num_tokens > num_slots:
+        raise ValueError("num_tokens cannot exceed the total number of cache slots")
+    slot_mapping_lst = random.sample(range(num_slots), num_tokens)
+    slot_mapping = torch.tensor(slot_mapping_lst, dtype=torch.long, device=device)
+
+    key_caches, value_caches = create_kv_caches_with_random_flash(
+        num_blocks,
+        block_size,
+        1,  # num_layers
+        num_heads,
+        head_size,
+        kv_cache_dtype,
+        dtype,
+        device=device,
+        cache_layout=kv_cache_layout,
+    )
+    key_cache, value_cache = key_caches[0], value_caches[0]
+
+    # compute per-kernel scaling factors for fp8 conversion (if used).
+    k_scale = (key.amax() / 64.0).to(torch.float32)
+    v_scale = (value.amax() / 64.0).to(torch.float32)
+
+    def run_cuda_benchmark(n_iters: int) -> float:
+        nonlocal key, value, key_cache, value_cache, slot_mapping
+        torch.cuda.synchronize()
+        start = time.perf_counter()
+        for _ in range(n_iters):
+            ops.reshape_and_cache_flash(
+                key,
+                value,
+                key_cache,
+                value_cache,
+                slot_mapping,
+                kv_cache_dtype,
+                k_scale,
+                v_scale,
+            )
+        torch.cuda.synchronize()
+        end = time.perf_counter()
+        return (end - start) / n_iters
+
+    # warm-up
+    run_cuda_benchmark(3)
+
+    lat = run_cuda_benchmark(num_iters)
+
+    # free tensors to mitigate OOM when sweeping
+    del key, value, key_cache, value_cache, slot_mapping
+    torch.cuda.empty_cache()
+
+    return lat
+
+
+def main(args):
+    rows = []
+    for layout in ["NHD", "HND"]:
+        for exp in range(1, 17):
+            n_tok = 2**exp
+            lat = run_benchmark(
+                num_tokens=n_tok,
+                num_heads=args.num_heads,
+                head_size=args.head_size,
+                block_size=args.block_size,
+                num_blocks=args.num_blocks,
+                dtype=STR_DTYPE_TO_TORCH_DTYPE[args.dtype],
+                kv_cache_dtype=args.kv_cache_dtype,
+                kv_cache_layout=layout,
+                num_iters=args.iters,
+                device="cuda",
+            )
+            rows.append([n_tok, layout, f"{lat * 1e6:.3f}"])
+
+    print(tabulate(rows, headers=["num_tokens", "layout", "latency (µs)"]))
+
+
+if __name__ == "__main__":
+    parser = FlexibleArgumentParser()
+
+    parser.add_argument("--num-heads", type=int, default=128)
+    parser.add_argument(
+        "--head-size",
+        type=int,
+        choices=[64, 80, 96, 112, 120, 128, 192, 256],
+        default=128,
+    )
+    parser.add_argument("--block-size", type=int, choices=[16, 32], default=16)
+    parser.add_argument("--num-blocks", type=int, default=128 * 512)
+
+    parser.add_argument(
+        "--dtype",
+        type=str,
+        choices=["half", "bfloat16", "float"],
+        default="bfloat16",
+    )
+
+    parser.add_argument(
+        "--kv-cache-dtype",
+        type=str,
+        choices=["auto", "fp8"],
+        default="auto",
+    )
+
+    parser.add_argument("--iters", type=int, default=100)
+    args = parser.parse_args()
+
+    main(args)

benchmarks/kernels/benchmark_trtllm_decode_attention.py

@@ -0,0 +1,253 @@
+# SPDX-License-Identifier: Apache-2.0
+# SPDX-FileCopyrightText: Copyright contributors to the vLLM project
+
+import csv
+import os
+import random
+from datetime import datetime
+
+import flashinfer
+import torch
+
+FLOAT32_BYTES = torch.finfo(torch.float).bits // 8
+
+# KV Cache Layout for TRT-LLM
+# kv_cache_shape = (num_blocks, 2, num_kv_heads, page_size, head_dim)
+
+
+def to_float8(x, dtype=torch.float8_e4m3fn):
+    finfo = torch.finfo(dtype)
+    min_val, max_val = x.aminmax()
+    amax = torch.maximum(min_val.abs(), max_val.abs()).clamp(min=1e-12)
+    scale = finfo.max / amax * 0.1
+    x_scl_sat = (x * scale).clamp(min=finfo.min, max=finfo.max)
+    return x_scl_sat.to(dtype), scale.float().reciprocal()
+
+
+@torch.no_grad()
+def benchmark_decode(
+    num_seqs,
+    max_seq_len,
+    page_size=16,
+    dtype=torch.bfloat16,
+    kv_layout="HND",
+    num_kv_heads=8,
+    kv_cache_dtype="auto",
+    head_dim=128,
+    warmup=10,
+    trials=20,
+):
+    torch.set_default_device("cuda")
+    device = "cuda"
+    torch.manual_seed(0)
+
+    HEAD_GRP_SIZE = 8
+    MAX_SEQ_LEN = max_seq_len
+
+    # large number to reduce kv_cache reuse
+    NUM_BLOCKS = int(256000 / page_size)
+
+    workspace_buffer = torch.empty(1024 * 1024 * 1024, dtype=torch.int8, device=device)
+
+    # For decode, batch_size is num_decode_token
+    num_qo_heads = num_kv_heads * HEAD_GRP_SIZE
+    sm_scale = float(1.0 / (head_dim**0.5))
+    q = torch.randn(num_seqs, num_qo_heads, head_dim, device=device, dtype=dtype)
+    kv_lens = [random.randint(1, MAX_SEQ_LEN) for _ in range(num_seqs)]
+
+    max_kv_len = max(kv_lens)
+    kv_lens_tensor = torch.tensor(kv_lens, dtype=torch.int, device=device)
+    max_num_blocks_per_seq = (max_kv_len + page_size - 1) // page_size
+
+    block_tables = torch.randint(
+        0, NUM_BLOCKS, (num_seqs, max_num_blocks_per_seq), dtype=torch.int32
+    )
+
+    kv_cache_shape = (NUM_BLOCKS, 2, num_kv_heads, page_size, head_dim)
+    kv_cache = torch.randn(size=kv_cache_shape, device=device, dtype=dtype)
+    k_scale = v_scale = 1.0
+
+    if kv_cache_dtype.startswith("fp8"):
+        kv_cache, _ = to_float8(kv_cache)
+
+    output_trtllm = torch.empty(q.shape, dtype=dtype)
+
+    # Benchmark TRT decode
+    def trt_decode():
+        return flashinfer.decode.trtllm_batch_decode_with_kv_cache(
+            q,
+            kv_cache,
+            workspace_buffer,
+            block_tables,
+            kv_lens_tensor,
+            max_kv_len,
+            bmm1_scale=k_scale * sm_scale,
+            bmm2_scale=v_scale,
+            out=output_trtllm,
+        )
+
+    def time_fn(fn, warmup=10, trials=20):
+        torch.cuda.synchronize()
+        start = torch.cuda.Event(enable_timing=True)
+        end = torch.cuda.Event(enable_timing=True)
+        times = []
+        for i in range(warmup):
+            fn()
+        for i in range(trials):
+            start.record()
+            fn()
+            end.record()
+            torch.cuda.synchronize()
+            times.append(start.elapsed_time(end))  # ms
+        return sum(times) / len(times), torch.std(torch.tensor(times))
+
+    # TRT Decode
+    trt_mean, trt_std = time_fn(trt_decode)
+
+    kv_indptr = [0]
+    kv_indices = []
+    kv_last_page_lens = []
+    for i in range(num_seqs):
+        seq_len = kv_lens[i]
+        assert seq_len > 0
+        num_blocks = (seq_len + page_size - 1) // page_size
+        kv_indices.extend(block_tables[i, :num_blocks])
+        kv_indptr.append(kv_indptr[-1] + num_blocks)
+        kv_last_page_len = seq_len % page_size
+        if kv_last_page_len == 0:
+            kv_last_page_len = page_size
+        kv_last_page_lens.append(kv_last_page_len)
+
+    kv_indptr = torch.tensor(kv_indptr, dtype=torch.int32)
+    kv_indices = torch.tensor(kv_indices, dtype=torch.int32)
+    kv_last_page_lens = torch.tensor(kv_last_page_lens, dtype=torch.int32)
+
+    output_baseline = torch.empty(q.shape, dtype=dtype)
+
+    wrapper = flashinfer.BatchDecodeWithPagedKVCacheWrapper(
+        workspace_buffer,
+        kv_layout,
+        use_tensor_cores=((num_qo_heads // num_kv_heads) > 4),
+    )
+
+    wrapper.plan(
+        kv_indptr,
+        kv_indices,
+        kv_last_page_lens,
+        num_qo_heads,
+        num_kv_heads,
+        head_dim,
+        page_size,
+        "NONE",
+        q_data_type=dtype,
+        kv_data_type=torch.float8_e4m3fn if kv_cache_dtype.startswith("fp8") else dtype,
+    )
+
+    def baseline_decode():
+        return wrapper.run(q, kv_cache, sm_scale, k_scale, v_scale, output_baseline)
+
+    baseline_mean, baseline_std = time_fn(baseline_decode)
+
+    # Calculate percentage speedup (positive means TRT is faster)
+    speedup_percent = (baseline_mean - trt_mean) / baseline_mean
+
+    print(
+        f"\t{num_seqs}\t{max_seq_len}\t{trt_mean:.3f}\t{trt_std.item():.3f}"
+        f"\t{baseline_mean:.3f}\t{baseline_std.item():.3f}\t{speedup_percent:.3f}"
+    )
+
+    # Return results for CSV writing
+    return {
+        "num_seqs": num_seqs,
+        "trt_mean": trt_mean,
+        "trt_std": trt_std.item(),
+        "baseline_mean": baseline_mean,
+        "baseline_std": baseline_std.item(),
+        "speedup_percent": speedup_percent,
+        "q_dtype": str(dtype),
+        "kv_cache_dtype": kv_cache_dtype,
+        "page_size": page_size,
+        "num_kv_heads": num_kv_heads,
+        "head_dim": head_dim,
+        "max_seq_len": max_seq_len,
+    }
+
+
+def write_results_to_csv(results, filename=None):
+    """Write benchmark results to CSV file."""
+    if filename is None:
+        timestamp = datetime.now().strftime("%Y%m%d_%H%M%S")
+        filename = f"flashinfer_trtllm_benchmark_{timestamp}.csv"
+
+    fieldnames = [
+        "num_seqs",
+        "trt_mean",
+        "trt_std",
+        "baseline_mean",
+        "baseline_std",
+        "speedup_percent",
+        "q_dtype",
+        "kv_cache_dtype",
+        "page_size",
+        "num_kv_heads",
+        "head_dim",
+        "max_seq_len",
+    ]
+
+    file_exists = os.path.exists(filename)
+
+    with open(filename, "a", newline="") as csvfile:
+        writer = csv.DictWriter(csvfile, fieldnames=fieldnames)
+
+        if not file_exists:
+            writer.writeheader()
+
+        for result in results:
+            writer.writerow(result)
+
+    print(f"Results written to {filename}")
+
+
+if __name__ == "__main__":
+    num_seqs = [1, 4, 8, 16, 32, 64, 128, 256]
+    max_seq_lens = [1024, 2048, 4096, 8192, 16384, 32768, 65536, 131072]
+    all_results = []
+
+    print(
+        "Running benchmark for q_dtype = bfloat16, kv_cache_dtype: bfloat16, "
+        "output_dtype: bfloat16"
+    )
+    print(
+        "\tnum_seqs\tmax_seq_len\ttrt_mean\ttrt_std\tbaseline_mean\t"
+        "baseline_std\tspeedup_percent"
+    )
+    for max_seq_len in max_seq_lens:
+        for bs in num_seqs:
+            result = benchmark_decode(
+                bs,
+                max_seq_len,
+                dtype=torch.bfloat16,
+                kv_cache_dtype="auto",
+            )
+            all_results.append(result)
+
+    print(
+        "Running benchmark for q_dtype = bfloat16, kv_cache_dtype: fp8, "
+        "output_dtype: bfloat16"
+    )
+    print(
+        "\tnum_seqs\tmax_seq_len\ttrt_mean\ttrt_std\tbaseline_mean\t"
+        "baseline_std\tspeedup_percent"
+    )
+    for max_seq_len in max_seq_lens:
+        for bs in num_seqs:
+            result = benchmark_decode(
+                bs,
+                max_seq_len,
+                dtype=torch.bfloat16,
+                kv_cache_dtype="fp8",
+            )
+            all_results.append(result)
+
+    # Write all results to CSV
+    write_results_to_csv(all_results)

benchmarks/kernels/benchmark_trtllm_prefill_attention.py

@@ -0,0 +1,250 @@
+# SPDX-License-Identifier: Apache-2.0
+# SPDX-FileCopyrightText: Copyright contributors to the vLLM project
+
+import csv
+import os
+import random
+from datetime import datetime
+
+import flashinfer
+import torch
+
+FLOAT32_BYTES = torch.finfo(torch.float).bits // 8
+
+# KV Cache Layout for TRT-LLM
+# kv_cache_shape = (num_blocks, 2, num_kv_heads, page_size, head_dim)
+
+
+def to_float8(x, dtype=torch.float8_e4m3fn):
+    finfo = torch.finfo(dtype)
+    min_val, max_val = x.aminmax()
+    amax = torch.maximum(min_val.abs(), max_val.abs()).clamp(min=1e-12)
+    scale = finfo.max / amax * 0.1
+    x_scl_sat = (x * scale).clamp(min=finfo.min, max=finfo.max)
+    return x_scl_sat.to(dtype), scale.float().reciprocal()
+
+
+@torch.no_grad()
+def benchmark_prefill(
+    num_seqs,
+    max_seq_len,
+    page_size=16,
+    dtype=torch.bfloat16,
+    kv_layout="HND",
+    num_kv_heads=8,
+    kv_cache_dtype="auto",
+    head_dim=128,
+    warmup=10,
+    trials=20,
+):
+    torch.set_default_device("cuda")
+    torch.manual_seed(0)
+
+    HEAD_GRP_SIZE = 8
+    MAX_SEQ_LEN = max_seq_len
+
+    # large number to reduce kv_cache reuse
+    NUM_BLOCKS = int(256000 / page_size)
+
+    workspace_buffer = torch.empty(1024 * 1024 * 1024, dtype=torch.int8)
+
+    num_qo_heads = num_kv_heads * HEAD_GRP_SIZE
+    sm_scale = float(1.0 / (head_dim**0.5))
+
+    q_lens = [random.randint(1, MAX_SEQ_LEN) for _ in range(num_seqs)]
+    q_lens[-1] = MAX_SEQ_LEN
+    max_q_len = max(q_lens)
+    q_indptr = torch.cat(
+        [
+            torch.tensor([0], dtype=torch.int32),
+            torch.cumsum(
+                torch.tensor(q_lens, dtype=torch.int32), dim=0, dtype=torch.int32
+            ),
+        ]
+    )
+    q = torch.randn(sum(q_lens), num_qo_heads, head_dim, dtype=dtype)
+
+    kv_lens = [random.randint(0, MAX_SEQ_LEN) for _ in range(num_seqs)]
+    kv_lens[-1] = MAX_SEQ_LEN
+
+    seq_lens = [q_len + kv_len for q_len, kv_len in zip(q_lens, kv_lens)]
+    max_seq_len = max(seq_lens)
+    seq_lens_tensor = torch.tensor(seq_lens, dtype=torch.int32)
+
+    max_num_blocks_per_seq = (max_seq_len + page_size - 1) // page_size
+    block_tables = torch.randint(
+        0, NUM_BLOCKS, (num_seqs, max_num_blocks_per_seq), dtype=torch.int32
+    )
+
+    kv_cache_shape = (NUM_BLOCKS, 2, num_kv_heads, page_size, head_dim)
+    kv_cache = torch.randn(size=kv_cache_shape, dtype=dtype)
+    k_scale = v_scale = 1.0
+
+    if kv_cache_dtype.startswith("fp8"):
+        kv_cache, _ = to_float8(kv_cache)
+
+    output_trtllm = torch.empty(q.shape, dtype=dtype)
+
+    kv_indptr = [0]
+    kv_indices = []
+    kv_last_page_lens = []
+    for i in range(num_seqs):
+        seq_len = seq_lens[i]
+        assert seq_len > 0
+        num_blocks = (seq_len + page_size - 1) // page_size
+        kv_indices.extend(block_tables[i, :num_blocks])
+        kv_indptr.append(kv_indptr[-1] + num_blocks)
+        kv_last_page_len = seq_len % page_size
+        if kv_last_page_len == 0:
+            kv_last_page_len = page_size
+        kv_last_page_lens.append(kv_last_page_len)
+
+    kv_indptr = torch.tensor(kv_indptr, dtype=torch.int32)
+    kv_indices = torch.tensor(kv_indices, dtype=torch.int32)
+    kv_last_page_lens = torch.tensor(kv_last_page_lens, dtype=torch.int32)
+
+    output_baseline = torch.empty(q.shape, dtype=dtype)
+
+    wrapper = flashinfer.BatchPrefillWithPagedKVCacheWrapper(
+        workspace_buffer, kv_layout
+    )
+    wrapper.plan(
+        q_indptr,
+        kv_indptr,
+        kv_indices,
+        kv_last_page_lens,
+        num_qo_heads,
+        num_kv_heads,
+        head_dim,
+        page_size,
+        causal=True,
+        sm_scale=sm_scale,
+        q_data_type=dtype,
+        kv_data_type=kv_cache.dtype,
+    )
+
+    def time_fn(fn, warmup=10, trials=20):
+        torch.cuda.synchronize()
+        start = torch.cuda.Event(enable_timing=True)
+        end = torch.cuda.Event(enable_timing=True)
+        times = []
+        for i in range(warmup):
+            fn()
+        for i in range(trials):
+            start.record()
+            fn()
+            end.record()
+            torch.cuda.synchronize()
+            times.append(start.elapsed_time(end))  # ms
+        return sum(times) / len(times), torch.std(torch.tensor(times))
+
+    def baseline_prefill():
+        return wrapper.run(
+            q, kv_cache, k_scale=k_scale, v_scale=v_scale, out=output_baseline
+        )
+
+    def trt_prefill():
+        return flashinfer.prefill.trtllm_batch_context_with_kv_cache(
+            query=q,
+            kv_cache=kv_cache,
+            workspace_buffer=workspace_buffer,
+            block_tables=block_tables,
+            seq_lens=seq_lens_tensor,
+            max_q_len=max_q_len,
+            max_kv_len=max_seq_len,
+            bmm1_scale=k_scale * sm_scale,
+            bmm2_scale=v_scale,
+            batch_size=num_seqs,
+            cum_seq_lens_q=q_indptr,
+            cum_seq_lens_kv=kv_indptr,
+            out=output_trtllm,
+        )
+
+    trt_mean, trt_std = time_fn(trt_prefill)
+    baseline_mean, baseline_std = time_fn(baseline_prefill)
+
+    # Calculate percentage speedup (positive means TRT is faster)
+    speedup_percent = (baseline_mean - trt_mean) / baseline_mean
+
+    print(
+        f"\t{num_seqs}\t{max_seq_len}\t{trt_mean:.5f}\t{trt_std.item():.5f}"
+        f"\t{baseline_mean:.5f}\t{baseline_std.item():.5f}\t{speedup_percent:.5f}"
+    )
+
+    # Return results for CSV writing
+    return {
+        "num_seqs": num_seqs,
+        "trt_mean": trt_mean,
+        "trt_std": trt_std.item(),
+        "baseline_mean": baseline_mean,
+        "baseline_std": baseline_std.item(),
+        "speedup_percent": speedup_percent,
+        "q_dtype": str(dtype),
+        "kv_cache_dtype": kv_cache_dtype,
+        "page_size": page_size,
+        "num_kv_heads": num_kv_heads,
+        "head_dim": head_dim,
+        "max_seq_len": max_seq_len,
+    }
+
+
+def write_results_to_csv(results, filename=None):
+    """Write benchmark results to CSV file."""
+    if filename is None:
+        timestamp = datetime.now().strftime("%Y%m%d_%H%M%S")
+        filename = f"flashinfer_trtllm_benchmark_{timestamp}.csv"
+
+    fieldnames = [
+        "num_seqs",
+        "trt_mean",
+        "trt_std",
+        "baseline_mean",
+        "baseline_std",
+        "speedup_percent",
+        "q_dtype",
+        "kv_cache_dtype",
+        "page_size",
+        "num_kv_heads",
+        "head_dim",
+        "max_seq_len",
+    ]
+
+    file_exists = os.path.exists(filename)
+
+    with open(filename, "a", newline="") as csvfile:
+        writer = csv.DictWriter(csvfile, fieldnames=fieldnames)
+
+        if not file_exists:
+            writer.writeheader()
+
+        for result in results:
+            writer.writerow(result)
+
+    print(f"Results written to {filename}")
+
+
+if __name__ == "__main__":
+    num_seqs = [1, 4, 8, 16, 32, 64, 128, 256]
+    max_seq_lens = [1024, 2048, 4096, 8192, 16384, 32768, 65536, 131072]
+    all_results = []
+
+    print(
+        "Running benchmark for q_dtype = bfloat16, kv_cache_dtype: bfloat16, "
+        "output_dtype: bfloat16"
+    )
+    print(
+        "\tnum_seqs\tmax_seq_len\ttrt_mean\ttrt_std\tbaseline_mean\t"
+        "baseline_std\tspeedup_percent"
+    )
+    for max_seq_len in max_seq_lens:
+        for bs in num_seqs:
+            result = benchmark_prefill(
+                bs,
+                max_seq_len,
+                dtype=torch.bfloat16,
+                kv_cache_dtype="auto",
+            )
+            all_results.append(result)
+
+    # Write all results to CSV
+    write_results_to_csv(all_results)

benchmarks/kernels/deepgemm/README.md

@@ -8,7 +8,7 @@ Currently this just includes dense GEMMs and only works on Hopper GPUs.
 
 You need to install vLLM in your usual fashion, then install DeepGEMM from source in its own directory:
 
-```
+```bash
 git clone --recursive https://github.com/deepseek-ai/DeepGEMM
 cd DeepGEMM
 python setup.py install
@@ -17,7 +17,7 @@ uv pip install -e .
 
 ## Usage
 
-```
+```console
 python benchmark_fp8_block_dense_gemm.py
 INFO 02-26 21:55:13 [__init__.py:207] Automatically detected platform cuda.
 ===== STARTING FP8 GEMM BENCHMARK =====

benchmarks/kernels/deepgemm/benchmark_fp8_block_dense_gemm.py

@@ -4,49 +4,16 @@
 # ruff: noqa: E501
 import time
 
-# Import DeepGEMM functions
-import deep_gemm
 import torch
-from deep_gemm import calc_diff, ceil_div, get_col_major_tma_aligned_tensor
 
-# Import vLLM functions
 from vllm import _custom_ops as ops
 from vllm.model_executor.layers.quantization.utils.fp8_utils import (
+    get_col_major_tma_aligned_tensor,
     per_token_group_quant_fp8,
     w8a8_block_fp8_matmul,
 )
 from vllm.triton_utils import triton
-
-
-# Copied from
-# https://github.com/deepseek-ai/DeepGEMM/blob/78cacf70d41d15d688bd493ebc85845f7f2a3d5d/tests/test_core.py#L9
-def per_token_cast_to_fp8(
-        x: torch.Tensor) -> tuple[torch.Tensor, torch.Tensor]:
-    """Convert tensor to FP8 format with per-token scaling."""
-    assert x.dim() == 2 and x.size(1) % 128 == 0
-    m, n = x.shape
-    x_view = x.view(m, -1, 128)
-    x_amax = x_view.abs().float().amax(dim=2).view(m, -1).clamp(1e-4)
-    return (x_view * (448.0 / x_amax.unsqueeze(2))).to(
-        torch.float8_e4m3fn).view(m, n), (x_amax / 448.0).view(m, -1)
-
-
-# Copied from
-# https://github.com/deepseek-ai/DeepGEMM/blob/78cacf70d41d15d688bd493ebc85845f7f2a3d5d/tests/test_core.py#L17
-def per_block_cast_to_fp8(
-        x: torch.Tensor) -> tuple[torch.Tensor, torch.Tensor]:
-    """Convert tensor to FP8 format with per-block scaling."""
-    assert x.dim() == 2
-    m, n = x.shape
-    x_padded = torch.zeros((ceil_div(m, 128) * 128, ceil_div(n, 128) * 128),
-                           dtype=x.dtype,
-                           device=x.device)
-    x_padded[:m, :n] = x
-    x_view = x_padded.view(-1, 128, x_padded.size(1) // 128, 128)
-    x_amax = x_view.abs().float().amax(dim=(1, 3), keepdim=True).clamp(1e-4)
-    x_scaled = (x_view * (448.0 / x_amax)).to(torch.float8_e4m3fn)
-    return x_scaled.view_as(x_padded)[:m, :n].contiguous(), (
-        x_amax / 448.0).view(x_view.size(0), x_view.size(2))
+from vllm.utils.deep_gemm import calc_diff, fp8_gemm_nt, per_block_cast_to_fp8
 
 
 def benchmark_shape(m: int,
@@ -69,14 +36,14 @@ def benchmark_shape(m: int,
 
     # Pre-quantize B for all implementations
     # (weights can be pre-quantized offline)
-    B_deepgemm, B_scale_deepgemm = per_block_cast_to_fp8(B)
-    B_vllm, B_scale_vllm = per_block_cast_to_fp8(B)
+    B_deepgemm, B_scale_deepgemm = per_block_cast_to_fp8(B, [128, 128], use_ue8m0=True)
+    B_vllm, B_scale_vllm = per_block_cast_to_fp8(B, [128, 128], use_ue8m0=True)
 
     # Block size configuration
     block_size = [128, 128]
 
     # Pre-quantize A for all implementations
-    A_deepgemm, A_scale_deepgemm = per_token_cast_to_fp8(A)
+    A_deepgemm, A_scale_deepgemm = per_token_group_quant_fp8(A, block_size[1])
     A_scale_deepgemm = get_col_major_tma_aligned_tensor(A_scale_deepgemm)
     C_deepgemm = torch.empty((m, n), device='cuda', dtype=torch.bfloat16)
     A_vllm, A_scale_vllm = per_token_group_quant_fp8(A, block_size[1])
@@ -85,21 +52,13 @@ def benchmark_shape(m: int,
 
     # === DeepGEMM Implementation ===
     def deepgemm_gemm():
-        # A quantization is inside the loop as it depends on activations
-        # A_deepgemm, A_scale_deepgemm = per_token_cast_to_fp8(A)
-        # A_deepgemm, A_scale_deepgemm = per_token_group_quant_fp8(
-        #     A, block_size[1])
-        # A_scale_aligned = get_col_major_tma_aligned_tensor(A_scale_deepgemm)
-        # C_deepgemm = torch.empty((m, n), device='cuda', dtype=torch.bfloat16)
-        deep_gemm.gemm_fp8_fp8_bf16_nt((A_deepgemm, A_scale_deepgemm),
+        fp8_gemm_nt((A_deepgemm, A_scale_deepgemm),
                                        (B_deepgemm, B_scale_deepgemm),
                                        C_deepgemm)
         return C_deepgemm
 
     # === vLLM Triton Implementation ===
     def vllm_triton_gemm():
-        # A quantization is inside the loop as it depends on activations
-        # A_vllm, A_scale_vllm = per_token_group_quant_fp8(A, block_size[1])
         return w8a8_block_fp8_matmul(A_vllm,
                                      B_vllm,
                                      A_scale_vllm,
@@ -109,9 +68,6 @@ def benchmark_shape(m: int,
 
     # === vLLM CUTLASS Implementation ===
     def vllm_cutlass_gemm():
-        # A quantization is inside the loop as it depends on activations
-        # A_vllm_cutlass, A_scale_vllm_cutlass = per_token_group_quant_fp8(
-        #     A, block_size[1], column_major_scales=True)
         return ops.cutlass_scaled_mm(A_vllm_cutlass,
                                      B_vllm.T,
                                      scale_a=A_scale_vllm_cutlass,

benchmarks/multi_turn/README.md

@@ -0,0 +1,71 @@
+# Benchmark KV Cache Offloading with Multi-Turn Conversations
+
+The requirements (pip) for `benchmark_serving_multi_turn.py` can be found in `requirements.txt`
+
+First start serving your model
+
+```bash
+export MODEL_NAME=/models/meta-llama/Meta-Llama-3.1-8B-Instruct/
+
+vllm serve $MODEL_NAME --disable-log-requests
+```
+
+## Synthetic Multi-Turn Conversations
+
+Download the following text file (used for generation of synthetic conversations)
+
+```bash
+wget https://www.gutenberg.org/ebooks/1184.txt.utf-8
+mv 1184.txt.utf-8 pg1184.txt
+```
+
+The filename `pg1184.txt` is used in `generate_multi_turn.json` (see `"text_files"`).
+
+But you may use other text files if you prefer (using this specific file is not required).
+
+Then run the benchmarking script
+
+```bash
+export MODEL_NAME=/models/meta-llama/Meta-Llama-3.1-8B-Instruct/
+
+python benchmark_serving_multi_turn.py --model $MODEL_NAME --input-file generate_multi_turn.json \
+--num-clients 2 --max-active-conversations 6
+```
+
+You can edit the file `generate_multi_turn.json` to change the conversation parameters (number of turns, etc.).
+
+If successful, you will see the following output
+
+```bash
+----------------------------------------------------------------------------------------------------
+Statistics summary:
+runtime_sec = 215.810
+requests_per_sec = 0.769
+----------------------------------------------------------------------------------------------------
+                   count     mean     std      min      25%      50%      75%      90%      99%      max
+ttft_ms            166.0    78.22   67.63    45.91    59.94    62.26    64.43    69.66   353.18   567.54
+tpot_ms            166.0    25.37    0.57    24.40    25.07    25.31    25.50    25.84    27.50    28.05
+latency_ms         166.0  2591.07  326.90  1998.53  2341.62  2573.01  2860.10  3003.50  3268.46  3862.94
+input_num_turns    166.0     7.43    4.57     1.00     3.00     7.00    11.00    13.00    17.00    17.00
+input_num_tokens   166.0  2006.20  893.56   522.00  1247.75  2019.00  2718.00  3233.00  3736.45  3899.00
+output_num_tokens  166.0   100.01   11.80    80.00    91.00    99.00   109.75   116.00   120.00   120.00
+output_num_chunks  166.0    99.01   11.80    79.00    90.00    98.00   108.75   115.00   119.00   119.00
+----------------------------------------------------------------------------------------------------
+```
+
+## ShareGPT Conversations
+
+To run with the ShareGPT data, download the following ShareGPT dataset:
+`https://huggingface.co/datasets/philschmid/sharegpt-raw/blob/main/sharegpt_20230401_clean_lang_split.json`
+
+Use the `convert_sharegpt_to_openai.py` script to convert the dataset to a format supported by `benchmark_serving_multi_turn.py`
+
+```bash
+python convert_sharegpt_to_openai.py sharegpt_20230401_clean_lang_split.json sharegpt_conv_128.json --seed=99 --max-items=128
+```
+
+The script will convert the ShareGPT dataset to a dataset with the standard user/assistant roles.
+
+The flag `--max-items=128` is used to sample 128 conversations from the original dataset (change as needed).
+
+Use the output JSON file `sharegpt_conv_128.json` as the `--input-file` for `benchmark_serving_multi_turn.py`.

benchmarks/multi_turn/bench_dataset.py

@@ -0,0 +1,493 @@
+# SPDX-License-Identifier: Apache-2.0
+# SPDX-FileCopyrightText: Copyright contributors to the vLLM project
+from abc import ABC, abstractmethod
+from statistics import mean
+from typing import Any, NamedTuple, Optional, Union
+
+import numpy as np  # type: ignore
+import pandas as pd  # type: ignore
+from bench_utils import (
+    TEXT_SEPARATOR,
+    Color,
+    logger,
+)
+from transformers import AutoTokenizer  # type: ignore
+
+# Conversation ID is a string (e.g: "UzTK34D")
+ConvId = str
+
+# A list of dicts (dicts with keys "id" and "messages")
+ShareGptConversations = list[dict[str, Any]]
+
+# A list of dicts (dicts with keys "role" and "content")
+MessagesList = list[dict[str, str]]
+
+# Map conversation ID to conversation messages
+ConversationsMap = list[ConvId, MessagesList]
+
+
+class Distribution(ABC):
+    @abstractmethod
+    def sample(self, size: int = 1) -> np.ndarray:
+        pass
+
+
+class UniformDistribution(Distribution):
+    def __init__(
+        self,
+        min_val: Union[int, float],
+        max_val: Union[int, float],
+        is_integer: bool = True,
+    ) -> None:
+        self.min_val = min_val
+        self.max_val = max_val
+        self.is_integer = is_integer
+
+    def sample(self, size: int = 1) -> np.ndarray:
+        if self.is_integer:
+            return np.random.randint(
+                int(self.min_val), int(self.max_val + 1), size=size
+            )
+        else:
+            return np.random.uniform(self.min_val, self.max_val, size=size)
+
+    def __repr__(self) -> str:
+        return f"UniformDistribution[{self.min_val}, {self.max_val}]"
+
+
+class ConstantDistribution(Distribution):
+    def __init__(self, value: Union[int, float]) -> None:
+        self.value = value
+        self.max_val = value
+
+    def sample(self, size: int = 1) -> np.ndarray:
+        return np.full(shape=size, fill_value=self.value)
+
+    def __repr__(self) -> str:
+        return f"Constant[{self.value}]"
+
+
+class ZipfDistribution(Distribution):
+    def __init__(self, alpha: float, max_val: Optional[int] = None) -> None:
+        self.alpha = alpha
+        self.max_val = max_val
+
+    def sample(self, size: int = 1) -> np.ndarray:
+        samples = np.random.zipf(self.alpha, size=size)
+        if self.max_val:
+            samples = np.minimum(samples, self.max_val)
+        return samples
+
+    def __repr__(self) -> str:
+        return f"ZipfDistribution[{self.alpha}]"
+
+
+class PoissonDistribution(Distribution):
+    def __init__(self, alpha: float, max_val: Optional[int] = None) -> None:
+        self.alpha = alpha
+        self.max_val = max_val
+
+    def sample(self, size: int = 1) -> np.ndarray:
+        samples = np.random.poisson(self.alpha, size=size)
+        if self.max_val:
+            samples = np.minimum(samples, self.max_val)
+        return samples
+
+    def __repr__(self) -> str:
+        return f"PoissonDistribution[{self.alpha}]"
+
+
+class LognormalDistribution(Distribution):
+    def __init__(
+        self, mean: float, sigma: float, max_val: Optional[int] = None
+    ) -> None:
+        self.mean = mean
+        self.sigma = sigma
+        self.max_val = max_val
+
+    def sample(self, size: int = 1) -> np.ndarray:
+        samples = np.random.lognormal(mean=self.mean, sigma=self.sigma, size=size)
+        if self.max_val:
+            samples = np.minimum(samples, self.max_val)
+
+        return np.round(samples).astype(int)
+
+    def __repr__(self) -> str:
+        return f"LognormalDistribution[{self.mean}, {self.sigma}]"
+
+
+class GenConvArgs(NamedTuple):
+    num_conversations: int
+    text_files: list[str]
+    input_num_turns: Distribution
+    input_common_prefix_num_tokens: Distribution
+    input_prefix_num_tokens: Distribution
+    input_num_tokens: Distribution
+    output_num_tokens: Distribution
+    print_stats: bool
+
+
+def verify_field_exists(
+    conf: dict, field_name: str, section: str, subsection: str
+) -> None:
+    if field_name not in conf:
+        raise ValueError(
+            f"Missing field '{field_name}' in {section=} and {subsection=}"
+        )
+
+
+def get_random_distribution(
+    conf: dict, section: str, subsection: str, optional: bool = False
+) -> Distribution:
+    # section can be "prompt_input" or "prompt_output" (both required)
+    conf = conf[section]
+
+    if optional and subsection not in conf:
+        # Optional subsection, if not found assume the value is always 0
+        return ConstantDistribution(0)
+
+    # subsection can be "num_turns", "num_tokens" or "prefix_num_tokens"
+    if subsection not in conf:
+        raise ValueError(f"Missing subsection {subsection} in section {section}")
+
+    conf = conf[subsection]
+
+    distribution = conf.get("distribution")
+    if distribution is None:
+        raise ValueError(
+            f"Missing field 'distribution' in {section=} and {subsection=}"
+        )
+
+    if distribution == "constant":
+        verify_field_exists(conf, "value", section, subsection)
+        return ConstantDistribution(conf["value"])
+
+    elif distribution == "zipf":
+        verify_field_exists(conf, "alpha", section, subsection)
+        max_val = conf.get("max", None)
+        return ZipfDistribution(conf["alpha"], max_val=max_val)
+
+    elif distribution == "poisson":
+        verify_field_exists(conf, "alpha", section, subsection)
+        max_val = conf.get("max", None)
+        return PoissonDistribution(conf["alpha"], max_val=max_val)
+
+    elif distribution == "lognormal":
+        verify_field_exists(conf, "mean", section, subsection)
+        verify_field_exists(conf, "sigma", section, subsection)
+        max_val = conf.get("max", None)
+        return LognormalDistribution(conf["mean"], conf["sigma"], max_val=max_val)
+
+    elif distribution == "uniform":
+        verify_field_exists(conf, "min", section, subsection)
+        verify_field_exists(conf, "max", section, subsection)
+
+        min_value = conf["min"]
+        max_value = conf["max"]
+
+        assert min_value > 0
+        assert min_value <= max_value
+
+        is_integer = isinstance(min_value, int) and isinstance(max_value, int)
+        return UniformDistribution(min_value, max_value, is_integer)
+    else:
+        raise ValueError(f"Unknown distribution: {distribution}")
+
+
+def parse_input_json_file(conf: dict) -> GenConvArgs:
+    # Validate the input file
+    assert isinstance(conf, dict)
+    required_fields = [
+        "filetype",
+        "num_conversations",
+        "text_files",
+        "prompt_input",
+        "prompt_output",
+    ]
+    for field in required_fields:
+        assert field in conf, f"Missing field {field} in input {conf}"
+
+    assert conf["filetype"] == "generate_conversations"
+
+    assert conf["num_conversations"] > 0, "num_conversations should be larger than zero"
+
+    text_files = conf["text_files"]
+
+    assert isinstance(text_files, list), "Field 'text_files' should be a list"
+    assert len(text_files) > 0, (
+        "Field 'text_files' should be a list with at least one file"
+    )
+
+    # Parse the parameters for the prompt input/output workload
+    input_num_turns = get_random_distribution(conf, "prompt_input", "num_turns")
+    input_num_tokens = get_random_distribution(conf, "prompt_input", "num_tokens")
+    input_common_prefix_num_tokens = get_random_distribution(
+        conf, "prompt_input", "common_prefix_num_tokens", optional=True
+    )
+    input_prefix_num_tokens = get_random_distribution(
+        conf, "prompt_input", "prefix_num_tokens"
+    )
+    output_num_tokens = get_random_distribution(conf, "prompt_output", "num_tokens")
+
+    print_stats: bool = conf.get("print_stats", False)
+    assert isinstance(print_stats, bool), (
+        "Field 'print_stats' should be either 'true' or 'false'"
+    )
+
+    args = GenConvArgs(
+        num_conversations=conf["num_conversations"],
+        text_files=text_files,
+        input_num_turns=input_num_turns,
+        input_common_prefix_num_tokens=input_common_prefix_num_tokens,
+        input_prefix_num_tokens=input_prefix_num_tokens,
+        input_num_tokens=input_num_tokens,
+        output_num_tokens=output_num_tokens,
+        print_stats=print_stats,
+    )
+    return args
+
+
+def print_conv_stats(conversations: ConversationsMap, tokenizer: AutoTokenizer) -> None:
+    # Collect statistics
+    conv_stats: list[dict[Any, Any]] = []
+    req_stats: list[int] = []
+
+    print("\nCollecting statistics...")
+    for messages in conversations.values():
+        # messages is a list of dicts
+        user_tokens: list[int] = []
+        assistant_tokens: list[int] = []
+        request_tokens: list[int] = []
+
+        req_tokens = 0
+        for m in messages:
+            content = m["content"]
+            num_tokens = len(tokenizer(content).input_ids)
+
+            if m["role"] == "user":
+                user_tokens.append(num_tokens)
+                # New user prompt including all chat history
+                req_tokens += num_tokens
+                request_tokens.append(req_tokens)
+
+            elif m["role"] == "assistant":
+                assistant_tokens.append(num_tokens)
+                # Update assistant answer
+                # (will be part of chat history for the next user prompt)
+                req_tokens += num_tokens
+
+        item_stats = {
+            "conversation_turns": len(messages),
+            "user_tokens": mean(user_tokens),
+            "assistant_tokens": mean(assistant_tokens),
+        }
+
+        conv_stats.append(item_stats)
+        req_stats.extend(request_tokens)
+
+    # Print statistics
+    percentiles = [0.25, 0.5, 0.75, 0.9, 0.99]
+
+    print(TEXT_SEPARATOR)
+    print(f"{Color.YELLOW}Conversations statistics:{Color.RESET}")
+    print(TEXT_SEPARATOR)
+    df = pd.DataFrame(conv_stats)
+    print(df.describe(percentiles=percentiles).transpose())
+    print(TEXT_SEPARATOR)
+    print(f"{Color.YELLOW}Request statistics:{Color.RESET}")
+    print(TEXT_SEPARATOR)
+    df = pd.DataFrame(req_stats, columns=["request_tokens"])
+    print(df.describe(percentiles=percentiles).transpose())
+    print(TEXT_SEPARATOR)
+
+
+def generate_conversations(
+    args: GenConvArgs, tokenizer: AutoTokenizer
+) -> ConversationsMap:
+    # Text for all user prompts
+    # (text from the input text files will be appended to this line)
+    base_prompt_text = "Please rewrite the following text and add more content: "
+    base_prompt_token_count = len(
+        tokenizer.encode(base_prompt_text, add_special_tokens=False)
+    )
+
+    logger.info(f"{Color.PURPLE}Generating conversations...{Color.RESET}")
+    logger.info(args)
+
+    list_of_tokens = []
+
+    for filename in args.text_files:
+        # Load text file that will be used to generate prompts
+        with open(filename) as file:
+            data = file.read()
+            tokens_in_file = tokenizer.encode(data, add_special_tokens=False)
+            list_of_tokens.extend(tokens_in_file)
+
+    conversations: ConversationsMap = {}
+    conv_id = 0
+
+    # Generate number of turns for every conversation
+    turn_count: np.ndarray = args.input_num_turns.sample(args.num_conversations)
+
+    # Turn count should be at least 2 (one user prompt and one assistant answer)
+    turn_count = np.maximum(turn_count, 2)
+
+    # Round up to an even number (every user prompt should have an answer)
+    turn_count = turn_count + (turn_count % 2)
+
+    # Generate number of prefix tokens for every conversation
+    conv_prefix_tokens: np.ndarray = args.input_prefix_num_tokens.sample(
+        args.num_conversations
+    )
+
+    # Used to reduce shared text between conversations
+    # (jump/skip over text sections between conversations)
+    base_offset = 0
+
+    # Common prefix size for all conversations (only 1 sample required)
+    common_prefix_text = ""
+    common_prefix_tokens: int = args.input_common_prefix_num_tokens.sample(1)[0]
+    if common_prefix_tokens > 0:
+        # Using "." at the end to separate sentences
+        common_prefix_text = (
+            tokenizer.decode(list_of_tokens[: common_prefix_tokens - 2]) + "."
+        )
+        base_offset += common_prefix_tokens
+
+    for conv_id in range(args.num_conversations):
+        # Generate a single conversation
+        messages: MessagesList = []
+
+        nturns = turn_count[conv_id]
+
+        # User prompt token count per turn (with lower limit)
+        input_token_count: np.ndarray = args.input_num_tokens.sample(nturns)
+        input_token_count = np.maximum(input_token_count, base_prompt_token_count)
+
+        # Assistant answer token count per turn (with lower limit)
+        output_token_count: np.ndarray = args.output_num_tokens.sample(nturns)
+        output_token_count = np.maximum(output_token_count, 1)
+
+        user_turn = True
+        for turn_id in range(nturns):
+            if user_turn:
+                role = "user"
+                num_tokens = input_token_count[turn_id]
+
+                # Generate the user prompt,
+                # use a unique prefix (the conv_id) for each conversation
+                # (to avoid shared prefix between conversations)
+                content = f"{conv_id} is a nice number... "
+
+                if len(common_prefix_text) > 0 and turn_id == 0:
+                    content = common_prefix_text + content
+
+                # Update the number of tokens left for the content
+                num_tokens -= len(tokenizer.encode(content, add_special_tokens=False))
+
+                if turn_id == 0:
+                    prefix_num_tokens = conv_prefix_tokens[conv_id]
+                    if prefix_num_tokens > 0:
+                        # Add prefix text (context) to the first turn
+                        start_offset = base_offset
+                        end_offset = start_offset + prefix_num_tokens
+                        assert len(list_of_tokens) > end_offset, (
+                            "Not enough input text to generate "
+                            f"{prefix_num_tokens} tokens for the "
+                            f"prefix text ({start_offset=}, {end_offset=})"
+                        )
+
+                        content += f"{conv_id}, " + tokenizer.decode(
+                            list_of_tokens[start_offset:end_offset]
+                        )
+                        base_offset += prefix_num_tokens
+
+                # Add the actual user prompt/question after the prefix text
+                content += base_prompt_text
+                num_tokens -= base_prompt_token_count
+
+                if num_tokens > 0:
+                    # Add text from the input file (to reach the desired token count)
+                    start_offset = base_offset + turn_id * input_token_count.max()
+                    end_offset = start_offset + num_tokens
+                    assert len(list_of_tokens) > end_offset, (
+                        f"Not enough input text to generate {num_tokens} tokens "
+                        f"for the prompt ({start_offset=}, {end_offset=})"
+                    )
+
+                    # Convert tokens back to text
+                    content += tokenizer.decode(list_of_tokens[start_offset:end_offset])
+            else:
+                role = "assistant"
+                # This content will not be used as input to the LLM server
+                # (actual answers will be used instead).
+                # Content is only required to determine the min_tokens/max_tokens
+                # (inputs to the LLM server).
+                num_tokens = output_token_count[turn_id]
+                assert len(list_of_tokens) > num_tokens, (
+                    f"Not enough input text to generate {num_tokens} "
+                    "tokens for assistant content"
+                )
+                content = tokenizer.decode(list_of_tokens[:num_tokens])
+
+            # Append the user/assistant message to the list of messages
+            messages.append({"role": role, "content": content})
+            user_turn = not user_turn
+
+        # Add the new conversation
+        conversations[f"CONV_ID_{conv_id}"] = messages
+
+        # Increase base offset for the next conversation
+        base_offset += nturns
+
+    if args.print_stats:
+        print_conv_stats(conversations, tokenizer)
+
+    return conversations
+
+
+def conversations_list_to_dict(input_list: ShareGptConversations) -> ConversationsMap:
+    conversations: ConversationsMap = {}
+
+    for item in input_list:
+        conv_id: str = item["id"]
+        assert isinstance(conv_id, str)
+
+        assert conv_id not in conversations, (
+            f"Conversation ID {conv_id} found more than once in the input"
+        )
+
+        messages: MessagesList = item["messages"]
+        assert isinstance(messages, list), (
+            f"Conversation messages should be a list (ID: {conv_id})"
+        )
+        assert len(messages) > 0, f"Conversation with no messages (ID: {conv_id})"
+
+        conversations[conv_id] = messages
+
+    logger.info(f"Using {len(conversations)} unique conversations (IDs)")
+    assert len(conversations) == len(input_list)
+
+    # Print statistics about the selected conversations
+    stats: list[dict[str, Any]] = []
+    for conv_data in conversations.values():
+        stats.append({"num_turns": len(conv_data)})
+
+    print(TEXT_SEPARATOR)
+    print(f"{Color.YELLOW}Conversations statistics:{Color.RESET}")
+    print(TEXT_SEPARATOR)
+    percentiles = [0.25, 0.5, 0.75, 0.9, 0.99, 0.999, 0.9999]
+    conv_stats = pd.DataFrame(stats).describe(percentiles=percentiles)
+    print(conv_stats.transpose())
+    print(TEXT_SEPARATOR)
+
+    return conversations
+
+
+def conversations_dict_to_list(input_dict: ConversationsMap) -> ShareGptConversations:
+    output: ShareGptConversations = []
+    for conv_id, conv_data in input_dict.items():
+        new_item = {"id": conv_id, "messages": conv_data}
+        output.append(new_item)
+
+    return output

benchmarks/multi_turn/bench_utils.py

@@ -0,0 +1,28 @@
+# SPDX-License-Identifier: Apache-2.0
+# SPDX-FileCopyrightText: Copyright contributors to the vLLM project
+import logging
+from enum import Enum
+
+
+class Color(Enum):
+    RED = "\033[91m"
+    GREEN = "\033[92m"
+    BLUE = "\033[94m"
+    PURPLE = "\033[95m"
+    CYAN = "\033[96m"
+    YELLOW = "\033[93m"
+    RESET = "\033[0m"
+
+    def __str__(self):
+        return self.value
+
+
+TEXT_SEPARATOR = "-" * 100
+
+# Configure the logger
+logging.basicConfig(
+    level=logging.INFO,
+    format="%(asctime)s [%(levelname)s] - %(message)s",
+    datefmt="%d-%m-%Y %H:%M:%S",
+)
+logger = logging.getLogger(__name__)

benchmarks/multi_turn/convert_sharegpt_to_openai.py

@@ -0,0 +1,354 @@
+# SPDX-License-Identifier: Apache-2.0
+# SPDX-FileCopyrightText: Copyright contributors to the vLLM project
+"""
+Download dataset from:
+https://huggingface.co/datasets/philschmid/sharegpt-raw/blob/main/sharegpt_20230401_clean_lang_split.json
+
+Convert to OpenAI API:
+export INPUT_FILE=sharegpt_20230401_clean_lang_split.json
+python convert_sharegpt_to_openai.py $INPUT_FILE sharegpt_conv_128.json --max-items=128
+"""
+
+import argparse
+import json
+import random
+from statistics import mean
+from typing import Any, Optional
+
+import pandas as pd  # type: ignore
+import tqdm  # type: ignore
+from transformers import AutoTokenizer  # type: ignore
+
+
+def has_non_english_chars(text: str) -> bool:
+    return not text.isascii()
+
+
+def content_is_valid(
+    content: str, min_content_len: Optional[int], max_content_len: Optional[int]
+) -> bool:
+    if min_content_len and len(content) < min_content_len:
+        return False
+
+    if max_content_len and len(content) > max_content_len:
+        return False
+
+    return has_non_english_chars(content)
+
+
+def print_stats(
+    conversations: "list[dict[Any, Any]]", tokenizer: Optional[AutoTokenizer] = None
+) -> None:
+    # Collect statistics
+    stats = []
+
+    print("\nCollecting statistics...")
+    for item in tqdm.tqdm(conversations):
+        # item has "id" and "messages"
+        messages = item["messages"]
+
+        user_turns = 0
+        assistant_turns = 0
+        user_words = 0
+        assistant_words = 0
+        conv_chars = 0
+
+        user_tokens: list[int] = []
+        assistant_tokens: list[int] = []
+
+        for m in messages:
+            content = m["content"]
+            conv_chars += len(content)
+            content_num_words = content.count(" ") + 1
+
+            num_tokens = 0
+            if tokenizer:
+                num_tokens = len(tokenizer(m["content"]).input_ids)
+
+            if m["role"] == "user":
+                user_turns += 1
+                user_words += content_num_words
+                if tokenizer:
+                    user_tokens.append(num_tokens)
+
+            elif m["role"] == "assistant":
+                assistant_turns += 1
+                assistant_words += content_num_words
+                if tokenizer:
+                    assistant_tokens.append(num_tokens)
+
+        # assert user_turns == assistant_turns, \
+        # f"Invalid conversation ID {item['id']}"
+
+        conv_words = user_words + assistant_words
+        item_stats = {
+            "user_turns": user_turns,
+            "assistant_turns": assistant_turns,
+            "user_words": user_words,
+            "assistant_words": assistant_words,
+            "conv_turns": len(messages),
+            "conv_words": conv_words,
+            "conv_characters": conv_chars,
+        }
+
+        if len(user_tokens) > 0:
+            item_stats["user_tokens"] = int(mean(user_tokens))
+
+        if len(assistant_tokens) > 0:
+            item_stats["assistant_tokens"] = int(mean(assistant_tokens))
+
+        stats.append(item_stats)
+
+    print("\nStatistics:")
+    percentiles = [0.25, 0.5, 0.75, 0.9, 0.99, 0.999, 0.9999]
+    df = pd.DataFrame(stats)
+    print(df.describe(percentiles=percentiles).transpose())
+
+
+def convert_sharegpt_to_openai(
+    seed: int,
+    input_file: str,
+    output_file: str,
+    max_items: Optional[int],
+    min_content_len: Optional[int] = None,
+    max_content_len: Optional[int] = None,
+    min_turns: Optional[int] = None,
+    max_turns: Optional[int] = None,
+    model: Optional[str] = None,
+) -> None:
+    if min_turns and max_turns:
+        assert min_turns <= max_turns
+
+    if min_content_len and max_content_len:
+        # Verify that min is not larger than max if both were given
+        assert min_content_len <= max_content_len
+
+    print(
+        f"Input parameters:\n{seed=}, {max_items=}, {min_content_len=},"
+        f" {max_content_len=}, {min_turns=}, {max_turns=}\n"
+    )
+
+    random.seed(seed)
+
+    tokenizer = None
+    if model is not None:
+        print(f"Loading tokenizer from: {model}")
+        tokenizer = AutoTokenizer.from_pretrained(model)
+
+    # Read the ShareGPT JSON file
+    print(f"Reading file: {input_file}")
+    with open(input_file, encoding="utf-8") as f:
+        # Should be a list of dicts
+        # Each dict should have "id" (string) and "conversations" (list of dicts)
+        sharegpt_data = json.load(f)
+
+    assert isinstance(sharegpt_data, list), "Input file should contain a list of dicts"
+
+    print(f"Total items in input file: {len(sharegpt_data):,}")
+
+    print(f"Shuffling dataset with seed {seed}")
+    random.shuffle(sharegpt_data)
+
+    # Map conversation ID to the all the messages
+    conversation_parts: dict[str, list[Any]] = {}
+
+    for item in tqdm.tqdm(sharegpt_data):
+        assert "id" in item, "Missing key 'id'"
+        assert "conversations" in item, "Missing key 'conversations'"
+
+        # Conversation ID (e.g: "hiWPlMD") and part/session (0, 1, 2, etc.)
+        conv_id, _ = item["id"].split("_")
+        new_turns = item["conversations"]
+
+        if conv_id not in conversation_parts:
+            # Start new conversation
+            conversation_parts[conv_id] = []
+        elif len(conversation_parts[conv_id]) > 0 and len(new_turns) > 0:
+            prev_turns = conversation_parts[conv_id][-1]
+            if prev_turns[-1]["from"] == new_turns[0]["from"]:
+                new_turns = new_turns[1:]
+
+        if len(new_turns) > 0:
+            # We assume that parts are in order in the ShareGPT dataset
+            conversation_parts[conv_id].append(new_turns)
+
+    dataset: list[dict[str, Any]] = []
+    for conv_id, conv_parts in conversation_parts.items():
+        new_item = {"id": conv_id}
+
+        conversations: list[dict[str, str]] = []
+
+        # Merge all parts
+        for conv_part in conv_parts:
+            conversations.extend(conv_part)
+
+        if len(conversations) > 0:
+            new_item["conversations"] = conversations
+            dataset.append(new_item)
+
+    print(f"Total unique conversations (IDs) in input file: {len(dataset):,}")
+
+    # Final output data
+    final_openai_dataset: list[dict] = []
+
+    # Filter conversations from the ShareGPT dataset and convert to OpenAI format
+    for item in tqdm.tqdm(dataset):
+        messages: list[dict] = []
+
+        assert "id" in item, "Missing key 'id'"
+        assert "conversations" in item, "Missing key 'conversations'"
+
+        conv_id = item["id"]
+        conversations = item["conversations"]
+
+        if min_turns is not None and len(conversations) < min_turns:
+            # Skip short conversations
+            continue
+
+        # Convert each message in the conversation, up to max_turns if specified
+        for i, turn in enumerate(conversations):
+            assert "from" in turn and "value" in turn, (
+                f"Invalid conversation ID {conv_id} - missing 'from' or 'value'"
+            )
+
+            role = None
+            turn_from = turn["from"]
+
+            if turn_from in {"human", "user"}:
+                role = "user"
+            elif turn_from in {"gpt", "bing", "chatgpt", "bard"}:
+                role = "assistant"
+            elif turn_from == "system":
+                role = "system"
+
+            assert role is not None, (
+                f"Invalid conversation ID {conv_id} - 'from'='{turn_from}' is invalid"
+            )
+
+            if i == 0 and role != "user":
+                # If the first message is from assistant (gpt), skip it.
+                # this happens when the conversation is a follow-up
+                # to a previous conversation (from the same user).
+                continue
+
+            if max_turns is not None and i >= max_turns:
+                break
+
+            # Convert message to OpenAI format (with "role" and "content")
+            content = turn["value"]
+            messages.append({"role": role, "content": content})
+
+        # Add the converted conversation to the OpenAI format
+        if len(messages) > 0:
+            valid_messages = True
+
+            # First turn should always be from the user
+            user_turn = True
+
+            for m in messages:
+                # Make sure that turns alternate between user and assistant
+                if (user_turn and m["role"] != "user") or (
+                    not user_turn and m["role"] != "assistant"
+                ):
+                    valid_messages = False
+                    break
+
+                user_turn = not user_turn
+
+                content = m["content"]
+                valid_messages = content_is_valid(
+                    content, min_content_len, max_content_len
+                )
+                if not valid_messages:
+                    break
+
+            if valid_messages is True:
+                final_openai_dataset.append({"id": conv_id, "messages": messages})
+
+    assert len(final_openai_dataset) > 0, "Final number of conversations is zero"
+
+    print_stats(final_openai_dataset)
+
+    print_stats_again = False
+    if max_items is not None and len(final_openai_dataset) > max_items:
+        print(f"\n\nSampling {max_items} items from the dataset...")
+        print_stats_again = True
+        final_openai_dataset = random.sample(final_openai_dataset, max_items)
+
+    if print_stats_again:
+        # Print stats after the dataset changed
+        print_stats(final_openai_dataset, tokenizer)
+
+    # Write the converted data to a new JSON file
+    final_size = len(final_openai_dataset)
+    print(f"\nTotal conversations converted (after filtering): {final_size:,}")
+    print(f"\nWriting file: {output_file}")
+    with open(output_file, "w", encoding="utf-8") as f:
+        json.dump(final_openai_dataset, f, ensure_ascii=False, indent=2)
+
+
+def main() -> None:
+    parser = argparse.ArgumentParser(
+        description="Convert ShareGPT dataset to OpenAI API format"
+    )
+    parser.add_argument("input_file", help="Path to the input ShareGPT JSON file")
+    parser.add_argument(
+        "output_file", help="Path to the output OpenAI format JSON file"
+    )
+    parser.add_argument(
+        "--seed", type=int, default=0, help="Seed for random number generators"
+    )
+    parser.add_argument(
+        "--max-items",
+        type=int,
+        default=None,
+        help="Maximum number of items in the output file",
+    )
+    parser.add_argument(
+        "--min-turns",
+        type=int,
+        default=None,
+        help="Minimum number of turns per conversation",
+    )
+    parser.add_argument(
+        "--max-turns",
+        type=int,
+        default=None,
+        help="Maximum number of turns per conversation",
+    )
+    parser.add_argument(
+        "--min-content-len",
+        type=int,
+        default=None,
+        help="Min number of characters in the messages' content",
+    )
+    parser.add_argument(
+        "--max-content-len",
+        type=int,
+        default=None,
+        help="Max number of characters in the messages' content",
+    )
+    parser.add_argument(
+        "--model",
+        type=str,
+        default=None,
+        help="LLM model, only the tokenizer will be used",
+    )
+
+    args = parser.parse_args()
+
+    convert_sharegpt_to_openai(
+        args.seed,
+        args.input_file,
+        args.output_file,
+        args.max_items,
+        args.min_content_len,
+        args.max_content_len,
+        args.min_turns,
+        args.max_turns,
+        args.model,
+    )
+
+
+if __name__ == "__main__":
+    main()

benchmarks/multi_turn/generate_multi_turn.json

@@ -0,0 +1,35 @@
+{
+    "filetype": "generate_conversations",
+    "num_conversations": 24,
+    "text_files": ["pg1184.txt"],
+    "print_stats": false,
+    "prompt_input": {
+        "num_turns": {
+            "distribution": "uniform",
+            "min": 12,
+            "max": 18
+        },
+        "common_prefix_num_tokens": {
+            "distribution": "constant",
+            "value": 500
+        },
+        "prefix_num_tokens": {
+            "distribution": "lognormal",
+            "mean": 6,
+            "sigma": 4,
+            "max": 1500
+        },
+        "num_tokens": {
+            "distribution": "uniform",
+            "min": 120,
+            "max": 160
+        }
+    },
+    "prompt_output": {
+        "num_tokens": {
+            "distribution": "uniform",
+            "min": 80,
+            "max": 120
+        }
+    }
+}

benchmarks/multi_turn/requirements.txt

@@ -0,0 +1,5 @@
+numpy>=1.24
+pandas>=2.0.0
+aiohttp>=3.10
+transformers>=4.46
+xlsxwriter>=3.2.1

cmake/cpu_extension.cmake

@@ -12,9 +12,8 @@ endif()
 #
 # Define environment variables for special configurations
 #
-if(DEFINED ENV{VLLM_CPU_AVX512BF16})
-    set(ENABLE_AVX512BF16 ON)
-endif()
+set(ENABLE_AVX512BF16 $ENV{VLLM_CPU_AVX512BF16})
+set(ENABLE_AVX512VNNI $ENV{VLLM_CPU_AVX512VNNI})
 
 include_directories("${CMAKE_SOURCE_DIR}/csrc")
 
@@ -59,6 +58,22 @@ function (find_isa CPUINFO TARGET OUT)
     endif()
 endfunction()
 
+
+function(check_sysctl TARGET OUT)
+    execute_process(COMMAND sysctl -n "${TARGET}"
+                    RESULT_VARIABLE SYSCTL_RET
+                    OUTPUT_VARIABLE SYSCTL_INFO
+                    ERROR_QUIET
+                    OUTPUT_STRIP_TRAILING_WHITESPACE)
+    if(SYSCTL_RET EQUAL 0 AND
+      (SYSCTL_INFO STREQUAL "1" OR SYSCTL_INFO GREATER 0))
+        set(${OUT} ON PARENT_SCOPE)
+    else()
+        set(${OUT} OFF PARENT_SCOPE)
+    endif()
+endfunction()
+
+
 function (is_avx512_disabled OUT)
     set(DISABLE_AVX512 $ENV{VLLM_CPU_DISABLE_AVX512})
     if(DISABLE_AVX512 AND DISABLE_AVX512 STREQUAL "true")
@@ -71,7 +86,10 @@ endfunction()
 is_avx512_disabled(AVX512_DISABLED)
 
 if (MACOSX_FOUND AND CMAKE_SYSTEM_PROCESSOR STREQUAL "arm64")
-    set(APPLE_SILICON_FOUND TRUE)
+    message(STATUS "Apple Silicon Detected")
+    set(ENABLE_NUMA OFF)
+    check_sysctl(hw.optional.neon ASIMD_FOUND)
+    check_sysctl(hw.optional.arm.FEAT_BF16 ARM_BF16_FOUND)
 else()
     find_isa(${CPUINFO} "avx2" AVX2_FOUND)
     find_isa(${CPUINFO} "avx512f" AVX512_FOUND)
@@ -83,7 +101,6 @@ else()
     find_isa(${CPUINFO} "S390" S390_FOUND)
 endif()
 
-
 if (AVX512_FOUND AND NOT AVX512_DISABLED)
     list(APPEND CXX_COMPILE_FLAGS
         "-mavx512f"
@@ -96,12 +113,30 @@ if (AVX512_FOUND AND NOT AVX512_DISABLED)
         if (CMAKE_CXX_COMPILER_ID STREQUAL "GNU" AND
             CMAKE_CXX_COMPILER_VERSION VERSION_GREATER_EQUAL 12.3)
             list(APPEND CXX_COMPILE_FLAGS "-mavx512bf16")
+            set(ENABLE_AVX512BF16 ON)
         else()
+            set(ENABLE_AVX512BF16 OFF)
             message(WARNING "Disable AVX512-BF16 ISA support, requires gcc/g++ >= 12.3")
         endif()
     else()
+        set(ENABLE_AVX512BF16 OFF)
         message(WARNING "Disable AVX512-BF16 ISA support, no avx512_bf16 found in local CPU flags." " If cross-compilation is required, please set env VLLM_CPU_AVX512BF16=1.")
     endif()
+
+    find_isa(${CPUINFO} "avx512_vnni" AVX512VNNI_FOUND)
+    if (AVX512VNNI_FOUND OR ENABLE_AVX512VNNI)
+        if (CMAKE_CXX_COMPILER_ID STREQUAL "GNU" AND
+            CMAKE_CXX_COMPILER_VERSION VERSION_GREATER_EQUAL 12.3)
+            list(APPEND CXX_COMPILE_FLAGS "-mavx512vnni")
+            set(ENABLE_AVX512VNNI ON)
+        else()
+            set(ENABLE_AVX512VNNI OFF)
+            message(WARNING "Disable AVX512-VNNI ISA support, requires gcc/g++ >= 12.3")
+        endif()
+    else()
+        set(ENABLE_AVX512VNNI OFF)
+        message(WARNING "Disable AVX512-VNNI ISA support, no avx512_vnni found in local CPU flags." " If cross-compilation is required, please set env VLLM_CPU_AVX512VNNI=1.")
+    endif()
     
 elseif (AVX2_FOUND)
     list(APPEND CXX_COMPILE_FLAGS "-mavx2")
@@ -132,9 +167,6 @@ elseif (ASIMD_FOUND)
         set(MARCH_FLAGS "-march=armv8.2-a+dotprod+fp16")  
     endif()
     list(APPEND CXX_COMPILE_FLAGS ${MARCH_FLAGS})     
-elseif(APPLE_SILICON_FOUND)
-    message(STATUS "Apple Silicon Detected")
-    set(ENABLE_NUMA OFF)
 elseif (S390_FOUND)
     message(STATUS "S390 detected")
     # Check for S390 VXE support
@@ -148,17 +180,32 @@ else()
 endif()
 
 #
-# Build oneDNN for W8A8 GEMM kernels (only for x86-AVX512 platforms)
-#
-if (AVX512_FOUND AND NOT AVX512_DISABLED)
+# Build oneDNN for W8A8 GEMM kernels (only for x86-AVX512 /ARM platforms)
+# Flag to enable ACL kernels for AARCH64 platforms
+if ( VLLM_BUILD_ACL STREQUAL "ON")
+    set(USE_ACL ON)
+else()
+    set(USE_ACL OFF)
+endif()
+
+if ((AVX512_FOUND AND NOT AVX512_DISABLED) OR ASIMD_FOUND)
     FetchContent_Declare(
         oneDNN
         GIT_REPOSITORY https://github.com/oneapi-src/oneDNN.git
-        GIT_TAG  v3.7.1
+        GIT_TAG  v3.8.1
         GIT_PROGRESS TRUE
         GIT_SHALLOW TRUE
     )
 
+    if(USE_ACL)
+        find_library(ARM_COMPUTE_LIBRARY NAMES arm_compute PATHS $ENV{ACL_ROOT_DIR}/build/)
+        if(NOT ARM_COMPUTE_LIBRARY)
+            message(FATAL_ERROR "Could not find ARM Compute Library: please set ACL_ROOT_DIR")
+        endif()
+        set(ONEDNN_AARCH64_USE_ACL "ON")
+        set(CMAKE_CXX_FLAGS "${CMAKE_CXX_FLAGS} -Wl,-rpath,$ENV{ACL_ROOT_DIR}/build/")
+        endif()
+
     set(ONEDNN_LIBRARY_TYPE "STATIC")
     set(ONEDNN_BUILD_DOC "OFF")
     set(ONEDNN_BUILD_EXAMPLES "OFF")
@@ -231,11 +278,29 @@ if (AVX512_FOUND AND NOT AVX512_DISABLED)
         "csrc/cpu/quant.cpp"
         "csrc/cpu/shm.cpp"
         ${VLLM_EXT_SRC})
+    if (ENABLE_AVX512BF16 AND ENABLE_AVX512VNNI)
+        set(VLLM_EXT_SRC
+            "csrc/cpu/sgl-kernels/gemm.cpp"
+            "csrc/cpu/sgl-kernels/gemm_int8.cpp"
+            "csrc/cpu/sgl-kernels/gemm_fp8.cpp"
+            "csrc/cpu/sgl-kernels/moe.cpp"
+            "csrc/cpu/sgl-kernels/moe_int8.cpp"
+            "csrc/cpu/sgl-kernels/moe_fp8.cpp"
+            ${VLLM_EXT_SRC})
+        add_compile_definitions(-DCPU_CAPABILITY_AVX512)
+    endif()
 elseif(POWER10_FOUND)
     set(VLLM_EXT_SRC
         "csrc/cpu/quant.cpp"
         ${VLLM_EXT_SRC})
 endif()
+if (ASIMD_FOUND)
+    set(VLLM_EXT_SRC
+        "csrc/cpu/quant.cpp"
+        ${VLLM_EXT_SRC})
+endif()
+
+message(STATUS "CPU extension source files: ${VLLM_EXT_SRC}")
 
 #
 # Define extension targets

cmake/external_projects/flashmla.cmake

@@ -19,7 +19,7 @@ else()
   FetchContent_Declare(
         flashmla
         GIT_REPOSITORY https://github.com/vllm-project/FlashMLA.git
-        GIT_TAG 575f7724b9762f265bbee5889df9c7d630801845
+        GIT_TAG 0e43e774597682284358ff2c54530757b654b8d1
         GIT_PROGRESS TRUE
         CONFIGURE_COMMAND ""
         BUILD_COMMAND ""
@@ -37,9 +37,9 @@ cuda_archs_loose_intersection(FLASH_MLA_ARCHS "9.0a" "${CUDA_ARCHS}")
 if(${CMAKE_CUDA_COMPILER_VERSION} VERSION_GREATER 12.3 AND FLASH_MLA_ARCHS)
     set(FlashMLA_SOURCES
         ${flashmla_SOURCE_DIR}/csrc/flash_api.cpp
-        ${flashmla_SOURCE_DIR}/csrc/flash_fwd_mla_bf16_sm90.cu
-        ${flashmla_SOURCE_DIR}/csrc/flash_fwd_mla_fp16_sm90.cu
-        ${flashmla_SOURCE_DIR}/csrc/flash_fwd_mla_metadata.cu)
+        ${flashmla_SOURCE_DIR}/csrc/kernels/splitkv_mla.cu
+        ${flashmla_SOURCE_DIR}/csrc/kernels/mla_combine.cu
+        ${flashmla_SOURCE_DIR}/csrc/kernels/get_mla_metadata.cu)
 
     set(FlashMLA_INCLUDES
         ${flashmla_SOURCE_DIR}/csrc/cutlass/include