more fixes

.circleci/config.yml

@@ -7,25 +7,12 @@ parameters:
     nightly:
         type: boolean
         default: false
-    GHA_Actor:
-        type: string
-        default: ""
-    GHA_Action:
-        type: string
-        default: ""
-    GHA_Event:
-        type: string
-        default: ""
-    GHA_Meta:
-        type: string
-        default: ""
 
 jobs:
     # Ensure running with CircleCI/huggingface
     check_circleci_user:
         docker:
             - image: python:3.10-slim
-        resource_class: small
         parallelism: 1
         steps:
             - run: echo $CIRCLE_PROJECT_USERNAME
@@ -47,44 +34,64 @@ jobs:
             - run: echo 'export "GIT_COMMIT_MESSAGE=$(git show -s --format=%s)"' >> "$BASH_ENV" && source "$BASH_ENV"
             - run: mkdir -p test_preparation
             - run: python utils/tests_fetcher.py | tee tests_fetched_summary.txt
+            - store_artifacts:
+                  path: ~/transformers/tests_fetched_summary.txt
+            - run: |
+                if [ -f test_list.txt ]; then
+                    cp test_list.txt test_preparation/test_list.txt
+                else
+                    touch test_preparation/test_list.txt
+                fi
+            - run: |
+                  if [ -f examples_test_list.txt ]; then
+                      mv examples_test_list.txt test_preparation/examples_test_list.txt
+                  else
+                      touch test_preparation/examples_test_list.txt
+                  fi
+            - run: |
+                  if [ -f filtered_test_list_cross_tests.txt ]; then
+                      mv filtered_test_list_cross_tests.txt test_preparation/filtered_test_list_cross_tests.txt
+                  else
+                      touch test_preparation/filtered_test_list_cross_tests.txt
+                  fi
+            - run: |
+                if [ -f doctest_list.txt ]; then
+                    cp doctest_list.txt test_preparation/doctest_list.txt
+                else
+                    touch test_preparation/doctest_list.txt
+                fi
+            - run: |
+                if [ -f test_repo_utils.txt ]; then
+                    mv test_repo_utils.txt test_preparation/test_repo_utils.txt
+                else
+                    touch test_preparation/test_repo_utils.txt
+                fi
             - run: python utils/tests_fetcher.py --filter_tests
+            - run: |
+                if [ -f test_list.txt ]; then
+                    mv test_list.txt test_preparation/filtered_test_list.txt
+                else
+                    touch test_preparation/filtered_test_list.txt
+                fi
+            - store_artifacts:
+                  path: test_preparation/test_list.txt
+            - store_artifacts:
+                  path: test_preparation/doctest_list.txt
+            - store_artifacts:
+                  path: ~/transformers/test_preparation/filtered_test_list.txt
+            - store_artifacts:
+                  path: test_preparation/examples_test_list.txt
             - run: export "GIT_COMMIT_MESSAGE=$(git show -s --format=%s)" && echo $GIT_COMMIT_MESSAGE && python .circleci/create_circleci_config.py --fetcher_folder test_preparation
             - run: |
-                if [ ! -s test_preparation/generated_config.yml ]; then
-                    echo "No tests to run, exiting early!"
-                    circleci-agent step halt
-                fi
-
+                  if [ ! -s test_preparation/generated_config.yml ]; then
+                      echo "No tests to run, exiting early!"
+                      circleci-agent step halt
+                  fi
             - store_artifacts:
-                path: test_preparation
-
-            - run:
-                name: "Retrieve Artifact Paths"
-                # [reference] https://circleci.com/docs/api/v2/index.html#operation/getJobArtifacts
-                # `CIRCLE_TOKEN` is defined as an environment variables set within a context, see `https://circleci.com/docs/contexts/`
-                command: |
-                    project_slug="gh/${CIRCLE_PROJECT_USERNAME}/${CIRCLE_PROJECT_REPONAME}"
-                    job_number=${CIRCLE_BUILD_NUM}
-                    url="https://circleci.com/api/v2/project/${project_slug}/${job_number}/artifacts"
-                    curl -o test_preparation/artifacts.json ${url} --header "Circle-Token: $CIRCLE_TOKEN"
-            - run:
-                name: "Prepare pipeline parameters"
-                command: |
-                    python utils/process_test_artifacts.py
-
-            # To avoid too long generated_config.yaml on the continuation orb, we pass the links to the artifacts as parameters.
-            # Otherwise the list of tests was just too big. Explicit is good but for that it was a limitation.
-            # We used:
-
-            # https://circleci.com/docs/api/v2/index.html#operation/getJobArtifacts : to get the job artifacts
-            # We could not pass a nested dict, which is why we create the test_file_... parameters for every single job
-
+                path: test_preparation/generated_config.yml
             - store_artifacts:
-                path: test_preparation/transformed_artifacts.json
-            - store_artifacts:
-                path: test_preparation/artifacts.json
+                path: test_preparation/filtered_test_list_cross_tests.txt
             - continuation/continue:
-                parameters:  test_preparation/transformed_artifacts.json
                 configuration_path: test_preparation/generated_config.yml
 
     # To run all tests for the nightly build
@@ -95,47 +102,22 @@ jobs:
         parallelism: 1
         steps:
             - checkout
-            - run: uv pip install -U -e .
-            - run: echo 'export "GIT_COMMIT_MESSAGE=$(git show -s --format=%s)"' >> "$BASH_ENV" && source "$BASH_ENV"
-            - run: mkdir -p test_preparation
-            - run: python utils/tests_fetcher.py --fetch_all | tee tests_fetched_summary.txt
-            - run: python utils/tests_fetcher.py --filter_tests
-            - run: export "GIT_COMMIT_MESSAGE=$(git show -s --format=%s)" && echo $GIT_COMMIT_MESSAGE && python .circleci/create_circleci_config.py --fetcher_folder test_preparation
+            - run: uv pip install -e .
             - run: |
-                if [ ! -s test_preparation/generated_config.yml ]; then
-                    echo "No tests to run, exiting early!"
-                    circleci-agent step halt
-                fi
-
+                  mkdir test_preparation
+                  echo -n "tests" > test_preparation/test_list.txt
+                  echo -n "all" > test_preparation/examples_test_list.txt
+                  echo -n "tests/repo_utils" > test_preparation/test_repo_utils.txt
+            - run: |
+                  echo -n "tests" > test_list.txt
+                  python utils/tests_fetcher.py --filter_tests
+                  mv test_list.txt test_preparation/filtered_test_list.txt
+            - run: python .circleci/create_circleci_config.py --fetcher_folder test_preparation
+            - run: cp test_preparation/generated_config.yml test_preparation/generated_config.txt
             - store_artifacts:
-                path: test_preparation
-
-            - run:
-                name: "Retrieve Artifact Paths"
-                command: |
-                    project_slug="gh/${CIRCLE_PROJECT_USERNAME}/${CIRCLE_PROJECT_REPONAME}"
-                    job_number=${CIRCLE_BUILD_NUM}
-                    url="https://circleci.com/api/v2/project/${project_slug}/${job_number}/artifacts"
-                    curl -o  test_preparation/artifacts.json ${url}
-            - run:
-                name: "Prepare pipeline parameters"
-                command: |
-                    python utils/process_test_artifacts.py
-
-            # To avoid too long generated_config.yaml on the continuation orb, we pass the links to the artifacts as parameters.
-            # Otherwise the list of tests was just too big. Explicit is good but for that it was a limitation.
-            # We used:
-
-            # https://circleci.com/docs/api/v2/index.html#operation/getJobArtifacts : to get the job artifacts
-            # We could not pass a nested dict, which is why we create the test_file_... parameters for every single job
-
-            - store_artifacts:
-                path: test_preparation/transformed_artifacts.json
-            - store_artifacts:
-                path: test_preparation/artifacts.json
+                  path: test_preparation/generated_config.txt
             - continuation/continue:
-                parameters:  test_preparation/transformed_artifacts.json
-                configuration_path: test_preparation/generated_config.yml
+                  configuration_path: test_preparation/generated_config.yml
 
     check_code_quality:
         working_directory: ~/transformers
@@ -148,7 +130,7 @@ jobs:
         parallelism: 1
         steps:
             - checkout
-            - run: uv pip install -e ".[quality]"
+            - run: uv pip install -e .
             - run:
                 name: Show installed libraries and their versions
                 command: pip freeze | tee installed.txt
@@ -156,7 +138,7 @@ jobs:
                   path: ~/transformers/installed.txt
             - run: python -c "from transformers import *" || (echo '🚨 import failed, this means you introduced unprotected imports! 🚨'; exit 1)
             - run: ruff check examples tests src utils
-            - run: ruff format examples tests src utils --check
+            - run: ruff format tests src utils --check
             - run: python utils/custom_init_isort.py --check_only
             - run: python utils/sort_auto_mappings.py --check_only
             - run: python utils/check_doc_toc.py
@@ -173,51 +155,36 @@ jobs:
         parallelism: 1
         steps:
             - checkout
-            - run: uv pip install -e ".[quality]"
+            - run: uv pip install -e .
             - run:
                 name: Show installed libraries and their versions
                 command: pip freeze | tee installed.txt
             - store_artifacts:
                   path: ~/transformers/installed.txt
             - run: python utils/check_copies.py
-            - run: python utils/check_modular_conversion.py
+            - run: python utils/check_table.py
             - run: python utils/check_dummies.py
             - run: python utils/check_repo.py
             - run: python utils/check_inits.py
-            - run: python utils/check_pipeline_typing.py
             - run: python utils/check_config_docstrings.py
             - run: python utils/check_config_attributes.py
             - run: python utils/check_doctest_list.py
             - run: make deps_table_check_updated
             - run: python utils/update_metadata.py --check-only
             - run: python utils/check_docstrings.py
+            - run: python utils/check_support_list.py
 
 workflows:
     version: 2
     setup_and_quality:
         when:
-            and:
-                - equal: [<<pipeline.project.git_url>>, https://github.com/huggingface/transformers]
-                - not: <<pipeline.parameters.nightly>>
+            not: <<pipeline.parameters.nightly>>
         jobs:
             - check_circleci_user
             - check_code_quality
             - check_repository_consistency
             - fetch_tests
 
-    setup_and_quality_2:
-        when:
-            not:
-                 equal: [<<pipeline.project.git_url>>, https://github.com/huggingface/transformers]
-        jobs:
-            - check_circleci_user
-            - check_code_quality
-            - check_repository_consistency
-            - fetch_tests:
-                # [reference] https://circleci.com/docs/contexts/
-                context:
-                    - TRANSFORMERS_CONTEXT
-
     nightly:
         when: <<pipeline.parameters.nightly>>
         jobs:

.circleci/create_circleci_config.py

@@ -28,54 +28,21 @@ COMMON_ENV_VARIABLES = {
     "TRANSFORMERS_IS_CI": True,
     "PYTEST_TIMEOUT": 120,
     "RUN_PIPELINE_TESTS": False,
-    # will be adjust in `CircleCIJob.to_dict`.
-    "RUN_FLAKY": True,
+    "RUN_PT_TF_CROSS_TESTS": False,
+    "RUN_PT_FLAX_CROSS_TESTS": False,
 }
 # Disable the use of {"s": None} as the output is way too long, causing the navigation on CircleCI impractical
-COMMON_PYTEST_OPTIONS = {"max-worker-restart": 0, "vvv": None, "rsfE":None}
+COMMON_PYTEST_OPTIONS = {"max-worker-restart": 0, "dist": "loadfile", "v": None}
 DEFAULT_DOCKER_IMAGE = [{"image": "cimg/python:3.8.12"}]
 
-# Strings that commonly appear in the output of flaky tests when they fail. These are used with `pytest-rerunfailures`
-# to rerun the tests that match these patterns.
-FLAKY_TEST_FAILURE_PATTERNS = [
-    "OSError",  # Machine/connection transient error
-    "Timeout",  # Machine/connection transient error
-    "ConnectionError",  # Connection transient error
-    "FileNotFoundError",  # Raised by `datasets` on Hub failures
-    "PIL.UnidentifiedImageError",  # Raised by `PIL.Image.open` on connection issues
-    "HTTPError",  # Also catches HfHubHTTPError
-    "AssertionError: Tensor-likes are not close!",  # `torch.testing.assert_close`, we might have unlucky random values
-    # TODO: error downloading tokenizer's `merged.txt` from hub can cause all the exceptions below. Throw and handle
-    # them under a single message.
-    "TypeError: expected str, bytes or os.PathLike object, not NoneType",
-    "TypeError: stat: path should be string, bytes, os.PathLike or integer, not NoneType",
-    "Converting from Tiktoken failed",
-    "KeyError: <class ",
-    "TypeError: not a string",
-]
-
 
 class EmptyJob:
     job_name = "empty"
 
     def to_dict(self):
-        steps = [{"run": 'ls -la'}]
-        if self.job_name == "collection_job":
-            steps.extend(
-                [
-                    "checkout",
-                    {"run": "pip install requests || true"},
-                    {"run": """while [[ $(curl --location --request GET "https://circleci.com/api/v2/workflow/$CIRCLE_WORKFLOW_ID/job" --header "Circle-Token: $CCI_TOKEN"| jq -r '.items[]|select(.name != "collection_job")|.status' | grep -c "running") -gt 0 ]]; do sleep 5; done || true"""},
-                    {"run": 'python utils/process_circleci_workflow_test_reports.py --workflow_id $CIRCLE_WORKFLOW_ID || true'},
-                    {"store_artifacts": {"path": "outputs"}},
-                    {"run": 'echo "All required jobs have now completed"'},
-                ]
-            )
-
         return {
             "docker": copy.deepcopy(DEFAULT_DOCKER_IMAGE),
-            "resource_class": "small",
-            "steps": steps,
+            "steps":["checkout"],
         }
 
 
@@ -83,15 +50,16 @@ class EmptyJob:
 class CircleCIJob:
     name: str
     additional_env: Dict[str, Any] = None
+    cache_name: str = None
+    cache_version: str = "0.8.2"
     docker_image: List[Dict[str, str]] = None
     install_steps: List[str] = None
     marker: Optional[str] = None
-    parallelism: Optional[int] = 0
-    pytest_num_workers: int = 8
+    parallelism: Optional[int] = 1
+    pytest_num_workers: int = 12
     pytest_options: Dict[str, Any] = None
-    resource_class: Optional[str] = "xlarge"
+    resource_class: Optional[str] = "2xlarge"
     tests_to_run: Optional[List[str]] = None
-    num_test_files_per_worker: Optional[int] = 10
     # This should be only used for doctest job!
     command_timeout: Optional[int] = None
 
@@ -99,6 +67,8 @@ class CircleCIJob:
         # Deal with defaults for mutable attributes.
         if self.additional_env is None:
             self.additional_env = {}
+        if self.cache_name is None:
+            self.cache_name = self.name
         if self.docker_image is None:
             # Let's avoid changing the default list and make a copy.
             self.docker_image = copy.deepcopy(DEFAULT_DOCKER_IMAGE)
@@ -109,163 +79,269 @@ class CircleCIJob:
                 self.docker_image[0]["image"] = f"{self.docker_image[0]['image']}:dev"
             print(f"Using {self.docker_image} docker image")
         if self.install_steps is None:
-            self.install_steps = ["uv venv && uv pip install ."]
+            self.install_steps = []
         if self.pytest_options is None:
             self.pytest_options = {}
         if isinstance(self.tests_to_run, str):
             self.tests_to_run = [self.tests_to_run]
-        else:
-            test_file = os.path.join("test_preparation" , f"{self.job_name}_test_list.txt")
-            print("Looking for ", test_file)
-            if os.path.exists(test_file):
-                with open(test_file) as f:
-                    expanded_tests = f.read().strip().split("\n")
-                self.tests_to_run = expanded_tests
-                print("Found:", expanded_tests)
-            else:
-                self.tests_to_run = []
-                print("not Found")
+        if self.parallelism is None:
+            self.parallelism = 1
 
     def to_dict(self):
         env = COMMON_ENV_VARIABLES.copy()
-        # Do not run tests decorated by @is_flaky on pull requests
-        env['RUN_FLAKY'] = os.environ.get("CIRCLE_PULL_REQUEST", "") == ""
         env.update(self.additional_env)
 
+        cache_branch_prefix = os.environ.get("CIRCLE_BRANCH", "pull")
+        if cache_branch_prefix != "main":
+            cache_branch_prefix = "pull"
+
         job = {
             "docker": self.docker_image,
             "environment": env,
         }
         if self.resource_class is not None:
             job["resource_class"] = self.resource_class
+        if self.parallelism is not None:
+            job["parallelism"] = self.parallelism
+        steps = [
+            "checkout",
+            {"attach_workspace": {"at": "test_preparation"}},
+        ]
+        steps.extend([{"run": l} for l in self.install_steps])
+        steps.append({"run": {"name": "Show installed libraries and their size", "command": """du -h -d 1 "$(pip -V | cut -d ' ' -f 4 | sed 's/pip//g')" | grep -vE "dist-info|_distutils_hack|__pycache__" | sort -h | tee installed.txt || true"""}})
+        steps.append({"run": {"name": "Show installed libraries and their versions", "command": """pip list --format=freeze | tee installed.txt || true"""}})
+
+        steps.append({"run":{"name":"Show biggest libraries","command":"""dpkg-query --show --showformat='${Installed-Size}\t${Package}\n' | sort -rh | head -25 | sort -h | awk '{ package=$2; sub(".*/", "", package); printf("%.5f GB %s\n", $1/1024/1024, package)}' || true"""}})
+        steps.append({"store_artifacts": {"path": "installed.txt"}})
 
         all_options = {**COMMON_PYTEST_OPTIONS, **self.pytest_options}
         pytest_flags = [f"--{key}={value}" if (value is not None or key in ["doctest-modules"]) else f"-{key}" for key, value in all_options.items()]
         pytest_flags.append(
             f"--make-reports={self.name}" if "examples" in self.name else f"--make-reports=tests_{self.name}"
         )
-                # Examples special case: we need to download NLTK files in advance to avoid cuncurrency issues
-        timeout_cmd = f"timeout {self.command_timeout} " if self.command_timeout else ""
-        marker_cmd = f"-m '{self.marker}'" if self.marker is not None else ""
-        junit_flags = f" -p no:warning -o junit_family=xunit1 --junitxml=test-results/junit.xml"
-        joined_flaky_patterns = "|".join(FLAKY_TEST_FAILURE_PATTERNS)
-        repeat_on_failure_flags = f"--reruns 5 --reruns-delay 2 --only-rerun '({joined_flaky_patterns})'"
-        parallel = f' << pipeline.parameters.{self.job_name}_parallelism >> '
-        steps = [
-            "checkout",
-            {"attach_workspace": {"at": "test_preparation"}},
-            {"run": "apt-get update && apt-get install -y curl"},
-            {"run": " && ".join(self.install_steps)},
-            {"run": {"name": "Download NLTK files", "command": """python -c "import nltk; nltk.download('punkt', quiet=True)" """} if "example" in self.name else "echo Skipping"},
-            {"run": {
-                    "name": "Show installed libraries and their size",
-                    "command": """du -h -d 1 "$(pip -V | cut -d ' ' -f 4 | sed 's/pip//g')" | grep -vE "dist-info|_distutils_hack|__pycache__" | sort -h | tee installed.txt || true"""}
-            },
-            {"run": {
-                "name": "Show installed libraries and their versions",
-                "command": """pip list --format=freeze | tee installed.txt || true"""}
-            },
-            {"run": {
-                "name": "Show biggest libraries",
-                "command": """dpkg-query --show --showformat='${Installed-Size}\t${Package}\n' | sort -rh | head -25 | sort -h | awk '{ package=$2; sub(".*/", "", package); printf("%.5f GB %s\n", $1/1024/1024, package)}' || true"""}
-            },
-            {"run": {"name": "Create `test-results` directory", "command": "mkdir test-results"}},
-            {"run": {"name": "Get files to test", "command":f'curl -L -o {self.job_name}_test_list.txt <<pipeline.parameters.{self.job_name}_test_list>> --header "Circle-Token: $CIRCLE_TOKEN"' if self.name != "pr_documentation_tests" else 'echo "Skipped"'}},
-                        {"run": {"name": "Split tests across parallel nodes: show current parallel tests",
-                    "command": f"TESTS=$(circleci tests split  --split-by=timings {self.job_name}_test_list.txt) && echo $TESTS > splitted_tests.txt && echo $TESTS | tr ' ' '\n'" if self.parallelism else f"awk '{{printf \"%s \", $0}}' {self.job_name}_test_list.txt > splitted_tests.txt"
-                    }
-            },
-            {"run": {"name": "fetch hub objects before pytest", "command": "python3 utils/fetch_hub_objects_for_ci.py"}},
-            {"run": {
-                "name": "Run tests",
-                "command": f"({timeout_cmd} python3 -m pytest {marker_cmd} -n {self.pytest_num_workers} {junit_flags} {repeat_on_failure_flags} {' '.join(pytest_flags)} $(cat splitted_tests.txt) | tee tests_output.txt)"}
-            },
-            {"run": {"name": "Expand to show skipped tests", "when": "always", "command": f"python3 .circleci/parse_test_outputs.py --file tests_output.txt --skip"}},
-            {"run": {"name": "Failed tests: show reasons",   "when": "always", "command": f"python3 .circleci/parse_test_outputs.py --file tests_output.txt --fail"}},
-            {"run": {"name": "Errors",                       "when": "always", "command": f"python3 .circleci/parse_test_outputs.py --file tests_output.txt --errors"}},
-            {"store_test_results": {"path": "test-results"}},
-            {"store_artifacts": {"path": "test-results/junit.xml"}},
-            {"store_artifacts": {"path": "reports"}},
-            {"store_artifacts": {"path": "tests.txt"}},
-            {"store_artifacts": {"path": "splitted_tests.txt"}},
-            {"store_artifacts": {"path": "installed.txt"}},
-        ]
-        if self.parallelism:
-            job["parallelism"] = parallel
+
+        steps.append({"run": {"name": "Create `test-results` directory", "command": "mkdir test-results"}})
+        test_command = ""
+        if self.command_timeout:
+            test_command = f"timeout {self.command_timeout} "
+        # junit familiy xunit1 is necessary to support splitting on test name or class name with circleci split
+        test_command += f"python3 -m pytest -rsfE -p no:warnings -o junit_family=xunit1 --tb=short --junitxml=test-results/junit.xml -n {self.pytest_num_workers} " + " ".join(pytest_flags)
+
+        if self.parallelism == 1:
+            if self.tests_to_run is None:
+                test_command += " << pipeline.parameters.tests_to_run >>"
+            else:
+                test_command += " " + " ".join(self.tests_to_run)
+        else:
+            # We need explicit list instead of `pipeline.parameters.tests_to_run` (only available at job runtime)
+            tests = self.tests_to_run
+            if tests is None:
+                folder = os.environ["test_preparation_dir"]
+                test_file = os.path.join(folder, "filtered_test_list.txt")
+                if os.path.exists(test_file): # We take this job's tests from the filtered test_list.txt
+                    with open(test_file) as f:
+                        tests = f.read().split(" ")
+
+            # expand the test list
+            if tests == ["tests"]:
+                tests = [os.path.join("tests", x) for x in os.listdir("tests")]
+            expanded_tests = []
+            for test in tests:
+                if test.endswith(".py"):
+                    expanded_tests.append(test)
+                elif test == "tests/models":
+                    if "tokenization" in self.name:
+                        expanded_tests.extend(glob.glob("tests/models/**/test_tokenization*.py", recursive=True))
+                    elif self.name in ["flax","torch","tf"]:
+                        name = self.name if self.name != "torch" else ""
+                        if self.name == "torch":
+                            all_tests = glob.glob(f"tests/models/**/test_modeling_{name}*.py", recursive=True)
+                            filtered = [k for k in all_tests if ("_tf_") not in k and "_flax_" not in k]
+                            expanded_tests.extend(filtered)
+                        else:
+                            expanded_tests.extend(glob.glob(f"tests/models/**/test_modeling_{name}*.py", recursive=True))
+                    else:
+                        expanded_tests.extend(glob.glob("tests/models/**/test_modeling*.py", recursive=True))
+                elif test == "tests/pipelines":
+                    expanded_tests.extend(glob.glob("tests/models/**/test_modeling*.py", recursive=True))
+                else:
+                    expanded_tests.append(test)
+            tests = " ".join(expanded_tests)
+
+            # Each executor to run ~10 tests
+            n_executors = max(len(expanded_tests) // 10, 1)
+            # Avoid empty test list on some executor(s) or launching too many executors
+            if n_executors > self.parallelism:
+                n_executors = self.parallelism
+            job["parallelism"] = n_executors
+
+            # Need to be newline separated for the command `circleci tests split` below
+            command = f'echo {tests} | tr " " "\\n" >> tests.txt'
+            steps.append({"run": {"name": "Get tests", "command": command}})
+
+            command = 'TESTS=$(circleci tests split tests.txt) && echo $TESTS > splitted_tests.txt'
+            steps.append({"run": {"name": "Split tests", "command": command}})
+
+            steps.append({"store_artifacts": {"path": "tests.txt"}})
+            steps.append({"store_artifacts": {"path": "splitted_tests.txt"}})
+
+            test_command = ""
+            if self.command_timeout:
+                test_command = f"timeout {self.command_timeout} "
+            test_command += f"python3 -m pytest -rsfE -p no:warnings --tb=short  -o junit_family=xunit1 --junitxml=test-results/junit.xml -n {self.pytest_num_workers} " + " ".join(pytest_flags)
+            test_command += " $(cat splitted_tests.txt)"
+        if self.marker is not None:
+            test_command += f" -m {self.marker}"
+
+        if self.name == "pr_documentation_tests":
+            # can't use ` | tee tee tests_output.txt` as usual
+            test_command += " > tests_output.txt"
+            # Save the return code, so we can check if it is timeout in the next step.
+            test_command += '; touch "$?".txt'
+            # Never fail the test step for the doctest job. We will check the results in the next step, and fail that
+            # step instead if the actual test failures are found. This is to avoid the timeout being reported as test
+            # failure.
+            test_command = f"({test_command}) || true"
+        else:
+            test_command = f"({test_command} | tee tests_output.txt)"
+        steps.append({"run": {"name": "Run tests", "command": test_command}})
+
+        steps.append({"run": {"name": "Skipped tests", "when": "always", "command": f"python3 .circleci/parse_test_outputs.py --file tests_output.txt --skip"}})
+        steps.append({"run": {"name": "Failed tests",  "when": "always", "command": f"python3 .circleci/parse_test_outputs.py --file tests_output.txt --fail"}})
+        steps.append({"run": {"name": "Errors",        "when": "always", "command": f"python3 .circleci/parse_test_outputs.py --file tests_output.txt --errors"}})
+
+        steps.append({"store_test_results": {"path": "test-results"}})
+        steps.append({"store_artifacts": {"path": "tests_output.txt"}})
+        steps.append({"store_artifacts": {"path": "test-results/junit.xml"}})
+        steps.append({"store_artifacts": {"path": "reports"}})
+
         job["steps"] = steps
         return job
 
     @property
     def job_name(self):
-        return self.name if ("examples" in self.name or "pipeline" in self.name or "pr_documentation" in self.name) else f"tests_{self.name}"
+        return self.name if "examples" in self.name else f"tests_{self.name}"
 
 
 # JOBS
+torch_and_tf_job = CircleCIJob(
+    "torch_and_tf",
+    docker_image=[{"image":"huggingface/transformers-torch-tf-light"}],
+    install_steps=["uv venv && uv pip install ."],
+    additional_env={"RUN_PT_TF_CROSS_TESTS": True},
+    marker="is_pt_tf_cross_test",
+    pytest_options={"rA": None, "durations": 0},
+)
+
+
+torch_and_flax_job = CircleCIJob(
+    "torch_and_flax",
+    additional_env={"RUN_PT_FLAX_CROSS_TESTS": True},
+    docker_image=[{"image":"huggingface/transformers-torch-jax-light"}],
+    install_steps=["uv venv && uv pip install ."],
+    marker="is_pt_flax_cross_test",
+    pytest_options={"rA": None, "durations": 0},
+)
+
 torch_job = CircleCIJob(
     "torch",
     docker_image=[{"image": "huggingface/transformers-torch-light"}],
-    marker="not generate",
-    parallelism=6,
-)
-
-generate_job = CircleCIJob(
-    "generate",
-    docker_image=[{"image": "huggingface/transformers-torch-light"}],
-    # networkx==3.3 (after #36957) cause some issues
-    # TODO: remove this once it works directly
     install_steps=["uv venv && uv pip install ."],
-    marker="generate",
     parallelism=6,
+    pytest_num_workers=4
 )
 
 tokenization_job = CircleCIJob(
     "tokenization",
     docker_image=[{"image": "huggingface/transformers-torch-light"}],
-    parallelism=8,
+    install_steps=["uv venv && uv pip install ."],
+    parallelism=6,
+    pytest_num_workers=4
 )
 
-processor_job = CircleCIJob(
-    "processors",
-    docker_image=[{"image": "huggingface/transformers-torch-light"}],
-    parallelism=8,
+
+tf_job = CircleCIJob(
+    "tf",
+    docker_image=[{"image":"huggingface/transformers-tf-light"}],
+    install_steps=["uv venv", "uv pip install -e."],
+    parallelism=6,
+    pytest_num_workers=4,
 )
 
+
+flax_job = CircleCIJob(
+    "flax",
+    docker_image=[{"image":"huggingface/transformers-jax-light"}],
+    install_steps=["uv venv && uv pip install ."],
+    parallelism=6,
+    pytest_num_workers=4
+)
+
+
 pipelines_torch_job = CircleCIJob(
     "pipelines_torch",
     additional_env={"RUN_PIPELINE_TESTS": True},
     docker_image=[{"image":"huggingface/transformers-torch-light"}],
+    install_steps=["uv venv && uv pip install ."],
     marker="is_pipeline_test",
-    parallelism=4,
 )
 
+
+pipelines_tf_job = CircleCIJob(
+    "pipelines_tf",
+    additional_env={"RUN_PIPELINE_TESTS": True},
+    docker_image=[{"image":"huggingface/transformers-tf-light"}],
+    install_steps=["uv venv && uv pip install ."],
+    marker="is_pipeline_test",
+)
+
+
 custom_tokenizers_job = CircleCIJob(
     "custom_tokenizers",
     additional_env={"RUN_CUSTOM_TOKENIZERS": True},
     docker_image=[{"image": "huggingface/transformers-custom-tokenizers"}],
+    install_steps=["uv venv","uv pip install -e ."],
+    parallelism=None,
+    resource_class=None,
+    tests_to_run=[
+        "./tests/models/bert_japanese/test_tokenization_bert_japanese.py",
+        "./tests/models/openai/test_tokenization_openai.py",
+        "./tests/models/clip/test_tokenization_clip.py",
+    ],
 )
 
 
 examples_torch_job = CircleCIJob(
     "examples_torch",
     additional_env={"OMP_NUM_THREADS": 8},
+    cache_name="torch_examples",
     docker_image=[{"image":"huggingface/transformers-examples-torch"}],
     # TODO @ArthurZucker remove this once docker is easier to build
     install_steps=["uv venv && uv pip install . && uv pip install -r examples/pytorch/_tests_requirements.txt"],
-    pytest_num_workers=4,
+    pytest_num_workers=1,
 )
 
+
+examples_tensorflow_job = CircleCIJob(
+    "examples_tensorflow",
+    cache_name="tensorflow_examples",
+    docker_image=[{"image":"huggingface/transformers-examples-tf"}],
+    install_steps=["uv venv && uv pip install . && uv pip install -r examples/tensorflow/_tests_requirements.txt"],
+    parallelism=8
+)
+
+
 hub_job = CircleCIJob(
     "hub",
     additional_env={"HUGGINGFACE_CO_STAGING": True},
     docker_image=[{"image":"huggingface/transformers-torch-light"}],
     install_steps=[
-        'uv venv && uv pip install .',
+        "uv venv && uv pip install .",
         'git config --global user.email "ci@dummy.com"',
         'git config --global user.name "ci"',
     ],
     marker="is_staging_test",
-    pytest_num_workers=2,
-    resource_class="medium",
+    pytest_num_workers=1,
 )
 
 
@@ -273,18 +349,27 @@ onnx_job = CircleCIJob(
     "onnx",
     docker_image=[{"image":"huggingface/transformers-torch-tf-light"}],
     install_steps=[
-        "uv venv",
-        "uv pip install .[testing,sentencepiece,onnxruntime,vision,rjieba]",
+        "uv venv && uv pip install .",
+        "uv pip install --upgrade eager pip",
+        "uv pip install .[torch,tf,testing,sentencepiece,onnxruntime,vision,rjieba]",
     ],
     pytest_options={"k onnx": None},
     pytest_num_workers=1,
-    resource_class="small",
 )
 
 
 exotic_models_job = CircleCIJob(
     "exotic_models",
+    install_steps=["uv venv && uv pip install ."],
     docker_image=[{"image":"huggingface/transformers-exotic-models"}],
+    tests_to_run=[
+        "tests/models/*layoutlmv*",
+        "tests/models/*nat",
+        "tests/models/deta",
+        "tests/models/udop",
+        "tests/models/nougat",
+    ],
+    pytest_num_workers=12,
     parallelism=4,
     pytest_options={"durations": 100},
 )
@@ -293,19 +378,11 @@ exotic_models_job = CircleCIJob(
 repo_utils_job = CircleCIJob(
     "repo_utils",
     docker_image=[{"image":"huggingface/transformers-consistency"}],
-    pytest_num_workers=4,
-    resource_class="large",
-)
-
-
-non_model_job = CircleCIJob(
-    "non_model",
-    docker_image=[{"image": "huggingface/transformers-torch-light"}],
-    # networkx==3.3 (after #36957) cause some issues
-    # TODO: remove this once it works directly
     install_steps=["uv venv && uv pip install ."],
-    marker="not generate",
-    parallelism=6,
+    parallelism=None,
+    pytest_num_workers=1,
+    resource_class="large",
+    tests_to_run="tests/repo_utils",
 )
 
 
@@ -314,18 +391,28 @@ non_model_job = CircleCIJob(
 # the bash output redirection.)
 py_command = 'from utils.tests_fetcher import get_doctest_files; to_test = get_doctest_files() + ["dummy.py"]; to_test = " ".join(to_test); print(to_test)'
 py_command = f"$(python3 -c '{py_command}')"
-command = f'echo """{py_command}""" > pr_documentation_tests_temp.txt'
+command = f'echo "{py_command}" > pr_documentation_tests_temp.txt'
 doc_test_job = CircleCIJob(
     "pr_documentation_tests",
     docker_image=[{"image":"huggingface/transformers-consistency"}],
     additional_env={"TRANSFORMERS_VERBOSITY": "error", "DATASETS_VERBOSITY": "error", "SKIP_CUDA_DOCTEST": "1"},
     install_steps=[
         # Add an empty file to keep the test step running correctly even no file is selected to be tested.
-        "uv venv && pip install .",
         "touch dummy.py",
-        command,
-        "cat pr_documentation_tests_temp.txt",
-        "tail -n1 pr_documentation_tests_temp.txt | tee pr_documentation_tests_test_list.txt"
+        {
+            "name": "Get files to test",
+            "command": command,
+        },
+        {
+            "name": "Show information in `Get files to test`",
+            "command":
+                "cat pr_documentation_tests_temp.txt"
+        },
+        {
+            "name": "Get the last line in `pr_documentation_tests.txt`",
+            "command":
+                "tail -n1 pr_documentation_tests_temp.txt | tee pr_documentation_tests.txt"
+        },
     ],
     tests_to_run="$(cat pr_documentation_tests.txt)",  # noqa
     pytest_options={"-doctest-modules": None, "doctest-glob": "*.md", "dist": "loadfile", "rvsA": None},
@@ -333,54 +420,121 @@ doc_test_job = CircleCIJob(
     pytest_num_workers=1,
 )
 
-REGULAR_TESTS = [torch_job, hub_job, onnx_job, tokenization_job, processor_job, generate_job, non_model_job] # fmt: skip
-EXAMPLES_TESTS = [examples_torch_job]
-PIPELINE_TESTS = [pipelines_torch_job]
+REGULAR_TESTS = [
+    torch_and_tf_job,
+    torch_and_flax_job,
+    torch_job,
+    tf_job,
+    flax_job,
+    custom_tokenizers_job,
+    hub_job,
+    onnx_job,
+    exotic_models_job,
+    tokenization_job
+]
+EXAMPLES_TESTS = [
+    examples_torch_job,
+    examples_tensorflow_job,
+]
+PIPELINE_TESTS = [
+    pipelines_torch_job,
+    pipelines_tf_job,
+]
 REPO_UTIL_TESTS = [repo_utils_job]
 DOC_TESTS = [doc_test_job]
-ALL_TESTS = REGULAR_TESTS + EXAMPLES_TESTS + PIPELINE_TESTS + REPO_UTIL_TESTS + DOC_TESTS + [custom_tokenizers_job] + [exotic_models_job]  # fmt: skip
 
 
 def create_circleci_config(folder=None):
     if folder is None:
         folder = os.getcwd()
+    # Used in CircleCIJob.to_dict() to expand the test list (for using parallelism)
     os.environ["test_preparation_dir"] = folder
-    jobs = [k for k in ALL_TESTS if os.path.isfile(os.path.join("test_preparation" , f"{k.job_name}_test_list.txt") )]
-    print("The following jobs will be run ", jobs)
+    jobs = []
+    all_test_file = os.path.join(folder, "test_list.txt")
+    if os.path.exists(all_test_file):
+        with open(all_test_file) as f:
+            all_test_list = f.read()
+    else:
+        all_test_list = []
+    if len(all_test_list) > 0:
+        jobs.extend(PIPELINE_TESTS)
+
+    test_file = os.path.join(folder, "filtered_test_list.txt")
+    if os.path.exists(test_file):
+        with open(test_file) as f:
+            test_list = f.read()
+    else:
+        test_list = []
+    if len(test_list) > 0:
+        jobs.extend(REGULAR_TESTS)
+
+        extended_tests_to_run = set(test_list.split())
+        # Extend the test files for cross test jobs
+        for job in jobs:
+            if job.job_name in ["tests_torch_and_tf", "tests_torch_and_flax"]:
+                for test_path in copy.copy(extended_tests_to_run):
+                    dir_path, fn = os.path.split(test_path)
+                    if fn.startswith("test_modeling_tf_"):
+                        fn = fn.replace("test_modeling_tf_", "test_modeling_")
+                    elif fn.startswith("test_modeling_flax_"):
+                        fn = fn.replace("test_modeling_flax_", "test_modeling_")
+                    else:
+                        if job.job_name == "test_torch_and_tf":
+                            fn = fn.replace("test_modeling_", "test_modeling_tf_")
+                        elif job.job_name == "test_torch_and_flax":
+                            fn = fn.replace("test_modeling_", "test_modeling_flax_")
+                    new_test_file = str(os.path.join(dir_path, fn))
+                    if os.path.isfile(new_test_file):
+                        if new_test_file not in extended_tests_to_run:
+                            extended_tests_to_run.add(new_test_file)
+        extended_tests_to_run = sorted(extended_tests_to_run)
+        for job in jobs:
+            if job.job_name in ["tests_torch_and_tf", "tests_torch_and_flax"]:
+                job.tests_to_run = extended_tests_to_run
+        fn = "filtered_test_list_cross_tests.txt"
+        f_path = os.path.join(folder, fn)
+        with open(f_path, "w") as fp:
+            fp.write(" ".join(extended_tests_to_run))
+
+    example_file = os.path.join(folder, "examples_test_list.txt")
+    if os.path.exists(example_file) and os.path.getsize(example_file) > 0:
+        with open(example_file, "r", encoding="utf-8") as f:
+            example_tests = f.read()
+        for job in EXAMPLES_TESTS:
+            framework = job.name.replace("examples_", "").replace("torch", "pytorch")
+            if example_tests == "all":
+                job.tests_to_run = [f"examples/{framework}"]
+            else:
+                job.tests_to_run = [f for f in example_tests.split(" ") if f.startswith(f"examples/{framework}")]
+
+            if len(job.tests_to_run) > 0:
+                jobs.append(job)
+
+    doctest_file = os.path.join(folder, "doctest_list.txt")
+    if os.path.exists(doctest_file):
+        with open(doctest_file) as f:
+            doctest_list = f.read()
+    else:
+        doctest_list = []
+    if len(doctest_list) > 0:
+        jobs.extend(DOC_TESTS)
+
+    repo_util_file = os.path.join(folder, "test_repo_utils.txt")
+    if os.path.exists(repo_util_file) and os.path.getsize(repo_util_file) > 0:
+        jobs.extend(REPO_UTIL_TESTS)
 
     if len(jobs) == 0:
         jobs = [EmptyJob()]
-    else:
-        print("Full list of job name inputs", {j.job_name + "_test_list":{"type":"string", "default":''} for j in jobs})
-        # Add a job waiting all the test jobs and aggregate their test summary files at the end
-        collection_job = EmptyJob()
-        collection_job.job_name = "collection_job"
-        jobs = [collection_job] + jobs
-
-    config = {
-        "version": "2.1",
-        "parameters": {
-            # Only used to accept the parameters from the trigger
-            "nightly": {"type": "boolean", "default": False},
-            # Only used to accept the parameters from GitHub Actions trigger
-            "GHA_Actor": {"type": "string", "default": ""},
-            "GHA_Action": {"type": "string", "default": ""},
-            "GHA_Event": {"type": "string", "default": ""},
-            "GHA_Meta": {"type": "string", "default": ""},
-            "tests_to_run": {"type": "string", "default": ""},
-            **{j.job_name + "_test_list":{"type":"string", "default":''} for j in jobs},
-            **{j.job_name + "_parallelism":{"type":"integer", "default":1} for j in jobs},
-        },
-        "jobs": {j.job_name: j.to_dict() for j in jobs}
+    config = {"version": "2.1"}
+    config["parameters"] = {
+        # Only used to accept the parameters from the trigger
+        "nightly": {"type": "boolean", "default": False},
+        "tests_to_run": {"type": "string", "default": test_list},
     }
-    if "CIRCLE_TOKEN" in os.environ:
-        # For private forked repo. (e.g. new model addition)
-        config["workflows"] = {"version": 2, "run_tests": {"jobs": [{j.job_name: {"context": ["TRANSFORMERS_CONTEXT"]}} for j in jobs]}}
-    else:
-        # For public repo. (e.g. `transformers`)
-        config["workflows"] = {"version": 2, "run_tests": {"jobs": [j.job_name for j in jobs]}}
+    config["jobs"] = {j.job_name: j.to_dict() for j in jobs}
+    config["workflows"] = {"version": 2, "run_tests": {"jobs": [j.job_name for j in jobs]}}
     with open(os.path.join(folder, "generated_config.yml"), "w") as f:
-        f.write(yaml.dump(config, sort_keys=False, default_flow_style=False).replace("' << pipeline", " << pipeline").replace(">> '", " >>"))
+        f.write(yaml.dump(config, indent=2, width=1000000, sort_keys=False))
 
 
 if __name__ == "__main__":

.circleci/parse_test_outputs.py

@@ -67,4 +67,4 @@ def main():
 
 
 if __name__ == "__main__":
-    main()
+    main()

.coveragerc

@@ -0,0 +1,12 @@
+[run]
+source=transformers
+omit =
+    # skip convertion scripts from testing for now
+    */convert_*
+    */__main__.py
+[report]
+exclude_lines =
+    pragma: no cover
+    raise
+    except
+    register_parameter

.github/ISSUE_TEMPLATE/bug-report.yml

@@ -16,7 +16,7 @@ body:
     id: system-info
     attributes:
       label: System Info
-      description: Please share your system info with us. You can run the command `transformers env` and copy-paste its output below.
+      description: Please share your system info with us. You can run the command `transformers-cli env` and copy-paste its output below.
       placeholder: transformers version, platform, python version, ...
     validations:
       required: true
@@ -37,31 +37,25 @@ body:
         Models:
 
           - text models: @ArthurZucker
-          - vision models: @amyeroberts, @qubvel
-          - speech models: @eustlb
+          - vision models: @amyeroberts
+          - speech models: @sanchit-gandhi
           - graph models: @clefourrier
 
         Library:
 
-          - flax: @gante and @Rocketknight1
+          - flax: @sanchit-gandhi
           - generate: @zucchini-nlp (visual-language models) or @gante (all others)
-          - pipelines: @Rocketknight1
+          - pipelines: @Narsil
           - tensorflow: @gante and @Rocketknight1
-          - tokenizers: @ArthurZucker and @itazap
-          - trainer: @zach-huggingface @SunMarc
+          - tokenizers: @ArthurZucker
+          - trainer: @muellerzr @SunMarc
 
         Integrations:
 
-          - deepspeed: HF Trainer/Accelerate: @SunMarc @zach-huggingface
+          - deepspeed: HF Trainer/Accelerate: @muellerzr
           - ray/raytune: @richardliaw, @amogkam
           - Big Model Inference: @SunMarc
-          - quantization (bitsandbytes, autogpt): @SunMarc @MekkCyber
-        
-        Devices/Backends:
-        
-          - AMD ROCm: @ivarflakstad
-          - Intel XPU: @IlyasMoutawwakil
-          - Ascend NPU: @ivarflakstad 
+          - quantization (bitsandbytes, autogpt): @SunMarc
 
         Documentation: @stevhliu
 
@@ -78,7 +72,7 @@ body:
 
         Maintained examples (not research project or legacy):
 
-          - Flax: @Rocketknight1
+          - Flax: @sanchit-gandhi
           - PyTorch: See Models above and tag the person corresponding to the modality of the example.
           - TensorFlow: @Rocketknight1
 
@@ -112,7 +106,6 @@ body:
       label: Reproduction
       description: |
         Please provide a code sample that reproduces the problem you ran into. It can be a Colab link or just a code snippet.
-        Please include relevant config information with your code, for example your Trainers, TRL, Peft, and DeepSpeed configs.
         If you have code snippets, error messages, stack traces please provide them here as well.
         Important! Use code tags to correctly format your code. See https://help.github.com/en/github/writing-on-github/creating-and-highlighting-code-blocks#syntax-highlighting
         Do not use screenshots, as they are hard to read and (more importantly) don't allow others to copy-and-paste your code.

.github/ISSUE_TEMPLATE/i18n.md

@@ -23,7 +23,7 @@ Some notes:
 * Please translate in a gender-neutral way.
 * Add your translations to the folder called `<languageCode>` inside the [source folder](https://github.com/huggingface/transformers/tree/main/docs/source).
 * Register your translation in `<languageCode>/_toctree.yml`; please follow the order of the [English version](https://github.com/huggingface/transformers/blob/main/docs/source/en/_toctree.yml).
-* Once you're finished, open a pull request and tag this issue by including #issue-number in the description, where issue-number is the number of this issue. Please ping @stevhliu for review.
+* Once you're finished, open a pull request and tag this issue by including #issue-number in the description, where issue-number is the number of this issue. Please ping @stevhliu and @MKhalusova for review.
 * 🙋 If you'd like others to help you with the translation, you can also post in the 🤗 [forums](https://discuss.huggingface.co/).
 
 ## Get Started section
@@ -34,7 +34,7 @@ Some notes:
 
 ## Tutorial section
 - [ ] [pipeline_tutorial.md](https://github.com/huggingface/transformers/blob/main/docs/source/en/pipeline_tutorial.md)
-- [ ]  [autoclass_tutorial.md](https://github.com/huggingface/transformers/blob/main/docs/source/en/autoclass_tutorial.md)
+- [ ]  [autoclass_tutorial.md](https://github.com/huggingface/transformers/blob/master/docs/source/autoclass_tutorial.md)
 - [ ]  [preprocessing.md](https://github.com/huggingface/transformers/blob/main/docs/source/en/preprocessing.md)
 - [ ]  [training.md](https://github.com/huggingface/transformers/blob/main/docs/source/en/training.md)
 - [ ]  [accelerate.md](https://github.com/huggingface/transformers/blob/main/docs/source/en/accelerate.md)

.github/ISSUE_TEMPLATE/migration.yml

@@ -6,7 +6,7 @@ body:
     id: system-info
     attributes:
       label: System Info
-      description: Please share your system info with us. You can run the command `transformers env` and copy-paste its output below.
+      description: Please share your system info with us. You can run the command `transformers-cli env` and copy-paste its output below.
       render: shell
       placeholder: transformers version, platform, python version, ...
     validations:

.github/PULL_REQUEST_TEMPLATE.md

@@ -40,26 +40,25 @@ members/contributors who may be interested in your PR.
 Models:
 
 - text models: @ArthurZucker
-- vision models: @amyeroberts, @qubvel
-- speech models: @eustlb
+- vision models: @amyeroberts
+- speech models: @sanchit-gandhi
 - graph models: @clefourrier
 
 Library:
 
-- flax: @gante and @Rocketknight1
+- flax: @sanchit-gandhi
 - generate: @zucchini-nlp (visual-language models) or @gante (all others)
-- pipelines: @Rocketknight1
+- pipelines: @Narsil
 - tensorflow: @gante and @Rocketknight1
 - tokenizers: @ArthurZucker
-- trainer: @zach-huggingface, @SunMarc and @qgallouedec
-- chat templates: @Rocketknight1
+- trainer: @muellerzr and @SunMarc
 
 Integrations:
 
-- deepspeed: HF Trainer/Accelerate: @SunMarc @zach-huggingface
+- deepspeed: HF Trainer/Accelerate: @muellerzr
 - ray/raytune: @richardliaw, @amogkam
 - Big Model Inference: @SunMarc
-- quantization (bitsandbytes, autogpt): @SunMarc @MekkCyber
+- quantization (bitsandbytes, autogpt): @SunMarc
 
 Documentation: @stevhliu
 
@@ -72,7 +71,7 @@ HF projects:
 
 Maintained examples (not research project or legacy):
 
-- Flax: @Rocketknight1
+- Flax: @sanchit-gandhi
 - PyTorch: See Models above and tag the person corresponding to the modality of the example.
 - TensorFlow: @Rocketknight1
 

.github/scripts/assign_reviewers.py

@@ -1,120 +0,0 @@
-# coding=utf-8
-# Copyright 2025 the HuggingFace Inc. team. All rights reserved.
-#
-# Licensed under the Apache License, Version 2.0 (the "License");
-# you may not use this file except in compliance with the License.
-# You may obtain a copy of the License at
-#
-#     http://www.apache.org/licenses/LICENSE-2.0
-#
-# Unless required by applicable law or agreed to in writing, software
-# distributed under the License is distributed on an "AS IS" BASIS,
-# WITHOUT WARRANTIES OR CONDITIONS OF ANY KIND, either express or implied.
-# See the License for the specific language governing permissions and
-# limitations under the License.
-
-import os
-import github
-import json
-from github import Github
-import re
-from collections import Counter
-from pathlib import Path
-
-def pattern_to_regex(pattern):
-    if pattern.startswith("/"):
-        start_anchor = True
-        pattern = re.escape(pattern[1:])
-    else:
-        start_anchor = False
-        pattern = re.escape(pattern)
-    # Replace `*` with "any number of non-slash characters"
-    pattern = pattern.replace(r"\*", "[^/]*")
-    if start_anchor:
-        pattern = r"^\/?" + pattern  # Allow an optional leading slash after the start of the string
-    return pattern
-
-def get_file_owners(file_path, codeowners_lines):
-    # Process lines in reverse (last matching pattern takes precedence)
-    for line in reversed(codeowners_lines):
-        # Skip comments and empty lines, strip inline comments
-        line = line.split('#')[0].strip()
-        if not line:
-            continue
-
-        # Split into pattern and owners
-        parts = line.split()
-        pattern = parts[0]
-        # Can be empty, e.g. for dummy files with explicitly no owner!
-        owners = [owner.removeprefix("@") for owner in parts[1:]]
-
-        # Check if file matches pattern
-        file_regex = pattern_to_regex(pattern)
-        if re.search(file_regex, file_path) is not None:
-            return owners  # Remember, can still be empty!
-    return []  # Should never happen, but just in case
-
-def pr_author_is_in_hf(pr_author, codeowners_lines):
-    # Check if the PR author is in the codeowners file
-    for line in codeowners_lines:
-        line = line.split('#')[0].strip()
-        if not line:
-            continue
-
-        # Split into pattern and owners
-        parts = line.split()
-        owners = [owner.removeprefix("@") for owner in parts[1:]]
-
-        if pr_author in owners:
-            return True
-    return False
-
-def main():
-    script_dir = Path(__file__).parent.absolute()
-    with open(script_dir / "codeowners_for_review_action") as f:
-        codeowners_lines = f.readlines()
-
-    g = Github(os.environ['GITHUB_TOKEN'])
-    repo = g.get_repo("huggingface/transformers")
-    with open(os.environ['GITHUB_EVENT_PATH']) as f:
-        event = json.load(f)
-
-    # The PR number is available in the event payload
-    pr_number = event['pull_request']['number']
-    pr = repo.get_pull(pr_number)
-    pr_author = pr.user.login
-    if pr_author_is_in_hf(pr_author, codeowners_lines):
-        print(f"PR author {pr_author} is in codeowners, skipping review request.")
-        return
-
-    existing_reviews = list(pr.get_reviews())
-    if existing_reviews:
-        print(f"Already has reviews: {[r.user.login for r in existing_reviews]}")
-        return
-
-    users_requested, teams_requested = pr.get_review_requests()
-    users_requested = list(users_requested)
-    if users_requested:
-        print(f"Reviewers already requested: {users_requested}")
-        return
-
-    locs_per_owner = Counter()
-    for file in pr.get_files():
-        owners = get_file_owners(file.filename, codeowners_lines)
-        for owner in owners:
-            locs_per_owner[owner] += file.changes
-
-    # Assign the top 2 based on locs changed as reviewers, but skip the owner if present
-    locs_per_owner.pop(pr_author, None)
-    top_owners = locs_per_owner.most_common(2)
-    print("Top owners", top_owners)
-    top_owners = [owner[0] for owner in top_owners]
-    try:
-        pr.create_review_request(top_owners)
-    except github.GithubException as e:
-        print(f"Failed to request review for {top_owners}: {e}")
-
-
-
-if __name__ == "__main__":
-    main()

.github/scripts/codeowners_for_review_action

@@ -1,370 +0,0 @@
-# Top-level rules are matched only if nothing else matches
-* @Rocketknight1 @ArthurZucker # if no one is pinged based on the other rules, he will do the dispatch
-*.md @stevhliu
-*tokenization* @ArthurZucker
-docs/ @stevhliu
-/benchmark/ @McPatate
-/docker/ @ydshieh @ArthurZucker
-
-# More high-level globs catch cases when specific rules later don't apply
-/src/transformers/models/*/processing* @molbap @yonigozlan @qubvel
-/src/transformers/models/*/image_processing* @qubvel
-/src/transformers/models/*/image_processing_*_fast* @yonigozlan
-
-# Owners of subsections of the library
-/src/transformers/generation/ @gante
-/src/transformers/pipeline/ @Rocketknight1 @yonigozlan
-/src/transformers/integrations/ @SunMarc @MekkCyber @zach-huggingface
-/src/transformers/quantizers/ @SunMarc @MekkCyber
-tests/ @ydshieh
-tests/generation/ @gante
-
-/src/transformers/models/auto/ @ArthurZucker
-/src/transformers/utils/ @ArthurZucker @Rocketknight1
-/src/transformers/loss/ @ArthurZucker
-/src/transformers/onnx/ @michaelbenayoun
-
-# Specific files come after the sections/globs, so they take priority
-/.circleci/config.yml @ArthurZucker @ydshieh
-/utils/tests_fetcher.py @ydshieh
-trainer.py @zach-huggingface @SunMarc
-trainer_utils.py @zach-huggingface @SunMarc
-/utils/modular_model_converter.py @Cyrilvallez @ArthurZucker
-
-# Owners of individual models are specific / high priority, and so they come last
-# mod* captures modeling and modular files
-
-# Text models
-/src/transformers/models/albert/mod*_albert* @ArthurZucker
-/src/transformers/models/bamba/mod*_bamba* @ArthurZucker
-/src/transformers/models/bart/mod*_bart* @ArthurZucker
-/src/transformers/models/barthez/mod*_barthez* @ArthurZucker
-/src/transformers/models/bartpho/mod*_bartpho* @ArthurZucker
-/src/transformers/models/bert/mod*_bert* @ArthurZucker
-/src/transformers/models/bert_generation/mod*_bert_generation* @ArthurZucker
-/src/transformers/models/bert_japanese/mod*_bert_japanese* @ArthurZucker
-/src/transformers/models/bertweet/mod*_bertweet* @ArthurZucker
-/src/transformers/models/big_bird/mod*_big_bird* @ArthurZucker
-/src/transformers/models/bigbird_pegasus/mod*_bigbird_pegasus* @ArthurZucker
-/src/transformers/models/biogpt/mod*_biogpt* @ArthurZucker
-/src/transformers/models/blenderbot/mod*_blenderbot* @ArthurZucker
-/src/transformers/models/blenderbot_small/mod*_blenderbot_small* @ArthurZucker
-/src/transformers/models/bloom/mod*_bloom* @ArthurZucker
-/src/transformers/models/bort/mod*_bort* @ArthurZucker
-/src/transformers/models/byt5/mod*_byt5* @ArthurZucker
-/src/transformers/models/camembert/mod*_camembert* @ArthurZucker
-/src/transformers/models/canine/mod*_canine* @ArthurZucker
-/src/transformers/models/codegen/mod*_codegen* @ArthurZucker
-/src/transformers/models/code_llama/mod*_code_llama* @ArthurZucker
-/src/transformers/models/cohere/mod*_cohere* @ArthurZucker
-/src/transformers/models/cohere2/mod*_cohere2* @ArthurZucker
-/src/transformers/models/convbert/mod*_convbert* @ArthurZucker
-/src/transformers/models/cpm/mod*_cpm* @ArthurZucker
-/src/transformers/models/cpmant/mod*_cpmant* @ArthurZucker
-/src/transformers/models/ctrl/mod*_ctrl* @ArthurZucker
-/src/transformers/models/dbrx/mod*_dbrx* @ArthurZucker
-/src/transformers/models/deberta/mod*_deberta* @ArthurZucker
-/src/transformers/models/deberta_v2/mod*_deberta_v2* @ArthurZucker
-/src/transformers/models/dialogpt/mod*_dialogpt* @ArthurZucker
-/src/transformers/models/diffllama/mod*_diffllama* @ArthurZucker
-/src/transformers/models/distilbert/mod*_distilbert* @ArthurZucker
-/src/transformers/models/dpr/mod*_dpr* @ArthurZucker
-/src/transformers/models/electra/mod*_electra* @ArthurZucker
-/src/transformers/models/encoder_decoder/mod*_encoder_decoder* @ArthurZucker
-/src/transformers/models/ernie/mod*_ernie* @ArthurZucker
-/src/transformers/models/ernie_m/mod*_ernie_m* @ArthurZucker
-/src/transformers/models/esm/mod*_esm* @ArthurZucker
-/src/transformers/models/falcon/mod*_falcon* @ArthurZucker
-/src/transformers/models/falcon3/mod*_falcon3* @ArthurZucker
-/src/transformers/models/falcon_mamba/mod*_falcon_mamba* @ArthurZucker
-/src/transformers/models/fastspeech2_conformer/mod*_fastspeech2_conformer* @ArthurZucker
-/src/transformers/models/flan_t5/mod*_flan_t5* @ArthurZucker
-/src/transformers/models/flan_ul2/mod*_flan_ul2* @ArthurZucker
-/src/transformers/models/flaubert/mod*_flaubert* @ArthurZucker
-/src/transformers/models/fnet/mod*_fnet* @ArthurZucker
-/src/transformers/models/fsmt/mod*_fsmt* @ArthurZucker
-/src/transformers/models/funnel/mod*_funnel* @ArthurZucker
-/src/transformers/models/fuyu/mod*_fuyu* @ArthurZucker
-/src/transformers/models/gemma/mod*_gemma* @ArthurZucker
-/src/transformers/models/gemma2/mod*_gemma2* @ArthurZucker
-/src/transformers/models/glm/mod*_glm* @ArthurZucker
-/src/transformers/models/openai_gpt/mod*_openai_gpt* @ArthurZucker
-/src/transformers/models/gpt_neo/mod*_gpt_neo* @ArthurZucker
-/src/transformers/models/gpt_neox/mod*_gpt_neox* @ArthurZucker
-/src/transformers/models/gpt_neox_japanese/mod*_gpt_neox_japanese* @ArthurZucker
-/src/transformers/models/gptj/mod*_gptj* @ArthurZucker
-/src/transformers/models/gpt2/mod*_gpt2* @ArthurZucker
-/src/transformers/models/gpt_bigcode/mod*_gpt_bigcode* @ArthurZucker
-/src/transformers/models/gptsan_japanese/mod*_gptsan_japanese* @ArthurZucker
-/src/transformers/models/gpt_sw3/mod*_gpt_sw3* @ArthurZucker
-/src/transformers/models/granite/mod*_granite* @ArthurZucker
-/src/transformers/models/granitemoe/mod*_granitemoe* @ArthurZucker
-/src/transformers/models/herbert/mod*_herbert* @ArthurZucker
-/src/transformers/models/ibert/mod*_ibert* @ArthurZucker
-/src/transformers/models/jamba/mod*_jamba* @ArthurZucker
-/src/transformers/models/jetmoe/mod*_jetmoe* @ArthurZucker
-/src/transformers/models/jukebox/mod*_jukebox* @ArthurZucker
-/src/transformers/models/led/mod*_led* @ArthurZucker
-/src/transformers/models/llama/mod*_llama* @ArthurZucker @Cyrilvallez
-/src/transformers/models/longformer/mod*_longformer* @ArthurZucker
-/src/transformers/models/longt5/mod*_longt5* @ArthurZucker
-/src/transformers/models/luke/mod*_luke* @ArthurZucker
-/src/transformers/models/m2m_100/mod*_m2m_100* @ArthurZucker
-/src/transformers/models/madlad_400/mod*_madlad_400* @ArthurZucker
-/src/transformers/models/mamba/mod*_mamba* @ArthurZucker
-/src/transformers/models/mamba2/mod*_mamba2* @ArthurZucker
-/src/transformers/models/marian/mod*_marian* @ArthurZucker
-/src/transformers/models/markuplm/mod*_markuplm* @ArthurZucker
-/src/transformers/models/mbart/mod*_mbart* @ArthurZucker
-/src/transformers/models/mega/mod*_mega* @ArthurZucker
-/src/transformers/models/megatron_bert/mod*_megatron_bert* @ArthurZucker
-/src/transformers/models/megatron_gpt2/mod*_megatron_gpt2* @ArthurZucker
-/src/transformers/models/mistral/mod*_mistral* @ArthurZucker
-/src/transformers/models/mixtral/mod*_mixtral* @ArthurZucker
-/src/transformers/models/mluke/mod*_mluke* @ArthurZucker
-/src/transformers/models/mobilebert/mod*_mobilebert* @ArthurZucker
-/src/transformers/models/modernbert/mod*_modernbert* @ArthurZucker
-/src/transformers/models/mpnet/mod*_mpnet* @ArthurZucker
-/src/transformers/models/mpt/mod*_mpt* @ArthurZucker
-/src/transformers/models/mra/mod*_mra* @ArthurZucker
-/src/transformers/models/mt5/mod*_mt5* @ArthurZucker
-/src/transformers/models/mvp/mod*_mvp* @ArthurZucker
-/src/transformers/models/myt5/mod*_myt5* @ArthurZucker
-/src/transformers/models/nemotron/mod*_nemotron* @ArthurZucker
-/src/transformers/models/nezha/mod*_nezha* @ArthurZucker
-/src/transformers/models/nllb/mod*_nllb* @ArthurZucker
-/src/transformers/models/nllb_moe/mod*_nllb_moe* @ArthurZucker
-/src/transformers/models/nystromformer/mod*_nystromformer* @ArthurZucker
-/src/transformers/models/olmo/mod*_olmo* @ArthurZucker
-/src/transformers/models/olmo2/mod*_olmo2* @ArthurZucker
-/src/transformers/models/olmoe/mod*_olmoe* @ArthurZucker
-/src/transformers/models/open_llama/mod*_open_llama* @ArthurZucker
-/src/transformers/models/opt/mod*_opt* @ArthurZucker
-/src/transformers/models/pegasus/mod*_pegasus* @ArthurZucker
-/src/transformers/models/pegasus_x/mod*_pegasus_x* @ArthurZucker
-/src/transformers/models/persimmon/mod*_persimmon* @ArthurZucker
-/src/transformers/models/phi/mod*_phi* @ArthurZucker
-/src/transformers/models/phi3/mod*_phi3* @ArthurZucker
-/src/transformers/models/phimoe/mod*_phimoe* @ArthurZucker
-/src/transformers/models/phobert/mod*_phobert* @ArthurZucker
-/src/transformers/models/plbart/mod*_plbart* @ArthurZucker
-/src/transformers/models/prophetnet/mod*_prophetnet* @ArthurZucker
-/src/transformers/models/qdqbert/mod*_qdqbert* @ArthurZucker
-/src/transformers/models/qwen2/mod*_qwen2* @ArthurZucker
-/src/transformers/models/qwen2_moe/mod*_qwen2_moe* @ArthurZucker
-/src/transformers/models/rag/mod*_rag* @ArthurZucker
-/src/transformers/models/realm/mod*_realm* @ArthurZucker
-/src/transformers/models/recurrent_gemma/mod*_recurrent_gemma* @ArthurZucker
-/src/transformers/models/reformer/mod*_reformer* @ArthurZucker
-/src/transformers/models/rembert/mod*_rembert* @ArthurZucker
-/src/transformers/models/retribert/mod*_retribert* @ArthurZucker
-/src/transformers/models/roberta/mod*_roberta* @ArthurZucker
-/src/transformers/models/roberta_prelayernorm/mod*_roberta_prelayernorm* @ArthurZucker
-/src/transformers/models/roc_bert/mod*_roc_bert* @ArthurZucker
-/src/transformers/models/roformer/mod*_roformer* @ArthurZucker
-/src/transformers/models/rwkv/mod*_rwkv* @ArthurZucker
-/src/transformers/models/splinter/mod*_splinter* @ArthurZucker
-/src/transformers/models/squeezebert/mod*_squeezebert* @ArthurZucker
-/src/transformers/models/stablelm/mod*_stablelm* @ArthurZucker
-/src/transformers/models/starcoder2/mod*_starcoder2* @ArthurZucker
-/src/transformers/models/switch_transformers/mod*_switch_transformers* @ArthurZucker
-/src/transformers/models/t5/mod*_t5* @ArthurZucker
-/src/transformers/models/t5v1.1/mod*_t5v1.1* @ArthurZucker
-/src/transformers/models/tapex/mod*_tapex* @ArthurZucker
-/src/transformers/models/transfo_xl/mod*_transfo_xl* @ArthurZucker
-/src/transformers/models/ul2/mod*_ul2* @ArthurZucker
-/src/transformers/models/umt5/mod*_umt5* @ArthurZucker
-/src/transformers/models/xmod/mod*_xmod* @ArthurZucker
-/src/transformers/models/xglm/mod*_xglm* @ArthurZucker
-/src/transformers/models/xlm/mod*_xlm* @ArthurZucker
-/src/transformers/models/xlm_prophetnet/mod*_xlm_prophetnet* @ArthurZucker
-/src/transformers/models/xlm_roberta/mod*_xlm_roberta* @ArthurZucker
-/src/transformers/models/xlm_roberta_xl/mod*_xlm_roberta_xl* @ArthurZucker
-/src/transformers/models/xlm_v/mod*_xlm_v* @ArthurZucker
-/src/transformers/models/xlnet/mod*_xlnet* @ArthurZucker
-/src/transformers/models/yoso/mod*_yoso* @ArthurZucker
-/src/transformers/models/zamba/mod*_zamba* @ArthurZucker
-
-# Vision models
-/src/transformers/models/beit/mod*_beit* @amyeroberts @qubvel
-/src/transformers/models/bit/mod*_bit* @amyeroberts @qubvel
-/src/transformers/models/conditional_detr/mod*_conditional_detr* @amyeroberts @qubvel
-/src/transformers/models/convnext/mod*_convnext* @amyeroberts @qubvel
-/src/transformers/models/convnextv2/mod*_convnextv2* @amyeroberts @qubvel
-/src/transformers/models/cvt/mod*_cvt* @amyeroberts @qubvel
-/src/transformers/models/deformable_detr/mod*_deformable_detr* @amyeroberts @qubvel
-/src/transformers/models/deit/mod*_deit* @amyeroberts @qubvel
-/src/transformers/models/depth_anything/mod*_depth_anything* @amyeroberts @qubvel
-/src/transformers/models/depth_anything_v2/mod*_depth_anything_v2* @amyeroberts @qubvel
-/src/transformers/models/deta/mod*_deta* @amyeroberts @qubvel
-/src/transformers/models/detr/mod*_detr* @amyeroberts @qubvel
-/src/transformers/models/dinat/mod*_dinat* @amyeroberts @qubvel
-/src/transformers/models/dinov2/mod*_dinov2* @amyeroberts @qubvel
-/src/transformers/models/dinov2_with_registers/mod*_dinov2_with_registers* @amyeroberts @qubvel
-/src/transformers/models/dit/mod*_dit* @amyeroberts @qubvel
-/src/transformers/models/dpt/mod*_dpt* @amyeroberts @qubvel
-/src/transformers/models/efficientformer/mod*_efficientformer* @amyeroberts @qubvel
-/src/transformers/models/efficientnet/mod*_efficientnet* @amyeroberts @qubvel
-/src/transformers/models/focalnet/mod*_focalnet* @amyeroberts @qubvel
-/src/transformers/models/glpn/mod*_glpn* @amyeroberts @qubvel
-/src/transformers/models/hiera/mod*_hiera* @amyeroberts @qubvel
-/src/transformers/models/ijepa/mod*_ijepa* @amyeroberts @qubvel
-/src/transformers/models/imagegpt/mod*_imagegpt* @amyeroberts @qubvel
-/src/transformers/models/levit/mod*_levit* @amyeroberts @qubvel
-/src/transformers/models/mask2former/mod*_mask2former* @amyeroberts @qubvel
-/src/transformers/models/maskformer/mod*_maskformer* @amyeroberts @qubvel
-/src/transformers/models/mobilenet_v1/mod*_mobilenet_v1* @amyeroberts @qubvel
-/src/transformers/models/mobilenet_v2/mod*_mobilenet_v2* @amyeroberts @qubvel
-/src/transformers/models/mobilevit/mod*_mobilevit* @amyeroberts @qubvel
-/src/transformers/models/mobilevitv2/mod*_mobilevitv2* @amyeroberts @qubvel
-/src/transformers/models/nat/mod*_nat* @amyeroberts @qubvel
-/src/transformers/models/poolformer/mod*_poolformer* @amyeroberts @qubvel
-/src/transformers/models/pvt/mod*_pvt* @amyeroberts @qubvel
-/src/transformers/models/pvt_v2/mod*_pvt_v2* @amyeroberts @qubvel
-/src/transformers/models/regnet/mod*_regnet* @amyeroberts @qubvel
-/src/transformers/models/resnet/mod*_resnet* @amyeroberts @qubvel
-/src/transformers/models/rt_detr/mod*_rt_detr* @amyeroberts @qubvel
-/src/transformers/models/segformer/mod*_segformer* @amyeroberts @qubvel
-/src/transformers/models/seggpt/mod*_seggpt* @amyeroberts @qubvel
-/src/transformers/models/superpoint/mod*_superpoint* @amyeroberts @qubvel
-/src/transformers/models/swiftformer/mod*_swiftformer* @amyeroberts @qubvel
-/src/transformers/models/swin/mod*_swin* @amyeroberts @qubvel
-/src/transformers/models/swinv2/mod*_swinv2* @amyeroberts @qubvel
-/src/transformers/models/swin2sr/mod*_swin2sr* @amyeroberts @qubvel
-/src/transformers/models/table_transformer/mod*_table_transformer* @amyeroberts @qubvel
-/src/transformers/models/textnet/mod*_textnet* @amyeroberts @qubvel
-/src/transformers/models/timm_wrapper/mod*_timm_wrapper* @amyeroberts @qubvel
-/src/transformers/models/upernet/mod*_upernet* @amyeroberts @qubvel
-/src/transformers/models/van/mod*_van* @amyeroberts @qubvel
-/src/transformers/models/vit/mod*_vit* @amyeroberts @qubvel
-/src/transformers/models/vit_hybrid/mod*_vit_hybrid* @amyeroberts @qubvel
-/src/transformers/models/vitdet/mod*_vitdet* @amyeroberts @qubvel
-/src/transformers/models/vit_mae/mod*_vit_mae* @amyeroberts @qubvel
-/src/transformers/models/vitmatte/mod*_vitmatte* @amyeroberts @qubvel
-/src/transformers/models/vit_msn/mod*_vit_msn* @amyeroberts @qubvel
-/src/transformers/models/vitpose/mod*_vitpose* @amyeroberts @qubvel
-/src/transformers/models/yolos/mod*_yolos* @amyeroberts @qubvel
-/src/transformers/models/zoedepth/mod*_zoedepth* @amyeroberts @qubvel
-
-# Audio models
-/src/transformers/models/audio_spectrogram_transformer/mod*_audio_spectrogram_transformer* @eustlb
-/src/transformers/models/bark/mod*_bark* @eustlb
-/src/transformers/models/clap/mod*_clap* @eustlb
-/src/transformers/models/dac/mod*_dac* @eustlb
-/src/transformers/models/encodec/mod*_encodec* @eustlb
-/src/transformers/models/hubert/mod*_hubert* @eustlb
-/src/transformers/models/mctct/mod*_mctct* @eustlb
-/src/transformers/models/mimi/mod*_mimi* @eustlb
-/src/transformers/models/mms/mod*_mms* @eustlb
-/src/transformers/models/moshi/mod*_moshi* @eustlb
-/src/transformers/models/musicgen/mod*_musicgen* @eustlb
-/src/transformers/models/musicgen_melody/mod*_musicgen_melody* @eustlb
-/src/transformers/models/pop2piano/mod*_pop2piano* @eustlb
-/src/transformers/models/seamless_m4t/mod*_seamless_m4t* @eustlb
-/src/transformers/models/seamless_m4t_v2/mod*_seamless_m4t_v2* @eustlb
-/src/transformers/models/sew/mod*_sew* @eustlb
-/src/transformers/models/sew_d/mod*_sew_d* @eustlb
-/src/transformers/models/speech_to_text/mod*_speech_to_text* @eustlb
-/src/transformers/models/speech_to_text_2/mod*_speech_to_text_2* @eustlb
-/src/transformers/models/speecht5/mod*_speecht5* @eustlb
-/src/transformers/models/unispeech/mod*_unispeech* @eustlb
-/src/transformers/models/unispeech_sat/mod*_unispeech_sat* @eustlb
-/src/transformers/models/univnet/mod*_univnet* @eustlb
-/src/transformers/models/vits/mod*_vits* @eustlb
-/src/transformers/models/wav2vec2/mod*_wav2vec2* @eustlb
-/src/transformers/models/wav2vec2_bert/mod*_wav2vec2_bert* @eustlb
-/src/transformers/models/wav2vec2_conformer/mod*_wav2vec2_conformer* @eustlb
-/src/transformers/models/wav2vec2_phoneme/mod*_wav2vec2_phoneme* @eustlb
-/src/transformers/models/wavlm/mod*_wavlm* @eustlb
-/src/transformers/models/whisper/mod*_whisper* @eustlb
-/src/transformers/models/xls_r/mod*_xls_r* @eustlb
-/src/transformers/models/xlsr_wav2vec2/mod*_xlsr_wav2vec2* @eustlb
-
-# Video models
-/src/transformers/models/timesformer/mod*_timesformer* @Rocketknight1
-/src/transformers/models/videomae/mod*_videomae* @Rocketknight1
-/src/transformers/models/vivit/mod*_vivit* @Rocketknight1
-
-# Multimodal models
-/src/transformers/models/align/mod*_align* @zucchini-nlp
-/src/transformers/models/altclip/mod*_altclip* @zucchini-nlp
-/src/transformers/models/aria/mod*_aria* @zucchini-nlp
-/src/transformers/models/blip/mod*_blip* @zucchini-nlp
-/src/transformers/models/blip_2/mod*_blip_2* @zucchini-nlp
-/src/transformers/models/bridgetower/mod*_bridgetower* @zucchini-nlp
-/src/transformers/models/bros/mod*_bros* @zucchini-nlp
-/src/transformers/models/chameleon/mod*_chameleon* @zucchini-nlp
-/src/transformers/models/chinese_clip/mod*_chinese_clip* @zucchini-nlp
-/src/transformers/models/clip/mod*_clip* @zucchini-nlp
-/src/transformers/models/clipseg/mod*_clipseg* @zucchini-nlp
-/src/transformers/models/clvp/mod*_clvp* @zucchini-nlp
-/src/transformers/models/colpali/mod*_colpali* @zucchini-nlp @yonigozlan
-/src/transformers/models/data2vec/mod*_data2vec* @zucchini-nlp
-/src/transformers/models/deplot/mod*_deplot* @zucchini-nlp
-/src/transformers/models/donut/mod*_donut* @zucchini-nlp
-/src/transformers/models/flava/mod*_flava* @zucchini-nlp
-/src/transformers/models/git/mod*_git* @zucchini-nlp
-/src/transformers/models/grounding_dino/mod*_grounding_dino* @qubvel
-/src/transformers/models/groupvit/mod*_groupvit* @zucchini-nlp
-/src/transformers/models/idefics/mod*_idefics* @zucchini-nlp
-/src/transformers/models/idefics2/mod*_idefics2* @zucchini-nlp
-/src/transformers/models/idefics3/mod*_idefics3* @zucchini-nlp
-/src/transformers/models/instructblip/mod*_instructblip* @zucchini-nlp
-/src/transformers/models/instructblipvideo/mod*_instructblipvideo* @zucchini-nlp
-/src/transformers/models/kosmos_2/mod*_kosmos_2* @zucchini-nlp
-/src/transformers/models/layoutlm/mod*_layoutlm* @NielsRogge
-/src/transformers/models/layoutlmv2/mod*_layoutlmv2* @NielsRogge
-/src/transformers/models/layoutlmv3/mod*_layoutlmv3* @NielsRogge
-/src/transformers/models/layoutxlm/mod*_layoutxlm* @NielsRogge
-/src/transformers/models/lilt/mod*_lilt* @zucchini-nlp
-/src/transformers/models/llava/mod*_llava* @zucchini-nlp @arthurzucker
-/src/transformers/models/llava_next/mod*_llava_next* @zucchini-nlp
-/src/transformers/models/llava_next_video/mod*_llava_next_video* @zucchini-nlp
-/src/transformers/models/llava_onevision/mod*_llava_onevision* @zucchini-nlp
-/src/transformers/models/lxmert/mod*_lxmert* @zucchini-nlp
-/src/transformers/models/matcha/mod*_matcha* @zucchini-nlp
-/src/transformers/models/mgp_str/mod*_mgp_str* @zucchini-nlp
-/src/transformers/models/mllama/mod*_mllama* @zucchini-nlp
-/src/transformers/models/nougat/mod*_nougat* @NielsRogge
-/src/transformers/models/omdet_turbo/mod*_omdet_turbo* @qubvel @yonigozlan
-/src/transformers/models/oneformer/mod*_oneformer* @zucchini-nlp
-/src/transformers/models/owlvit/mod*_owlvit* @qubvel
-/src/transformers/models/owlv2/mod*_owlv2* @qubvel
-/src/transformers/models/paligemma/mod*_paligemma* @zucchini-nlp @molbap
-/src/transformers/models/perceiver/mod*_perceiver* @zucchini-nlp
-/src/transformers/models/pix2struct/mod*_pix2struct* @zucchini-nlp
-/src/transformers/models/pixtral/mod*_pixtral* @zucchini-nlp @ArthurZucker
-/src/transformers/models/qwen2_audio/mod*_qwen2_audio* @zucchini-nlp @ArthurZucker
-/src/transformers/models/qwen2_vl/mod*_qwen2_vl* @zucchini-nlp @ArthurZucker
-/src/transformers/models/sam/mod*_sam* @zucchini-nlp @ArthurZucker
-/src/transformers/models/siglip/mod*_siglip* @zucchini-nlp
-/src/transformers/models/speech_encoder_decoder/mod*_speech_encoder_decoder* @zucchini-nlp
-/src/transformers/models/tapas/mod*_tapas* @NielsRogge
-/src/transformers/models/trocr/mod*_trocr* @zucchini-nlp
-/src/transformers/models/tvlt/mod*_tvlt* @zucchini-nlp
-/src/transformers/models/tvp/mod*_tvp* @zucchini-nlp
-/src/transformers/models/udop/mod*_udop* @zucchini-nlp
-/src/transformers/models/video_llava/mod*_video_llava* @zucchini-nlp
-/src/transformers/models/vilt/mod*_vilt* @zucchini-nlp
-/src/transformers/models/vipllava/mod*_vipllava* @zucchini-nlp
-/src/transformers/models/vision_encoder_decoder/mod*_vision_encoder_decoder* @Rocketknight1
-/src/transformers/models/vision_text_dual_encoder/mod*_vision_text_dual_encoder* @Rocketknight1
-/src/transformers/models/visual_bert/mod*_visual_bert* @zucchini-nlp
-/src/transformers/models/xclip/mod*_xclip* @zucchini-nlp
-
-# Reinforcement learning models
-/src/transformers/models/decision_transformer/mod*_decision_transformer* @Rocketknight1
-/src/transformers/models/trajectory_transformer/mod*_trajectory_transformer* @Rocketknight1
-
-# Time series models
-/src/transformers/models/autoformer/mod*_autoformer* @Rocketknight1
-/src/transformers/models/informer/mod*_informer* @Rocketknight1
-/src/transformers/models/patchtsmixer/mod*_patchtsmixer* @Rocketknight1
-/src/transformers/models/patchtst/mod*_patchtst* @Rocketknight1
-/src/transformers/models/time_series_transformer/mod*_time_series_transformer* @Rocketknight1
-
-# Graph models
-/src/transformers/models/graphormer/mod*_graphormer* @clefourrier
-
-# Finally, files with no owners that shouldn't generate pings, usually automatically generated and checked in the CI
-utils/dummy*

.github/workflows/add-model-like.yml

@@ -23,7 +23,7 @@ jobs:
           sudo apt -y update && sudo apt install -y libsndfile1-dev
 
       - name: Load cached virtual environment
-        uses: actions/cache@v4
+        uses: actions/cache@v2
         id: cache
         with:
           path: ~/venv/
@@ -54,7 +54,7 @@ jobs:
       - name: Create model files
         run: |
           . ~/venv/bin/activate
-          transformers add-new-model-like --config_file tests/fixtures/add_distilbert_like_config.json --path_to_repo .
+          transformers-cli add-new-model-like --config_file tests/fixtures/add_distilbert_like_config.json --path_to_repo .
           make style
           make fix-copies
 

.github/workflows/assign-reviewers.yml

@@ -1,26 +0,0 @@
-name: Assign PR Reviewers
-on:
-  pull_request_target:
-    branches:
-      - main
-    types: [ready_for_review]
-
-jobs:
-  assign_reviewers:
-    permissions:
-       pull-requests: write
-    runs-on: ubuntu-22.04
-    steps:
-      - uses: actions/checkout@v4
-      - name: Set up Python
-        uses: actions/setup-python@v5
-        with:
-          python-version: '3.13'
-      - name: Install dependencies
-        run: |
-          python -m pip install --upgrade pip
-          pip install PyGithub
-      - name: Run assignment script
-        env:
-          GITHUB_TOKEN: ${{ secrets.GITHUB_TOKEN }}
-        run: python .github/scripts/assign_reviewers.py

.github/workflows/benchmark.yml

@@ -1,76 +1,42 @@
 name: Self-hosted runner (benchmark)
 
 on:
-  push:
-    branches: [main]
-  pull_request:
-    types: [ opened, labeled, reopened, synchronize ]
-
-concurrency:
-  group: ${{ github.workflow }}-${{ github.head_ref || github.run_id }}
-  cancel-in-progress: true
+  schedule:
+    - cron: "17 2 * * *"
+  workflow_call:
 
 env:
   HF_HOME: /mnt/cache
+  TF_FORCE_GPU_ALLOW_GROWTH: true
+
 
 jobs:
   benchmark:
     name: Benchmark
-    strategy:
-      matrix:
-        # group: [aws-g5-4xlarge-cache, aws-p4d-24xlarge-plus] (A100 runner is not enabled)
-        group: [aws-g5-4xlarge-cache]
-    runs-on:
-      group: ${{ matrix.group }}
-    if: |
-      (github.event_name == 'pull_request' && contains( github.event.pull_request.labels.*.name, 'run-benchmark') )||
-      (github.event_name == 'push' && github.ref == 'refs/heads/main')
+    runs-on: [single-gpu, nvidia-gpu, a10, ci]
     container:
-      image: huggingface/transformers-pytorch-gpu
-      options: --gpus all --privileged --ipc host
+      image: huggingface/transformers-all-latest-gpu
+      options: --gpus all --privileged --ipc host -v /mnt/cache/.cache/huggingface:/mnt/cache/
     steps:
-      - name: Get repo
-        uses: actions/checkout@v4
-        with:
-          ref: ${{ github.event.pull_request.head.sha || github.sha }}
-
-      - name: Install libpq-dev & psql
+      - name: Update clone
+        working-directory: /transformers
         run: |
-          apt update
-          apt install -y libpq-dev postgresql-client
-
-      - name: Install benchmark script dependencies
-        run: python3 -m pip install -r benchmark/requirements.txt
+          git fetch && git checkout ${{ github.sha }}
 
       - name: Reinstall transformers in edit mode (remove the one installed during docker image build)
         working-directory: /transformers
-        run: python3 -m pip uninstall -y transformers && python3 -m pip install -e ".[torch]"
+        run: python3 -m pip uninstall -y transformers && python3 -m pip install -e .
 
-      - name: Run database init script
+      - name: Benchmark (daily)
+        if: github.event_name == 'schedule'
+        working-directory: /transformers
         run: |
-          psql -f benchmark/init_db.sql
-        env:
-          PGDATABASE: metrics
-          PGHOST: ${{ secrets.TRANSFORMERS_BENCHMARKS_PGHOST }}
-          PGUSER: transformers_benchmarks
-          PGPASSWORD: ${{ secrets.TRANSFORMERS_BENCHMARKS_PGPASSWORD }}
+          python3 -m pip install optimum-benchmark>=0.2.0
+          HF_TOKEN=${{ secrets.TRANSFORMERS_BENCHMARK_TOKEN }} python3 benchmark/benchmark.py --repo_id hf-internal-testing/benchmark_results --path_in_repo $(date +'%Y-%m-%d') --config-dir benchmark/config --config-name generation --commit=${{ github.sha }} backend.model=google/gemma-2b backend.cache_implementation=null,static backend.torch_compile=false,true --multirun
 
-      - name: Run benchmark
+      - name: Benchmark (merged to main event)
+        if: github.event_name == 'push' && github.ref_name == 'main'
+        working-directory: /transformers
         run: |
-          git config --global --add safe.directory /__w/transformers/transformers
-          if [ "$GITHUB_EVENT_NAME" = "pull_request" ]; then
-            commit_id=$(echo "${{ github.event.pull_request.head.sha }}")
-          elif [ "$GITHUB_EVENT_NAME" = "push" ]; then
-            commit_id=$GITHUB_SHA
-          fi
-          commit_msg=$(git show -s --format=%s | cut -c1-70)
-          python3 benchmark/benchmarks_entrypoint.py "huggingface/transformers" "$BRANCH_NAME" "$commit_id" "$commit_msg"
-        env:
-          HF_TOKEN: ${{ secrets.HF_HUB_READ_TOKEN }}
-          # Enable this to see debug logs
-          # HF_HUB_VERBOSITY: debug
-          # TRANSFORMERS_VERBOSITY: debug
-          PGHOST: ${{ secrets.TRANSFORMERS_BENCHMARKS_PGHOST }}
-          PGUSER: transformers_benchmarks
-          PGPASSWORD: ${{ secrets.TRANSFORMERS_BENCHMARKS_PGPASSWORD }}
-          BRANCH_NAME: ${{ github.head_ref || github.ref_name }}
+          python3 -m pip install optimum-benchmark>=0.2.0
+          HF_TOKEN=${{ secrets.TRANSFORMERS_BENCHMARK_TOKEN }} python3 benchmark/benchmark.py --repo_id hf-internal-testing/benchmark_results_merge_event --path_in_repo $(date +'%Y-%m-%d') --config-dir benchmark/config --config-name generation --commit=${{ github.sha }} backend.model=google/gemma-2b backend.cache_implementation=null,static backend.torch_compile=false,true --multirun

.github/workflows/build-ci-docker-images.yml

@@ -26,7 +26,7 @@ jobs:
 
     strategy:
       matrix:
-        file: ["quality", "consistency", "custom-tokenizers", "torch-light", "tf-light", "exotic-models", "torch-tf-light", "jax-light", "examples-torch",  "examples-tf"]
+        file: ["quality", "consistency", "custom-tokenizers", "torch-light", "tf-light", "exotic-models", "torch-tf-light", "torch-jax-light", "jax-light", "examples-torch",  "examples-tf"]
     continue-on-error: true
 
     steps:
@@ -34,11 +34,11 @@ jobs:
         name: Set tag
         run: |
               if ${{contains(github.event.head_commit.message, '[build-ci-image]')}}; then
-                  echo "TAG=huggingface/transformers-${{ matrix.file }}:dev" >> "$GITHUB_ENV"
+                  echo "TAG=huggingface/transformers-${{ matrix.file }}:dev" >> "$GITHUB_ENV" 
                   echo "setting it to DEV!"
               else
                   echo "TAG=huggingface/transformers-${{ matrix.file }}" >> "$GITHUB_ENV"
-
+                  
               fi
       -
         name: Set up Docker Buildx
@@ -74,4 +74,4 @@ jobs:
           slack_channel: "#transformers-ci-circleci-images"
           title: 🤗 New docker images for CircleCI are pushed.
           status: ${{ job.status }}
-          slack_token: ${{ secrets.SLACK_CIFEEDBACK_BOT_TOKEN }}
+          slack_token: ${{ secrets.SLACK_CIFEEDBACK_BOT_TOKEN }}

.github/workflows/build-docker-images.yml

@@ -19,9 +19,8 @@ concurrency:
 
 jobs:
   latest-docker:
-    name: "Latest PyTorch [dev]"
-    runs-on:
-      group: aws-general-8-plus
+    name: "Latest PyTorch + TensorFlow [dev]"
+    runs-on: [intel-cpu, 8-cpu, ci]
     steps:
       -
         name: Set up Docker Buildx
@@ -63,14 +62,13 @@ jobs:
         uses: huggingface/hf-workflows/.github/actions/post-slack@main
         with:
           slack_channel: ${{ secrets.CI_SLACK_CHANNEL_DOCKER }}
-          title: 🤗 Results of the transformers-all-latest-gpu-push-ci docker build
+          title: 🤗 Results of the transformers-all-latest-gpu-push-ci docker build 
           status: ${{ job.status }}
           slack_token: ${{ secrets.SLACK_CIFEEDBACK_BOT_TOKEN }}
 
   latest-torch-deepspeed-docker:
     name: "Latest PyTorch + DeepSpeed"
-    runs-on:
-      group: aws-g4dn-2xlarge-cache
+    runs-on: [intel-cpu, 8-cpu, ci]
     steps:
       -
         name: Set up Docker Buildx
@@ -99,15 +97,14 @@ jobs:
         uses: huggingface/hf-workflows/.github/actions/post-slack@main
         with:
           slack_channel: ${{ secrets.CI_SLACK_CHANNEL_DOCKER}}
-          title: 🤗 Results of the transformers-pytorch-deepspeed-latest-gpu docker build
+          title: 🤗 Results of the transformers-pytorch-deepspeed-latest-gpu docker build 
           status: ${{ job.status }}
           slack_token: ${{ secrets.SLACK_CIFEEDBACK_BOT_TOKEN }}
 
   # Can't build 2 images in a single job `latest-torch-deepspeed-docker` (for `nvcr.io/nvidia`)
   latest-torch-deepspeed-docker-for-push-ci-daily-build:
     name: "Latest PyTorch + DeepSpeed (Push CI - Daily Build)"
-    runs-on:
-      group: aws-general-8-plus
+    runs-on: [intel-cpu, 8-cpu, ci]
     steps:
       -
         name: Set up Docker Buildx
@@ -140,7 +137,7 @@ jobs:
         uses: huggingface/hf-workflows/.github/actions/post-slack@main
         with:
           slack_channel: ${{ secrets.CI_SLACK_CHANNEL_DOCKER }}
-          title: 🤗 Results of the transformers-pytorch-deepspeed-latest-gpu-push-ci docker build
+          title: 🤗 Results of the transformers-pytorch-deepspeed-latest-gpu-push-ci docker build 
           status: ${{ job.status }}
           slack_token: ${{ secrets.SLACK_CIFEEDBACK_BOT_TOKEN }}
 
@@ -148,8 +145,7 @@ jobs:
     name: "Doc builder"
     # Push CI doesn't need this image
     if: inputs.image_postfix != '-push-ci'
-    runs-on:
-      group: aws-general-8-plus
+    runs-on: [intel-cpu, 8-cpu, ci]
     steps:
       -
         name: Set up Docker Buildx
@@ -176,7 +172,7 @@ jobs:
         uses: huggingface/hf-workflows/.github/actions/post-slack@main
         with:
           slack_channel: ${{ secrets.CI_SLACK_CHANNEL_DOCKER }}
-          title: 🤗 Results of the huggingface/transformers-doc-builder docker build
+          title: 🤗 Results of the huggingface/transformers-doc-builder docker build 
           status: ${{ job.status }}
           slack_token: ${{ secrets.SLACK_CIFEEDBACK_BOT_TOKEN }}
 
@@ -184,8 +180,7 @@ jobs:
     name: "Latest PyTorch [dev]"
     # Push CI doesn't need this image
     if: inputs.image_postfix != '-push-ci'
-    runs-on:
-      group: aws-general-8-plus
+    runs-on: [intel-cpu, 8-cpu, ci]
     steps:
       -
         name: Set up Docker Buildx
@@ -214,28 +209,27 @@ jobs:
         uses: huggingface/hf-workflows/.github/actions/post-slack@main
         with:
           slack_channel: ${{ secrets.CI_SLACK_CHANNEL_DOCKER }}
-          title: 🤗 Results of the huggingface/transformers-pytorch-gpudocker build
+          title: 🤗 Results of the huggingface/transformers-pytorch-gpudocker build 
           status: ${{ job.status }}
           slack_token: ${{ secrets.SLACK_CIFEEDBACK_BOT_TOKEN }}
 
   latest-pytorch-amd:
     name: "Latest PyTorch (AMD) [dev]"
-    runs-on:
-      group: aws-general-8-plus
+    runs-on: [intel-cpu, 8-cpu, ci]
     steps:
-      -
+      - 
         name: Set up Docker Buildx
         uses: docker/setup-buildx-action@v3
-      -
+      - 
         name: Check out code
         uses: actions/checkout@v4
-      -
+      - 
         name: Login to DockerHub
         uses: docker/login-action@v3
         with:
           username: ${{ secrets.DOCKERHUB_USERNAME }}
           password: ${{ secrets.DOCKERHUB_PASSWORD }}
-      -
+      - 
         name: Build and push
         uses: docker/build-push-action@v5
         with:
@@ -263,14 +257,15 @@ jobs:
         uses: huggingface/hf-workflows/.github/actions/post-slack@main
         with:
           slack_channel: ${{ secrets.CI_SLACK_CHANNEL_DOCKER }}
-          title: 🤗 Results of the huggingface/transformers-pytorch-amd-gpu-push-ci build
+          title: 🤗 Results of the huggingface/transformers-pytorch-amd-gpu-push-ci build 
           status: ${{ job.status }}
           slack_token: ${{ secrets.SLACK_CIFEEDBACK_BOT_TOKEN }}
 
-  latest-pytorch-deepspeed-amd:
-    name: "PyTorch + DeepSpeed (AMD) [dev]"
-    runs-on:
-      group: aws-general-8-plus
+  latest-tensorflow:
+    name: "Latest TensorFlow [dev]"
+    # Push CI doesn't need this image
+    if: inputs.image_postfix != '-push-ci'
+    runs-on: [intel-cpu, 8-cpu, ci]
     steps:
       -
         name: Set up Docker Buildx
@@ -285,6 +280,41 @@ jobs:
           username: ${{ secrets.DOCKERHUB_USERNAME }}
           password: ${{ secrets.DOCKERHUB_PASSWORD }}
       -
+        name: Build and push
+        uses: docker/build-push-action@v5
+        with:
+          context: ./docker/transformers-tensorflow-gpu
+          build-args: |
+            REF=main
+          push: true
+          tags: huggingface/transformers-tensorflow-gpu
+
+      - name: Post to Slack
+        if: always()
+        uses: huggingface/hf-workflows/.github/actions/post-slack@main
+        with:
+          slack_channel: ${{ secrets.CI_SLACK_CHANNEL_DOCKER }}
+          title: 🤗 Results of the huggingface/transformers-tensorflow-gpu build 
+          status: ${{ job.status }}
+          slack_token: ${{ secrets.SLACK_CIFEEDBACK_BOT_TOKEN }}
+
+  latest-pytorch-deepspeed-amd:
+    name: "PyTorch + DeepSpeed (AMD) [dev]"
+    runs-on: [intel-cpu, 8-cpu, ci]
+    steps:
+      - 
+        name: Set up Docker Buildx
+        uses: docker/setup-buildx-action@v3
+      - 
+        name: Check out code
+        uses: actions/checkout@v4
+      - 
+        name: Login to DockerHub
+        uses: docker/login-action@v3
+        with:
+          username: ${{ secrets.DOCKERHUB_USERNAME }}
+          password: ${{ secrets.DOCKERHUB_PASSWORD }}
+      - 
         name: Build and push
         uses: docker/build-push-action@v5
         with:
@@ -312,7 +342,7 @@ jobs:
         uses: huggingface/hf-workflows/.github/actions/post-slack@main
         with:
           slack_channel: ${{ secrets.CI_SLACK_CHANNEL_DOCKER }}
-          title: 🤗 Results of the transformers-pytorch-deepspeed-amd-gpu build
+          title: 🤗 Results of the transformers-pytorch-deepspeed-amd-gpu build 
           status: ${{ job.status }}
           slack_token: ${{ secrets.SLACK_CIFEEDBACK_BOT_TOKEN }}
 
@@ -320,8 +350,7 @@ jobs:
     name: "Latest Pytorch + Quantization [dev]"
      # Push CI doesn't need this image
     if: inputs.image_postfix != '-push-ci'
-    runs-on:
-      group: aws-general-8-plus
+    runs-on: [intel-cpu, 8-cpu, ci]
     steps:
       -
         name: Set up Docker Buildx
@@ -350,6 +379,6 @@ jobs:
         uses: huggingface/hf-workflows/.github/actions/post-slack@main
         with:
           slack_channel: ${{ secrets.CI_SLACK_CHANNEL_DOCKER }}
-          title: 🤗 Results of the transformers-quantization-latest-gpu build
+          title: 🤗 Results of the transformers-quantization-latest-gpu build 
           status: ${{ job.status }}
           slack_token: ${{ secrets.SLACK_CIFEEDBACK_BOT_TOKEN }}

.github/workflows/build-nightly-ci-docker-images.yml

@@ -13,8 +13,7 @@ concurrency:
 jobs:
   latest-with-torch-nightly-docker:
     name: "Nightly PyTorch + Stable TensorFlow"
-    runs-on:
-      group: aws-general-8-plus
+    runs-on: [intel-cpu, 8-cpu, ci]
     steps:
       -
         name: Set up Docker Buildx
@@ -41,8 +40,7 @@ jobs:
 
   nightly-torch-deepspeed-docker:
     name: "Nightly PyTorch + DeepSpeed"
-    runs-on:
-      group: aws-g4dn-2xlarge-cache
+    runs-on: [intel-cpu, 8-cpu, ci]
     steps:
       -
         name: Set up Docker Buildx
@@ -64,4 +62,4 @@ jobs:
           build-args: |
             REF=main
           push: true
-          tags: huggingface/transformers-pytorch-deepspeed-nightly-gpu
+          tags: huggingface/transformers-pytorch-deepspeed-nightly-gpu

.github/workflows/build-past-ci-docker-images.yml

@@ -16,8 +16,7 @@ jobs:
       fail-fast: false
       matrix:
         version: ["1.13", "1.12", "1.11"]
-    runs-on:
-      group: aws-general-8-plus
+    runs-on: [intel-cpu, 8-cpu, ci]
     steps:
       -
         name: Set up Docker Buildx
@@ -61,8 +60,7 @@ jobs:
       fail-fast: false
       matrix:
         version: ["2.11", "2.10", "2.9", "2.8", "2.7", "2.6", "2.5"]
-    runs-on:
-      group: aws-general-8-plus
+    runs-on: [intel-cpu, 8-cpu, ci]
     steps:
       -
         name: Set up Docker Buildx

.github/workflows/build_documentation.yml

@@ -1,7 +1,6 @@
 name: Build documentation
 
 on:
-  workflow_dispatch:
   push:
     branches:
       - main
@@ -16,7 +15,7 @@ jobs:
       commit_sha: ${{ github.sha }}
       package: transformers
       notebook_folder: transformers_doc
-      languages: ar de en es fr hi it ko pt tr zh ja te
+      languages: de en es fr hi it ko pt tr zh ja te
       custom_container: huggingface/transformers-doc-builder
     secrets:
       token: ${{ secrets.HUGGINGFACE_PUSH }}

.github/workflows/build_pr_documentation.yml

@@ -14,4 +14,5 @@ jobs:
       commit_sha: ${{ github.event.pull_request.head.sha }}
       pr_number: ${{ github.event.number }}
       package: transformers
-      languages: en
+      languages: de en es fr hi it ko pt tr zh ja te
+      custom_container: huggingface/transformers-doc-builder

.github/workflows/check_failed_tests.yml

@@ -1,205 +0,0 @@
-name: Process failed tests
-
-on:
-  workflow_call:
-    inputs:
-      docker:
-        required: true
-        type: string
-      start_sha:
-        required: true
-        type: string
-      job:
-        required: true
-        type: string
-      slack_report_channel:
-        required: true
-        type: string
-      ci_event:
-        required: true
-        type: string
-      report_repo_id:
-        required: true
-        type: string
-
-
-env:
-  HF_HOME: /mnt/cache
-  TRANSFORMERS_IS_CI: yes
-  OMP_NUM_THREADS: 8
-  MKL_NUM_THREADS: 8
-  RUN_SLOW: yes
-  # For gated repositories, we still need to agree to share information on the Hub repo. page in order to get access.
-  # This token is created under the bot `hf-transformers-bot`.
-  HF_HUB_READ_TOKEN: ${{ secrets.HF_HUB_READ_TOKEN }}
-  SIGOPT_API_TOKEN: ${{ secrets.SIGOPT_API_TOKEN }}
-  TF_FORCE_GPU_ALLOW_GROWTH: true
-  CUDA_VISIBLE_DEVICES: 0,1
-
-
-jobs:
-  check_new_failures:
-    name: " "
-    runs-on:
-      group: aws-g4dn-4xlarge-cache
-    container:
-      image: ${{ inputs.docker }}
-      options: --gpus all --shm-size "16gb" --ipc host -v /mnt/cache/.cache/huggingface:/mnt/cache/
-    steps:
-      - uses: actions/download-artifact@v4
-        with:
-          name: ci_results_${{ inputs.job }}
-          path: /transformers/ci_results_${{ inputs.job }}
-
-      - name: Check file
-        working-directory: /transformers
-        run: |
-          if [ -f ci_results_${{ inputs.job }}/new_failures.json ]; then
-            echo "`ci_results_${{ inputs.job }}/new_failures.json` exists, continue ..."
-            echo "process=true" >> $GITHUB_ENV
-          else
-            echo "`ci_results_${{ inputs.job }}/new_failures.json` doesn't exist, abort."
-            echo "process=false" >> $GITHUB_ENV
-          fi
-
-      - uses: actions/download-artifact@v4
-        if: ${{ env.process == 'true' }}
-        with:
-          pattern: setup_values*
-          path: setup_values
-          merge-multiple: true
-
-      - name: Prepare some setup values
-        if: ${{ env.process == 'true' }}
-        run: |
-          if [ -f setup_values/prev_workflow_run_id.txt ]; then
-            echo "PREV_WORKFLOW_RUN_ID=$(cat setup_values/prev_workflow_run_id.txt)" >> $GITHUB_ENV
-          else
-            echo "PREV_WORKFLOW_RUN_ID=" >> $GITHUB_ENV
-          fi
-
-          if [ -f setup_values/other_workflow_run_id.txt ]; then
-            echo "OTHER_WORKFLOW_RUN_ID=$(cat setup_values/other_workflow_run_id.txt)" >> $GITHUB_ENV
-          else
-            echo "OTHER_WORKFLOW_RUN_ID=" >> $GITHUB_ENV
-          fi
-
-      - name: Update clone
-        working-directory: /transformers
-        if: ${{ env.process == 'true' }}
-        run: git fetch && git checkout ${{ github.sha }}
-
-      - name: Get target commit
-        working-directory: /transformers/utils
-        if: ${{ env.process == 'true' }}
-        run: |
-          echo "END_SHA=$(TOKEN=${{ secrets.ACCESS_REPO_INFO_TOKEN }} python3 -c 'import os; from get_previous_daily_ci import get_last_daily_ci_run_commit; commit=get_last_daily_ci_run_commit(token=os.environ["TOKEN"], workflow_run_id=os.environ["PREV_WORKFLOW_RUN_ID"]); print(commit)')" >> $GITHUB_ENV
-
-      - name: Checkout to `start_sha`
-        working-directory: /transformers
-        if: ${{ env.process == 'true' }}
-        run: git fetch && git checkout ${{ inputs.start_sha }}
-
-      - name: Reinstall transformers in edit mode (remove the one installed during docker image build)
-        working-directory: /transformers
-        if: ${{ env.process == 'true' }}
-        run: python3 -m pip uninstall -y transformers && python3 -m pip install -e .
-
-      - name: NVIDIA-SMI
-        if: ${{ env.process == 'true' }}
-        run: |
-          nvidia-smi
-
-      - name: Environment
-        working-directory: /transformers
-        if: ${{ env.process == 'true' }}
-        run: |
-          python3 utils/print_env.py
-
-      - name: Show installed libraries and their versions
-        working-directory: /transformers
-        if: ${{ env.process == 'true' }}
-        run: pip freeze
-
-      - name: Check failed tests
-        working-directory: /transformers
-        if: ${{ env.process == 'true' }}
-        run: python3 utils/check_bad_commit.py --start_commit ${{ inputs.start_sha }} --end_commit ${{ env.END_SHA }} --file ci_results_${{ inputs.job }}/new_failures.json --output_file new_failures_with_bad_commit.json
-
-      - name: Show results
-        working-directory: /transformers
-        if: ${{ env.process == 'true' }}
-        run: |
-          ls -l new_failures_with_bad_commit.json
-          cat new_failures_with_bad_commit.json
-
-      - name: Checkout back
-        working-directory: /transformers
-        if: ${{ env.process == 'true' }}
-        run: |
-          git checkout ${{ inputs.start_sha }}
-
-      - name: Process report
-        shell: bash
-        working-directory: /transformers
-        if: ${{ env.process == 'true' }}
-        env:
-          ACCESS_REPO_INFO_TOKEN: ${{ secrets.ACCESS_REPO_INFO_TOKEN }}
-          TRANSFORMERS_CI_RESULTS_UPLOAD_TOKEN: ${{ secrets.TRANSFORMERS_CI_RESULTS_UPLOAD_TOKEN }}
-          JOB_NAME: ${{ inputs.job }}
-          REPORT_REPO_ID: ${{ inputs.report_repo_id }}
-        run: |
-          python3 utils/process_bad_commit_report.py
-
-      - name: Process report
-        shell: bash
-        working-directory: /transformers
-        if: ${{ env.process == 'true' }}
-        env:
-          ACCESS_REPO_INFO_TOKEN: ${{ secrets.ACCESS_REPO_INFO_TOKEN }}
-          TRANSFORMERS_CI_RESULTS_UPLOAD_TOKEN: ${{ secrets.TRANSFORMERS_CI_RESULTS_UPLOAD_TOKEN }}
-          JOB_NAME: ${{ inputs.job }}
-          REPORT_REPO_ID: ${{ inputs.report_repo_id }}
-        run: |
-          {
-            echo 'REPORT_TEXT<<EOF'
-            python3 utils/process_bad_commit_report.py
-            echo EOF
-          } >> "$GITHUB_ENV"
-
-      - name: Prepare Slack report title
-        working-directory: /transformers
-        if: ${{ env.process == 'true' }}
-        run: |
-          pip install slack_sdk
-          echo "title=$(python3 -c 'import sys; sys.path.append("utils"); from utils.notification_service import job_to_test_map; ci_event = "${{ inputs.ci_event }}"; job = "${{ inputs.job }}"; test_name = job_to_test_map[job]; title = f"New failed tests of {ci_event}" + ":" + f" {test_name}"; print(title)')" >> $GITHUB_ENV
-
-      - name: Send processed report
-        if: ${{ env.process == 'true' && !endsWith(env.REPORT_TEXT, '{}') }}
-        uses: slackapi/slack-github-action@6c661ce58804a1a20f6dc5fbee7f0381b469e001
-        with:
-          # Slack channel id, channel name, or user id to post message.
-          # See also: https://api.slack.com/methods/chat.postMessage#channels
-          channel-id: '#${{ inputs.slack_report_channel }}'
-          # For posting a rich message using Block Kit
-          payload: |
-            {
-              "blocks": [
-                {
-                  "type": "header",
-                  "text": {
-                    "type": "plain_text",
-                    "text": "${{ env.title }}"
-                  }
-                },
-                {
-                  "type": "section",
-                  "text": {
-                    "type": "mrkdwn",
-                    "text": "${{ env.REPORT_TEXT }}"
-                  }
-                }
-              ]
-            }
-        env:
-          SLACK_BOT_TOKEN: ${{ secrets.SLACK_CIFEEDBACK_BOT_TOKEN }}

.github/workflows/check_tiny_models.yml

@@ -23,7 +23,7 @@ jobs:
 
       - uses: actions/checkout@v4
       - name: Set up Python 3.8
-        uses: actions/setup-python@v5
+        uses: actions/setup-python@v4
         with:
           # Semantic version range syntax or exact version of a Python version
           python-version: '3.8'

.github/workflows/doctest_job.yml

@@ -27,8 +27,7 @@ jobs:
       fail-fast: false
       matrix:
         split_keys: ${{ fromJson(inputs.split_keys) }}
-    runs-on: 
-      group: aws-g4dn-4xlarge-cache
+    runs-on: [single-gpu, nvidia-gpu, t4, ci]
     container:
       image: huggingface/transformers-all-latest-gpu
       options: --gpus 0 --shm-size "16gb" --ipc host -v /mnt/cache/.cache/huggingface:/mnt/cache/

.github/workflows/doctests.yml

@@ -14,8 +14,7 @@ env:
 jobs:
   setup:
     name: Setup
-    runs-on: 
-      group: aws-g4dn-4xlarge-cache
+    runs-on: [single-gpu, nvidia-gpu, t4, ci]
     container:
       image: huggingface/transformers-all-latest-gpu
       options: --gpus 0 --shm-size "16gb" --ipc host -v /mnt/cache/.cache/huggingface:/mnt/cache/
@@ -86,4 +85,4 @@ jobs:
         uses: actions/upload-artifact@v4
         with:
           name: doc_test_results
-          path: doc_test_results
+          path: doc_test_results

.github/workflows/model_jobs.yml

@@ -12,16 +12,12 @@ on:
       slice_id:
         required: true
         type: number
-      runner_map:
-        required: false
+      runner:
+        required: true
         type: string
       docker:
         required: true
         type: string
-      report_name_prefix:
-        required: false
-        default: run_models_gpu
-        type: string
 
 env:
   HF_HOME: /mnt/cache
@@ -34,6 +30,7 @@ env:
   HF_HUB_READ_TOKEN: ${{ secrets.HF_HUB_READ_TOKEN }}
   SIGOPT_API_TOKEN: ${{ secrets.SIGOPT_API_TOKEN }}
   TF_FORCE_GPU_ALLOW_GROWTH: true
+  RUN_PT_TF_CROSS_TESTS: 1
   CUDA_VISIBLE_DEVICES: 0,1
 
 jobs:
@@ -44,8 +41,7 @@ jobs:
       fail-fast: false
       matrix:
         folders: ${{ fromJson(inputs.folder_slices)[inputs.slice_id] }}
-    runs-on:
-      group: ${{ fromJson(inputs.runner_map)[matrix.folders][inputs.machine_type] }}
+    runs-on: ['${{ inputs.machine_type }}', nvidia-gpu, t4, '${{ inputs.runner }}']
     container:
       image: ${{ inputs.docker }}
       options: --gpus all --shm-size "16gb" --ipc host -v /mnt/cache/.cache/huggingface:/mnt/cache/
@@ -101,42 +97,25 @@ jobs:
         working-directory: /transformers
         run: pip freeze
 
-      - name: Set `machine_type` for report and artifact names
-        working-directory: /transformers
-        shell: bash
-        run: |
-          echo "${{ inputs.machine_type }}"
-
-          if [ "${{ inputs.machine_type }}" = "aws-g4dn-4xlarge-cache" ]; then
-            machine_type=single-gpu
-          elif [ "${{ inputs.machine_type }}" = "aws-g4dn-12xlarge-cache" ]; then
-            machine_type=multi-gpu
-          else
-            machine_type=${{ inputs.machine_type }}
-          fi
-
-          echo "$machine_type"
-          echo "machine_type=$machine_type" >> $GITHUB_ENV
-
       - name: Run all tests on GPU
         working-directory: /transformers
-        run: python3 -m pytest -rsfE -v --make-reports=${{ env.machine_type }}_${{ inputs.report_name_prefix }}_${{ matrix.folders }}_test_reports tests/${{ matrix.folders }}
+        run: python3 -m pytest -rsfE -v --make-reports=${{ inputs.machine_type }}_run_models_gpu_${{ matrix.folders }}_test_reports tests/${{ matrix.folders }}
 
       - name: Failure short reports
         if: ${{ failure() }}
         continue-on-error: true
-        run: cat /transformers/reports/${{ env.machine_type }}_${{ inputs.report_name_prefix }}_${{ matrix.folders }}_test_reports/failures_short.txt
+        run: cat /transformers/reports/${{ inputs.machine_type }}_run_models_gpu_${{ matrix.folders }}_test_reports/failures_short.txt
 
       - name: Run test
         shell: bash
         run: |
-          mkdir -p /transformers/reports/${{ env.machine_type }}_${{ inputs.report_name_prefix }}_${{ matrix.folders }}_test_reports
-          echo "hello" > /transformers/reports/${{ env.machine_type }}_${{ inputs.report_name_prefix }}_${{ matrix.folders }}_test_reports/hello.txt
-          echo "${{ env.machine_type }}_${{ inputs.report_name_prefix }}_${{ matrix.folders }}_test_reports"
+          mkdir -p /transformers/reports/${{ inputs.machine_type }}_run_models_gpu_${{ matrix.folders }}_test_reports
+          echo "hello" > /transformers/reports/${{ inputs.machine_type }}_run_models_gpu_${{ matrix.folders }}_test_reports/hello.txt
+          echo "${{ inputs.machine_type }}_run_models_gpu_${{ matrix.folders }}_test_reports"
 
-      - name: "Test suite reports artifacts: ${{ env.machine_type }}_${{ inputs.report_name_prefix }}_${{ env.matrix_folders }}_test_reports"
+      - name: "Test suite reports artifacts: ${{ inputs.machine_type }}_run_models_gpu_${{ env.matrix_folders }}_test_reports"
         if: ${{ always() }}
         uses: actions/upload-artifact@v4
         with:
-          name: ${{ env.machine_type }}_${{ inputs.report_name_prefix }}_${{ env.matrix_folders }}_test_reports
-          path: /transformers/reports/${{ env.machine_type }}_${{ inputs.report_name_prefix }}_${{ matrix.folders }}_test_reports
+          name: ${{ inputs.machine_type }}_run_models_gpu_${{ env.matrix_folders }}_test_reports
+          path: /transformers/reports/${{ inputs.machine_type }}_run_models_gpu_${{ matrix.folders }}_test_reports

.github/workflows/model_jobs_intel_gaudi.yml

@@ -1,121 +0,0 @@
-name: model jobs
-
-on:
-  workflow_call:
-    inputs:
-      folder_slices:
-        required: true
-        type: string
-      slice_id:
-        required: true
-        type: number
-      runner:
-        required: true
-        type: string
-      machine_type:
-        required: true
-        type: string
-      report_name_prefix:
-        required: false
-        default: run_models_gpu
-        type: string
-
-env:
-  RUN_SLOW: yes
-  PT_HPU_LAZY_MODE: 0
-  TRANSFORMERS_IS_CI: yes
-  PT_ENABLE_INT64_SUPPORT: 1
-  HF_HUB_READ_TOKEN: ${{ secrets.HF_HUB_READ_TOKEN }}
-  SIGOPT_API_TOKEN: ${{ secrets.SIGOPT_API_TOKEN }}
-  HF_HOME: /mnt/cache/.cache/huggingface
-
-jobs:
-  run_models_gpu:
-    name: " "
-    strategy:
-      max-parallel: 8
-      fail-fast: false
-      matrix:
-        folders: ${{ fromJson(inputs.folder_slices)[inputs.slice_id] }}
-    runs-on:
-      group: ${{ inputs.runner }}
-    container:
-      image: vault.habana.ai/gaudi-docker/1.21.1/ubuntu22.04/habanalabs/pytorch-installer-2.6.0:latest
-      options: --runtime=habana
-        -v /mnt/cache/.cache/huggingface:/mnt/cache/.cache/huggingface
-        --env OMPI_MCA_btl_vader_single_copy_mechanism=none
-        --env HABANA_VISIBLE_DEVICES
-        --env HABANA_VISIBLE_MODULES
-        --cap-add=sys_nice
-        --shm-size=64G
-    steps:
-      - name: Echo input and matrix info
-        shell: bash
-        run: |
-          echo "${{ inputs.folder_slices }}"
-          echo "${{ matrix.folders }}"
-          echo "${{ toJson(fromJson(inputs.folder_slices)[inputs.slice_id]) }}"
-
-      - name: Echo folder ${{ matrix.folders }}
-        shell: bash
-        run: |
-          echo "${{ matrix.folders }}"
-          matrix_folders=${{ matrix.folders }}
-          matrix_folders=${matrix_folders/'models/'/'models_'}
-          echo "$matrix_folders"
-          echo "matrix_folders=$matrix_folders" >> $GITHUB_ENV
-
-      - name: Checkout
-        uses: actions/checkout@v4
-        with:
-          fetch-depth: 0
-
-      - name: Install dependencies
-        run: |
-          pip install -e .[testing,torch] "numpy<2.0.0" scipy scikit-learn
-
-      - name: HL-SMI
-        run: |
-          hl-smi
-          echo "HABANA_VISIBLE_DEVICES=${HABANA_VISIBLE_DEVICES}"
-          echo "HABANA_VISIBLE_MODULES=${HABANA_VISIBLE_MODULES}"
-
-      - name: Environment
-        run: python3 utils/print_env.py
-
-      - name: Show installed libraries and their versions
-        run: pip freeze
-
-      - name: Set `machine_type` for report and artifact names
-        shell: bash
-        run: |
-          if [ "${{ inputs.machine_type }}" = "1gaudi" ]; then
-            machine_type=single-gpu
-          elif [ "${{ inputs.machine_type }}" = "2gaudi" ]; then
-            machine_type=multi-gpu
-          else
-            machine_type=${{ inputs.machine_type }}
-          fi
-          echo "machine_type=$machine_type" >> $GITHUB_ENV
-
-      - name: Run all tests on Gaudi
-        run: python3 -m pytest -v --make-reports=${{ env.machine_type }}_${{ inputs.report_name_prefix }}_${{ matrix.folders }}_test_reports tests/${{ matrix.folders }}
-
-      - name: Failure short reports
-        if: ${{ failure() }}
-        continue-on-error: true
-        run: cat reports/${{ env.machine_type }}_${{ inputs.report_name_prefix }}_${{ matrix.folders }}_test_reports/failures_short.txt
-
-      - name: Run test
-        shell: bash
-        run: |
-          mkdir -p reports/${{ env.machine_type }}_${{ inputs.report_name_prefix }}_${{ matrix.folders }}_test_reports
-          echo "hello" > reports/${{ env.machine_type }}_${{ inputs.report_name_prefix }}_${{ matrix.folders }}_test_reports/hello.txt
-          echo "${{ env.machine_type }}_${{ inputs.report_name_prefix }}_${{ matrix.folders }}_test_reports"
-
-      - name: "Test suite reports artifacts: ${{ env.machine_type }}_${{ inputs.report_name_prefix }}_${{ env.matrix_folders }}_test_reports"
-        if: ${{ always() }}
-        uses: actions/upload-artifact@v4
-        with:
-          name: ${{ env.machine_type }}_${{ inputs.report_name_prefix }}_${{ env.matrix_folders }}_test_reports
-          path: reports/${{ env.machine_type }}_${{ inputs.report_name_prefix }}_${{ matrix.folders }}_test_reports

.github/workflows/new_model_pr_merged_notification.yml

@@ -1,68 +0,0 @@
-# Used to notify core maintainers about new model PR being merged
-name: New model PR merged notification
-
-on:
-  push:
-    branches:
-      - main
-    paths:
-      - 'src/transformers/models/*/modeling_*'
-
-jobs:
-  notify_new_model:
-    name: Notify new model
-    runs-on: ubuntu-22.04
-    steps:
-      - uses: actions/checkout@v4
-        with:
-          fetch-depth: 0
-      - name: Check new model
-        shell: bash
-        run: |
-          python -m pip install gitpython
-          python -c 'from utils.pr_slow_ci_models import get_new_model; new_model = get_new_model(diff_with_last_commit=True); print(new_model)' | tee output.txt
-          echo "NEW_MODEL=$(tail -n 1 output.txt)" >> $GITHUB_ENV
-          echo "COMMIT_SHA=$(git log -1 --format=%H)" >> $GITHUB_ENV
-
-      - name: print commit sha
-        if: ${{ env.NEW_MODEL != ''}}
-        shell: bash
-        run: |
-          echo "$COMMIT_SHA"
-
-      - name: print new model
-        if: ${{ env.NEW_MODEL != ''}}
-        shell: bash
-        run: |
-          echo "$NEW_MODEL"
-
-      - name: Notify
-        if: ${{ env.NEW_MODEL != ''}}
-        uses: slackapi/slack-github-action@6c661ce58804a1a20f6dc5fbee7f0381b469e001
-        with:
-          # Slack channel id, channel name, or user id to post message.
-          # See also: https://api.slack.com/methods/chat.postMessage#channels
-          channel-id: transformers-new-model-notification
-          # For posting a rich message using Block Kit
-          payload: |
-            {
-              "blocks": [
-                {
-                  "type": "header",
-                  "text": {
-                    "type": "plain_text",
-                    "text": "New model!",
-                    "emoji": true
-                  }
-                },
-                {
-                  "type": "section",
-                  "text": {
-                    "type": "mrkdwn",
-                    "text": "<https://github.com/huggingface/transformers/commit/${{ env.COMMIT_SHA }}|New model: ${{ env.NEW_MODEL }}> GH_ArthurZucker, GH_lysandrejik, GH_ydshieh\ncommit SHA: ${{ env.COMMIT_SHA }}"
-                  }
-                }
-              ]
-            }
-        env:
-          SLACK_BOT_TOKEN: ${{ secrets.SLACK_CIFEEDBACK_BOT_TOKEN }}

.github/workflows/pr-style-bot.yml

@@ -1,18 +0,0 @@
-# To run this bot, comment "@bot /style" on a PR
-name: Style Bot
-
-on:
-  issue_comment:
-    types: [created]
-
-permissions:
-  pull-requests: write
-
-jobs:
-  style:
-    uses: huggingface/huggingface_hub/.github/workflows/style-bot-action.yml@main
-    with:
-      python_quality_dependencies: "[quality]"
-      style_command_type: "default"
-    secrets:
-      bot_token: ${{ secrets.HF_STYLE_BOT_ACTION }}

.github/workflows/push-important-models.yml

@@ -7,13 +7,14 @@ on:
 env:
   OUTPUT_SLACK_CHANNEL_ID: "C06L2SGMEEA"
   HF_HUB_READ_TOKEN: ${{ secrets.HF_HUB_READ_TOKEN }}
-  HF_HOME: /mnt/cache
-  TRANSFORMERS_IS_CI: yes
-  OMP_NUM_THREADS: 8
-  MKL_NUM_THREADS: 8
-  RUN_SLOW: yes # For gated repositories, we still need to agree to share information on the Hub repo. page in order to get access. # This token is created under the bot `hf-transformers-bot`.
-  SIGOPT_API_TOKEN: ${{ secrets.SIGOPT_API_TOKEN }}
-  TF_FORCE_GPU_ALLOW_GROWTH: true
+  HF_HOME: /mnt/cache 
+  TRANSFORMERS_IS_CI: yes 
+  OMP_NUM_THREADS: 8 
+  MKL_NUM_THREADS: 8 
+  RUN_SLOW: yes # For gated repositories, we still need to agree to share information on the Hub repo. page in order to get access. # This token is created under the bot `hf-transformers-bot`. 
+  SIGOPT_API_TOKEN: ${{ secrets.SIGOPT_API_TOKEN }} 
+  TF_FORCE_GPU_ALLOW_GROWTH: true 
+  RUN_PT_TF_CROSS_TESTS: 1
 
 jobs:
   get_modified_models:
@@ -24,13 +25,13 @@ jobs:
     steps:
       - name: Check out code
         uses: actions/checkout@v4
-
+      
       - name: Get changed files
         id: changed-files
-        uses: tj-actions/changed-files@1c8e6069583811afb28f97afeaf8e7da80c6be5c
+        uses: tj-actions/changed-files@3f54ebb830831fc121d3263c1857cfbdc310cdb9 #v42
         with:
           files: src/transformers/models/**
-
+      
       - name: Run step if only the files listed above change
         if: steps.changed-files.outputs.any_changed == 'true'
         id: set-matrix
@@ -51,49 +52,48 @@ jobs:
   test_modified_files:
     needs: get_modified_models
     name: Slow & FA2 tests
-    runs-on:
-      group: aws-g5-4xlarge-cache
+    runs-on: [single-gpu, nvidia-gpu, a10, ci]
     container:
       image: huggingface/transformers-all-latest-gpu
       options: --gpus all --privileged --ipc host -v /mnt/cache/.cache/huggingface:/mnt/cache/
     if: ${{ needs.get_modified_models.outputs.matrix != '[]' && needs.get_modified_models.outputs.matrix != '' && fromJson(needs.get_modified_models.outputs.matrix)[0] != null }}
     strategy:
       fail-fast: false
-      matrix:
+      matrix: 
         model-name: ${{ fromJson(needs.get_modified_models.outputs.matrix) }}
 
     steps:
       - name: Check out code
         uses: actions/checkout@v4
-
+      
       - name: Install locally transformers & other libs
         run: |
           apt install sudo
           sudo -H pip install --upgrade pip
-          sudo -H pip uninstall -y transformers
-          sudo -H pip install -U -e ".[testing]"
+          sudo -H pip uninstall -y transformers 
+          sudo -H pip install -U -e ".[testing]" 
           MAX_JOBS=4 pip install flash-attn --no-build-isolation
           pip install bitsandbytes
-
+      
       - name: NVIDIA-SMI
         run: |
           nvidia-smi
-
+      
       - name: Show installed libraries and their versions
         run: pip freeze
-
+      
       - name: Run FA2 tests
         id: run_fa2_tests
         run:
           pytest -rsfE -m "flash_attn_test" --make-reports=${{ matrix.model-name }}_fa2_tests/ tests/${{ matrix.model-name }}/test_modeling_*
-
+      
       - name: "Test suite reports artifacts: ${{ matrix.model-name }}_fa2_tests"
         if: ${{ always() }}
         uses: actions/upload-artifact@v4
         with:
           name: ${{ matrix.model-name }}_fa2_tests
           path: /transformers/reports/${{ matrix.model-name }}_fa2_tests
-
+      
       - name: Post to Slack
         if: always()
         uses: huggingface/hf-workflows/.github/actions/post-slack@main
@@ -102,13 +102,13 @@ jobs:
           title: 🤗 Results of the FA2 tests - ${{ matrix.model-name }}
           status: ${{ steps.run_fa2_tests.conclusion}}
           slack_token: ${{ secrets.CI_SLACK_BOT_TOKEN }}
-
+      
       - name: Run integration tests
         id: run_integration_tests
         if: always()
         run:
           pytest -rsfE -k "IntegrationTest"  --make-reports=tests_integration_${{ matrix.model-name }} tests/${{ matrix.model-name }}/test_modeling_*
-
+      
       - name: "Test suite reports artifacts: tests_integration_${{ matrix.model-name }}"
         if: ${{ always() }}
         uses: actions/upload-artifact@v4
@@ -118,7 +118,7 @@ jobs:
 
       - name: Post to Slack
         if: always()
-        uses: huggingface/hf-workflows/.github/actions/post-slack@main
+        uses: huggingface/hf-workflows/.github/actions/post-slack@main 
         with:
           slack_channel: ${{ env.OUTPUT_SLACK_CHANNEL_ID }}
           title: 🤗 Results of the Integration tests - ${{ matrix.model-name }}
@@ -133,3 +133,10 @@ jobs:
           slackChannel: ${{ secrets.SLACK_CIFEEDBACK_CHANNEL }}
           slackToken: ${{ secrets.SLACK_CIFEEDBACK_BOT_TOKEN }}
           waitForSSH: true
+
+  benchmark:
+    name: Benchmark workflow
+    needs: get_modified_models
+    if: ${{ needs.get_modified_models.outputs.matrix != '[]' && needs.get_modified_models.outputs.matrix != '' && fromJson(needs.get_modified_models.outputs.matrix)[0] != null }}
+    uses: ./.github/workflows/benchmark.yml
+    secrets: inherit

.github/workflows/release-conda.yml

@@ -19,7 +19,7 @@ jobs:
 
     steps:
       - name: Checkout repository
-        uses: actions/checkout@v4
+        uses: actions/checkout@v1
 
       - name: Install miniconda
         uses: conda-incubator/setup-miniconda@v2

.github/workflows/self-comment-ci.yml

@@ -1,416 +0,0 @@
-name: PR comment GitHub CI
-
-on:
-  issue_comment:
-    types:
-      - created
-    branches-ignore:
-      - main
-concurrency:
-  group: ${{ github.workflow }}-${{ github.event.issue.number }}-${{ startsWith(github.event.comment.body, 'run-slow') || startsWith(github.event.comment.body, 'run slow') || startsWith(github.event.comment.body, 'run_slow') }}
-  cancel-in-progress: true
-permissions: read-all
-
-env:
-  HF_HOME: /mnt/cache
-  TRANSFORMERS_IS_CI: yes
-  OMP_NUM_THREADS: 8
-  MKL_NUM_THREADS: 8
-  RUN_SLOW: yes
-  # For gated repositories, we still need to agree to share information on the Hub repo. page in order to get access.
-  # This token is created under the bot `hf-transformers-bot`.
-  HF_HUB_READ_TOKEN: ${{ secrets.HF_HUB_READ_TOKEN }}
-  SIGOPT_API_TOKEN: ${{ secrets.SIGOPT_API_TOKEN }}
-  TF_FORCE_GPU_ALLOW_GROWTH: true
-  CUDA_VISIBLE_DEVICES: 0,1
-
-jobs:
-  get-pr-number:
-    runs-on: ubuntu-22.04
-    name: Get PR number
-    # For security: only allow team members to run
-    if: ${{ github.event.issue.state == 'open' && contains(fromJSON('["ydshieh", "ArthurZucker", "zucchini-nlp", "qubvel", "molbap", "gante", "LysandreJik", "Cyrilvallez", "Rocketknight1", "SunMarc", "muellerzr", "eustlb", "MekkCyber", "manueldeprada", "vasqu"]'), github.actor) && (startsWith(github.event.comment.body, 'run-slow') || startsWith(github.event.comment.body, 'run slow') || startsWith(github.event.comment.body, 'run_slow')) }}
-    outputs:
-      PR_NUMBER: ${{ steps.set_pr_number.outputs.PR_NUMBER }}
-    steps:
-      - name: Get PR number
-        shell: bash
-        run: |
-          if [[ "${{ github.event.issue.number }}" != "" && "${{ github.event.issue.pull_request }}" != "" ]]; then
-            echo "PR_NUMBER=${{ github.event.issue.number }}" >> $GITHUB_ENV
-          else
-            echo "PR_NUMBER=" >> $GITHUB_ENV
-          fi
-
-      - name: Check PR number
-        shell: bash
-        run: |
-          echo "${{ env.PR_NUMBER }}"
-
-      - name: Set PR number
-        id: set_pr_number
-        run: echo "PR_NUMBER=${{ env.PR_NUMBER }}" >> "$GITHUB_OUTPUT"
-
-  get-sha:
-    runs-on: ubuntu-22.04
-    needs: get-pr-number
-    if: ${{ needs.get-pr-number.outputs.PR_NUMBER != ''}}
-    outputs:
-      PR_HEAD_SHA: ${{ steps.get_sha.outputs.PR_HEAD_SHA }}
-      PR_MERGE_SHA: ${{ steps.get_sha.outputs.PR_MERGE_SHA }}
-    steps:
-      - uses: actions/checkout@v4
-        with:
-          fetch-depth: "0"
-          ref: "refs/pull/${{needs.get-pr-number.outputs.PR_NUMBER}}/merge"
-
-      - name: Get SHA (and verify timestamps against the issue comment date)
-        id: get_sha
-        env:
-          PR_NUMBER: ${{ needs.get-pr-number.outputs.PR_NUMBER }}
-          COMMENT_DATE: ${{ github.event.comment.created_at }}
-        run: |
-            git fetch origin refs/pull/$PR_NUMBER/head:refs/remotes/pull/$PR_NUMBER/head
-            git checkout refs/remotes/pull/$PR_NUMBER/head
-            echo "PR_HEAD_SHA: $(git log -1 --format=%H)"
-            echo "PR_HEAD_SHA=$(git log -1 --format=%H)" >> "$GITHUB_OUTPUT"
-            git fetch origin refs/pull/$PR_NUMBER/merge:refs/remotes/pull/$PR_NUMBER/merge
-            git checkout refs/remotes/pull/$PR_NUMBER/merge
-            echo "PR_MERGE_SHA: $(git log -1 --format=%H)"
-            echo "PR_MERGE_SHA=$(git log -1 --format=%H)" >> "$GITHUB_OUTPUT"
-            PR_MERGE_COMMIT_TIMESTAMP=$(git log -1 --date=unix --format=%cd)
-            echo "PR_MERGE_COMMIT_TIMESTAMP: $PR_MERGE_COMMIT_TIMESTAMP"
-            COMMENT_TIMESTAMP=$(date -d "${COMMENT_DATE}" +"%s")
-            echo "COMMENT_DATE: $COMMENT_DATE"
-            echo "COMMENT_TIMESTAMP: $COMMENT_TIMESTAMP"
-            if [ $COMMENT_TIMESTAMP -le $PR_MERGE_COMMIT_TIMESTAMP ]; then
-              echo "Last commit on the pull request is newer than the issue comment triggering this run! Abort!";
-              exit -1;
-            fi
-
-  # use a python script to handle this complex logic
-  # case 1: `run-slow` (auto. infer with limited number of models, but in particular, new model)
-  # case 2: `run-slow model_1, model_2`
-  get-tests:
-    runs-on: ubuntu-22.04
-    needs: [get-pr-number, get-sha]
-    if: ${{ needs.get-pr-number.outputs.PR_NUMBER != ''}}
-    outputs:
-      models: ${{ steps.models_to_run.outputs.models }}
-      quantizations: ${{ steps.models_to_run.outputs.quantizations }}
-    steps:
-      - uses: actions/checkout@v4
-        with:
-          fetch-depth: "0"
-          ref: "refs/pull/${{needs.get-pr-number.outputs.PR_NUMBER}}/merge"
-
-      - name: Verify merge commit SHA
-        env:
-          VERIFIED_PR_MERGE_SHA: ${{ needs.get-sha.outputs.PR_MERGE_SHA }}
-        run: |
-            PR_MERGE_SHA=$(git log -1 --format=%H)
-            if [ $PR_MERGE_SHA != $VERIFIED_PR_MERGE_SHA ]; then
-              echo "The merged commit SHA is not the same as the verified one! Security issue detected, abort the workflow!";
-              exit -1;
-            fi
-
-      - name: Get models to test
-        env:
-          PR_COMMENT: ${{ github.event.comment.body }}
-        run: |
-          python -m pip install GitPython
-          python utils/pr_slow_ci_models.py --message "$PR_COMMENT" | tee output.txt
-          echo "models=$(tail -n 1 output.txt)" >> $GITHUB_ENV
-          python utils/pr_slow_ci_models.py --message "$PR_COMMENT" --quantization | tee output2.txt
-          echo "quantizations=$(tail -n 1 output2.txt)" >> $GITHUB_ENV
-
-      - name: Show models to test
-        id: models_to_run
-        run: |
-          echo "${{ env.models }}"
-          echo "models=${{ env.models }}" >> $GITHUB_ENV
-          echo "models=${{ env.models }}" >> $GITHUB_OUTPUT
-          echo "${{ env.quantizations }}"
-          echo "quantizations=${{ env.quantizations }}" >> $GITHUB_OUTPUT
-
-  reply_to_comment:
-    name: Reply to the comment
-    if: ${{ needs.get-tests.outputs.models != '[]'  || needs.get-tests.outputs.quantizations != '[]' }}
-    needs: [get-pr-number, get-tests]
-    permissions:
-      pull-requests: write
-    runs-on: ubuntu-22.04
-    steps:
-      - name: Reply to the comment
-        env:
-          GH_TOKEN: ${{ secrets.GITHUB_TOKEN }}
-          MODELS: ${{ needs.get-tests.outputs.models }}
-          BODY: "\n\nmodels: ${{ needs.get-tests.outputs.models }}\nquantizations: ${{ needs.get-tests.outputs.quantizations }}"
-        run: |
-          gh api \
-            --method POST \
-            -H "Accept: application/vnd.github+json" \
-            -H "X-GitHub-Api-Version: 2022-11-28" \
-            repos/${{ github.repository }}/issues/${{ needs.get-pr-number.outputs.PR_NUMBER }}/comments \
-            -f "body=This comment contains run-slow, running the specified jobs: ${{ env.BODY }} ..."
-
-  create_run:
-    name: Create run
-    if: ${{ needs.get-tests.outputs.models != '[]' || needs.get-tests.outputs.quantizations != '[]' }}
-    needs: [get-sha, get-tests, reply_to_comment]
-    permissions:
-      statuses: write
-    runs-on: ubuntu-22.04
-    steps:
-      - name: Create Run
-        id: create_run
-        env:
-          GH_TOKEN: ${{ secrets.GITHUB_TOKEN }}
-          # Create a commit status (pending) for a run of this workflow. The status has to be updated later in `update_run_status`.
-          # See https://docs.github.com/en/rest/commits/statuses?apiVersion=2022-11-28#create-a-commit-status
-          GITHUB_RUN_URL: https://github.com/${{ github.repository }}/actions/runs/${{ github.run_id }}
-        run: |
-          gh api \
-            --method POST \
-            -H "Accept: application/vnd.github+json" \
-            -H "X-GitHub-Api-Version: 2022-11-28" \
-            repos/${{ github.repository }}/statuses/${{ needs.get-sha.outputs.PR_HEAD_SHA }} \
-            -f "target_url=$GITHUB_RUN_URL" -f "state=pending" -f "description=Slow CI job" -f "context=pytest/custom-tests"
-
-  run_models_gpu:
-    name: Run all tests for the model
-    if: ${{ needs.get-tests.outputs.models != '[]' }}
-    needs: [get-pr-number, get-sha, get-tests, create_run]
-    strategy:
-      fail-fast: false
-      matrix:
-        folders: ${{ fromJson(needs.get-tests.outputs.models) }}
-        machine_type: [aws-g4dn-4xlarge-cache, aws-g4dn-12xlarge-cache]
-    runs-on:
-       group: '${{ matrix.machine_type }}'
-    container:
-      image: huggingface/transformers-all-latest-gpu
-      options: --gpus all --shm-size "16gb" --ipc host -v /mnt/cache/.cache/huggingface:/mnt/cache/
-    steps:
-      - name: Echo input and matrix info
-        shell: bash
-        run: |
-          echo "${{ matrix.folders }}"
-
-      - name: Echo folder ${{ matrix.folders }}
-        shell: bash
-        # For folders like `models/bert`, set an env. var. (`matrix_folders`) to `models_bert`, which will be used to
-        # set the artifact folder names (because the character `/` is not allowed).
-        run: |
-          echo "${{ matrix.folders }}"
-          matrix_folders=${{ matrix.folders }}
-          matrix_folders=${matrix_folders/'models/'/'models_'}
-          echo "$matrix_folders"
-          echo "matrix_folders=$matrix_folders" >> $GITHUB_ENV
-
-      - name: Checkout to PR merge commit
-        working-directory: /transformers
-        run: |
-          git fetch origin refs/pull/${{ needs.get-pr-number.outputs.PR_NUMBER }}/merge:refs/remotes/pull/${{ needs.get-pr-number.outputs.PR_NUMBER }}/merge
-          git checkout refs/remotes/pull/${{ needs.get-pr-number.outputs.PR_NUMBER }}/merge
-          git log -1 --format=%H
-
-      - name: Verify merge commit SHA
-        env:
-          VERIFIED_PR_MERGE_SHA: ${{ needs.get-sha.outputs.PR_MERGE_SHA }}
-        working-directory: /transformers
-        run: |
-          PR_MERGE_SHA=$(git log -1 --format=%H)
-          if [ $PR_MERGE_SHA != $VERIFIED_PR_MERGE_SHA ]; then
-            echo "The merged commit SHA is not the same as the verified one! Security issue detected, abort the workflow!";
-            exit -1;
-          fi
-
-      - name: Reinstall transformers in edit mode (remove the one installed during docker image build)
-        working-directory: /transformers
-        run: python3 -m pip uninstall -y transformers && python3 -m pip install -e .
-
-      - name: NVIDIA-SMI
-        run: |
-          nvidia-smi
-
-      - name: Set `machine_type` for report and artifact names
-        working-directory: /transformers
-        shell: bash
-        run: |
-          echo "${{ matrix.machine_type }}"
-          if [ "${{ matrix.machine_type }}" = "aws-g4dn-4xlarge-cache" ]; then
-            machine_type=single-gpu
-          elif [ "${{ matrix.machine_type }}" = "aws-g4dn-12xlarge-cache" ]; then
-            machine_type=multi-gpu
-          else
-            machine_type=${{ matrix.machine_type }}
-          fi
-          echo "$machine_type"
-          echo "machine_type=$machine_type" >> $GITHUB_ENV
-
-      - name: Environment
-        working-directory: /transformers
-        run: |
-          python3 utils/print_env.py
-
-      - name: Show installed libraries and their versions
-        working-directory: /transformers
-        run: pip freeze
-
-      - name: Run all tests on GPU
-        working-directory: /transformers
-        run: |
-          export CUDA_VISIBLE_DEVICES="$(python3 utils/set_cuda_devices_for_ci.py --test_folder ${{ matrix.folders }})"
-          echo $CUDA_VISIBLE_DEVICES
-          python3 -m pytest -v -rsfE --make-reports=${{ env.machine_type }}_run_models_gpu_${{ matrix.folders }}_test_reports tests/${{ matrix.folders }}
-
-      - name: Failure short reports
-        if: ${{ failure() }}
-        continue-on-error: true
-        run: cat /transformers/reports/${{ env.machine_type }}_run_models_gpu_${{ matrix.folders }}_test_reports/failures_short.txt
-
-      - name: Make sure report directory exists
-        shell: bash
-        run: |
-          mkdir -p /transformers/reports/${{ env.machine_type }}_run_models_gpu_${{ matrix.folders }}_test_reports
-          echo "hello" > /transformers/reports/${{ env.machine_type }}_run_models_gpu_${{ matrix.folders }}_test_reports/hello.txt
-          echo "${{ env.machine_type }}_run_models_gpu_${{ matrix.folders }}_test_reports"
-
-      - name: "Test suite reports artifacts: ${{ env.machine_type }}_run_models_gpu_${{ env.matrix_folders }}_test_reports"
-        if: ${{ always() }}
-        uses: actions/upload-artifact@v4
-        with:
-          name: ${{ env.machine_type }}_run_models_gpu_${{ env.matrix_folders }}_test_reports
-          path: /transformers/reports/${{ env.machine_type }}_run_models_gpu_${{ matrix.folders }}_test_reports
-
-  run_quantization_torch_gpu:
-    name: Run all tests for a quantization
-    if: ${{ needs.get-tests.outputs.quantizations != '[]' }}
-    needs: [get-pr-number, get-sha, get-tests, create_run]
-    strategy:
-      fail-fast: false
-      matrix:
-        folders: ${{ fromJson(needs.get-tests.outputs.quantizations) }}
-        machine_type: [aws-g4dn-4xlarge-cache, aws-g4dn-12xlarge-cache]
-    runs-on:
-      group: '${{ matrix.machine_type }}'
-    container:
-      image: huggingface/transformers-quantization-latest-gpu
-      options: --gpus all --shm-size "16gb" --ipc host -v /mnt/cache/.cache/huggingface:/mnt/cache/
-    steps:
-      - name: Echo folder ${{ matrix.folders }}
-        shell: bash
-        run: |
-          echo "${{ matrix.folders }}"
-          matrix_folders=${{ matrix.folders }}
-          matrix_folders=${matrix_folders/'quantization/'/'quantization_'}
-          echo "$matrix_folders"
-          echo "matrix_folders=$matrix_folders" >> $GITHUB_ENV
-
-      - name: Checkout to PR merge commit
-        working-directory: /transformers
-        run: |
-          git fetch origin refs/pull/${{ needs.get-pr-number.outputs.PR_NUMBER }}/merge:refs/remotes/pull/${{ needs.get-pr-number.outputs.PR_NUMBER }}/merge
-          git checkout refs/remotes/pull/${{ needs.get-pr-number.outputs.PR_NUMBER }}/merge
-          git log -1 --format=%H
-
-      - name: Verify merge commit SHA
-        env:
-          VERIFIED_PR_MERGE_SHA: ${{ needs.get-sha.outputs.PR_MERGE_SHA }}
-        working-directory: /transformers
-        run: |
-          PR_MERGE_SHA=$(git log -1 --format=%H)
-          if [ $PR_MERGE_SHA != $VERIFIED_PR_MERGE_SHA ]; then
-            echo "The merged commit SHA is not the same as the verified one! Security issue detected, abort the workflow!";
-            exit -1;
-          fi
-
-      - name: Reinstall transformers in edit mode (remove the one installed during docker image build)
-        working-directory: /transformers
-        run: python3 -m pip uninstall -y transformers && python3 -m pip install -e .
-      - name: NVIDIA-SMI
-        run: |
-          nvidia-smi
-
-      - name: Set `machine_type` for report and artifact names
-        working-directory: /transformers
-        shell: bash
-        run: |
-          echo "${{ matrix.machine_type }}"
-          if [ "${{ matrix.machine_type }}" = "aws-g4dn-4xlarge-cache" ]; then
-            machine_type=single-gpu
-          elif [ "${{ matrix.machine_type }}" = "aws-g4dn-12xlarge-cache" ]; then
-            machine_type=multi-gpu
-          else
-            machine_type=${{ matrix.machine_type }}
-          fi
-          echo "$machine_type"
-          echo "machine_type=$machine_type" >> $GITHUB_ENV
-
-      - name: Environment
-        working-directory: /transformers
-        run: |
-          python3 utils/print_env.py
-
-      - name: Show installed libraries and their versions
-        working-directory: /transformers
-        run: pip freeze
-
-      - name: Run quantization tests on GPU
-        working-directory: /transformers
-        run: |
-          python3 -m pytest -v --make-reports=${{ env.machine_type }}_run_quantization_torch_gpu_${{ matrix.folders }}_test_reports tests/${{ matrix.folders }}
-
-      - name: Failure short reports
-        if: ${{ failure() }}
-        continue-on-error: true
-        run: cat /transformers/reports/${{ env.machine_type }}_run_quantization_torch_gpu_${{ matrix.folders }}_test_reports/failures_short.txt
-
-      - name: Make sure report directory exists
-        shell: bash
-        run: |
-          mkdir -p /transformers/reports/${{ env.machine_type }}_run_quantization_gpu_${{ matrix.folders }}_test_reports
-          echo "hello" > /transformers/reports/${{ env.machine_type }}_run_quantization_gpu_${{ matrix.folders }}_test_reports/hello.txt
-          echo "${{ env.machine_type }}_run_quantization_gpu_${{ matrix.folders }}_test_reports"
-
-      - name: "Test suite reports artifacts: ${{ env.machine_type }}_run_quantization_torch_gpu_${{ env.matrix_folders }}_test_reports"
-        if: ${{ always() }}
-        uses: actions/upload-artifact@v4
-        with:
-          name: ${{ env.machine_type }}_run_quantization_torch_gpu_${{ env.matrix_folders }}_test_reports
-          path: /transformers/reports/${{ env.machine_type }}_run_quantization_torch_gpu_${{ matrix.folders }}_test_reports
-
-  update_run_status:
-    name: Update Check Run Status
-    needs: [get-sha, create_run, run_models_gpu, run_quantization_torch_gpu]
-    permissions:
-      statuses: write
-    if: ${{ always() && needs.create_run.result == 'success' }}
-    runs-on: ubuntu-22.04
-    env:
-      GH_TOKEN: ${{ secrets.GITHUB_TOKEN }}
-      GITHUB_RUN_URL: https://github.com/${{ github.repository }}/actions/runs/${{ github.run_id }}
-      STATUS_OK: ${{ contains(fromJSON('["skipped", "success"]'), needs.run_models_gpu.result) && contains(fromJSON('["skipped", "success"]'), needs.run_quantization_torch_gpu.result) }}
-    steps:
-      - name: Get `run_models_gpu` job status
-        run: |
-          echo "${{ needs.run_models_gpu.result }}"
-          echo "${{ needs.run_quantization_torch_gpu.result }}"
-          echo $STATUS_OK
-          if [ "$STATUS_OK" = "true" ]; then
-            echo "STATUS=success" >> $GITHUB_ENV
-          else
-            echo "STATUS=failure" >> $GITHUB_ENV
-          fi
-
-      - name: Update PR commit statuses
-        run: |
-          echo "${{ needs.run_models_gpu.result }}"
-          echo "${{ env.STATUS }}"
-          gh api \
-            --method POST \
-            -H "Accept: application/vnd.github+json" \
-            -H "X-GitHub-Api-Version: 2022-11-28" \
-            repos/${{ github.repository }}/statuses/${{ needs.get-sha.outputs.PR_HEAD_SHA }} \
-            -f "target_url=$GITHUB_RUN_URL" -f "state=${{ env.STATUS }}" -f "description=Slow CI job" -f "context=pytest/custom-tests"

.github/workflows/self-nightly-past-ci-caller.yml

@@ -21,6 +21,39 @@ jobs:
           echo "$(python3 -c 'print(int(${{ github.run_number }}) % 10)')"
           echo "run_number=$(python3 -c 'print(int(${{ github.run_number }}) % 10)')" >> $GITHUB_OUTPUT
 
+  run_past_ci_pytorch_1-13:
+    name: PyTorch 1.13
+    needs: get_number
+    if: needs.get_number.outputs.run_number == 0 && (cancelled() != true) && ((github.event_name == 'schedule') || ((github.event_name == 'push') && startsWith(github.ref_name, 'run_past_ci')))
+    uses: ./.github/workflows/self-past-caller.yml
+    with:
+      framework: pytorch
+      version: "1.13"
+      sha: ${{ github.sha }}
+    secrets: inherit
+
+  run_past_ci_pytorch_1-12:
+    name: PyTorch 1.12
+    needs: get_number
+    if: needs.get_number.outputs.run_number == 1 && (cancelled() != true) && ((github.event_name == 'schedule') || ((github.event_name == 'push') && startsWith(github.ref_name, 'run_past_ci')))
+    uses: ./.github/workflows/self-past-caller.yml
+    with:
+      framework: pytorch
+      version: "1.12"
+      sha: ${{ github.sha }}
+    secrets: inherit
+
+  run_past_ci_pytorch_1-11:
+    name: PyTorch 1.11
+    needs: get_number
+    if: needs.get_number.outputs.run_number == 2 && (cancelled() != true) && ((github.event_name == 'schedule') || ((github.event_name == 'push') && startsWith(github.ref_name, 'run_past_ci')))
+    uses: ./.github/workflows/self-past-caller.yml
+    with:
+      framework: pytorch
+      version: "1.11"
+      sha: ${{ github.sha }}
+    secrets: inherit
+
   run_past_ci_tensorflow_2-11:
     name: TensorFlow 2.11
     needs: get_number

.github/workflows/self-pr-slow-ci.yml

@@ -0,0 +1,135 @@
+name: PR slow CI
+
+on:
+  pull_request:
+    paths:
+      - "src/transformers/models/*/modeling_*.py"
+      - "tests/**/test_*.py"
+
+concurrency:
+  group: ${{ github.workflow }}-${{ github.head_ref || github.run_id }}
+  cancel-in-progress: true
+
+env:
+  HF_HOME: /mnt/cache
+  TRANSFORMERS_IS_CI: yes
+  OMP_NUM_THREADS: 8
+  MKL_NUM_THREADS: 8
+  RUN_SLOW: yes
+  # For gated repositories, we still need to agree to share information on the Hub repo. page in order to get access.
+  # This token is created under the bot `hf-transformers-bot`.
+  HF_HUB_READ_TOKEN: ${{ secrets.HF_HUB_READ_TOKEN }}
+  SIGOPT_API_TOKEN: ${{ secrets.SIGOPT_API_TOKEN }}
+  TF_FORCE_GPU_ALLOW_GROWTH: true
+  RUN_PT_TF_CROSS_TESTS: 1
+  CUDA_VISIBLE_DEVICES: 0,1
+
+jobs:
+  find_models_to_run:
+      runs-on: ubuntu-22.04
+      name: Find models to run slow tests
+      # Triggered only if the required label `run-slow` is added
+      if: ${{ contains(github.event.pull_request.labels.*.name, 'run-slow') }}
+      outputs:
+        models: ${{ steps.models_to_run.outputs.models }}
+      steps:
+        - uses: actions/checkout@v4
+          with:
+            fetch-depth: "0"
+            ref: ${{ github.event.pull_request.head.sha }}
+
+        - name: Get commit message
+          run: |
+            echo "commit_message=$(git show -s --format=%s)" >> $GITHUB_ENV
+
+        - name: Get models to run slow tests
+          run: |
+            echo "${{ env.commit_message }}"
+            python -m pip install GitPython
+            python utils/pr_slow_ci_models.py --commit_message "${{ env.commit_message }}" | tee output.txt
+            echo "models=$(tail -n 1 output.txt)" >> $GITHUB_ENV
+
+        - name: Models to run slow tests
+          id: models_to_run
+          run: |
+            echo "${{ env.models }}"
+            echo "models=${{ env.models }}" >> $GITHUB_OUTPUT
+
+  run_models_gpu:
+      name: Run all tests for the model
+      # Triggered only `find_models_to_run` is triggered (label `run-slow` is added) which gives the models to run
+      # (either a new model PR or via a commit message)
+      if: ${{ needs.find_models_to_run.outputs.models != '[]' }}
+      needs: find_models_to_run
+      strategy:
+        fail-fast: false
+        matrix:
+          folders: ${{ fromJson(needs.find_models_to_run.outputs.models) }}
+          machine_type: [single-gpu, multi-gpu]
+      runs-on: ['${{ matrix.machine_type }}', nvidia-gpu, t4, ci]
+      container:
+        image: huggingface/transformers-all-latest-gpu
+        options: --gpus all --shm-size "16gb" --ipc host -v /mnt/cache/.cache/huggingface:/mnt/cache/
+      steps:
+      - name: Echo input and matrix info
+        shell: bash
+        run: |
+          echo "${{ matrix.folders }}"
+
+      - name: Echo folder ${{ matrix.folders }}
+        shell: bash
+        # For folders like `models/bert`, set an env. var. (`matrix_folders`) to `models_bert`, which will be used to
+        # set the artifact folder names (because the character `/` is not allowed).
+        run: |
+          echo "${{ matrix.folders }}"
+          matrix_folders=${{ matrix.folders }}
+          matrix_folders=${matrix_folders/'models/'/'models_'}
+          echo "$matrix_folders"
+          echo "matrix_folders=$matrix_folders" >> $GITHUB_ENV
+
+      - name: Update clone
+        working-directory: /transformers
+        run: git fetch && git fetch origin pull/${{ github.event.pull_request.number }}/head:pull/${{ github.event.pull_request.number }}/merge && git checkout pull/${{ github.event.pull_request.number }}/merge
+
+      - name: Reinstall transformers in edit mode (remove the one installed during docker image build)
+        working-directory: /transformers
+        run: python3 -m pip uninstall -y transformers && python3 -m pip install -e .
+
+      - name: NVIDIA-SMI
+        run: |
+          nvidia-smi
+
+      - name: Environment
+        working-directory: /transformers
+        run: |
+          python3 utils/print_env.py
+
+      - name: Show installed libraries and their versions
+        working-directory: /transformers
+        run: pip freeze
+
+      - name: Run all tests on GPU
+        working-directory: /transformers
+        run: |
+          export CUDA_VISIBLE_DEVICES="$(python3 utils/set_cuda_devices_for_ci.py --test_folder ${{ matrix.folders }})"
+          echo $CUDA_VISIBLE_DEVICES
+          python3 -m pytest -v -rsfE --make-reports=${{ matrix.machine_type }}_run_models_gpu_${{ matrix.folders }}_test_reports tests/${{ matrix.folders }}
+
+      - name: Failure short reports
+        if: ${{ failure() }}
+        continue-on-error: true
+        run: cat /transformers/reports/${{ matrix.machine_type }}_run_models_gpu_${{ matrix.folders }}_test_reports/failures_short.txt
+
+      - name: Make sure report directory exists
+        shell: bash
+        run: |
+          mkdir -p /transformers/reports/${{ matrix.machine_type }}_run_models_gpu_${{ matrix.folders }}_test_reports
+          echo "hello" > /transformers/reports/${{ matrix.machine_type }}_run_models_gpu_${{ matrix.folders }}_test_reports/hello.txt
+          echo "${{ matrix.machine_type }}_run_models_gpu_${{ matrix.folders }}_test_reports"
+
+      - name: "Test suite reports artifacts: ${{ matrix.machine_type }}_run_models_gpu_${{ env.matrix_folders }}_test_reports"
+        if: ${{ always() }}
+        uses: actions/upload-artifact@v4
+        with:
+          name: ${{ matrix.machine_type }}_run_models_gpu_${{ env.matrix_folders }}_test_reports
+          path: /transformers/reports/${{ matrix.machine_type }}_run_models_gpu_${{ matrix.folders }}_test_reports

.github/workflows/self-push-amd-mi210-caller.yml

@@ -1,25 +1,25 @@
-name: Self-hosted runner (AMD mi210 CI caller)
-
-on:
-  #workflow_run:
-  #  workflows: ["Self-hosted runner (push-caller)"]
-  #  branches: ["main"]
-  #  types: [completed]
-  push:
-    branches:
-      - run_amd_push_ci_caller*
-    paths:
-      - "src/**"
-      - "tests/**"
-      - ".github/**"
-      - "templates/**"
-      - "utils/**"
-
-jobs:
-  run_amd_ci:
-    name: AMD mi210
-    if: (cancelled() != true) && ((github.event_name == 'workflow_run') || ((github.event_name == 'push') && startsWith(github.ref_name, 'run_amd_push_ci_caller')))
-    uses: ./.github/workflows/self-push-amd.yml
-    with:
-      gpu_flavor: mi210
-    secrets: inherit
+name: Self-hosted runner (AMD mi210 CI caller)
+
+on:
+  workflow_run:
+    workflows: ["Self-hosted runner (push-caller)"]
+    branches: ["main"]
+    types: [completed]
+  push:
+    branches:
+      - run_amd_push_ci_caller*
+    paths:
+      - "src/**"
+      - "tests/**"
+      - ".github/**"
+      - "templates/**"
+      - "utils/**"
+
+jobs:
+  run_amd_ci:
+    name: AMD mi210
+    if: (cancelled() != true) && ((github.event_name == 'workflow_run') || ((github.event_name == 'push') && startsWith(github.ref_name, 'run_amd_push_ci_caller')))
+    uses: ./.github/workflows/self-push-amd.yml
+    with:
+      gpu_flavor: mi210
+    secrets: inherit

.github/workflows/self-push-amd-mi250-caller.yml

@@ -1,25 +1,25 @@
-name: Self-hosted runner (AMD mi250 CI caller)
-
-on:
-  #workflow_run:
-  #  workflows: ["Self-hosted runner (push-caller)"]
-  #  branches: ["main"]
-  #  types: [completed]
-  push:
-    branches:
-      - run_amd_push_ci_caller*
-    paths:
-      - "src/**"
-      - "tests/**"
-      - ".github/**"
-      - "templates/**"
-      - "utils/**"
-
-jobs:
-  run_amd_ci:
-    name: AMD mi250
-    if: (cancelled() != true) && ((github.event_name == 'workflow_run') || ((github.event_name == 'push') && startsWith(github.ref_name, 'run_amd_push_ci_caller')))
-    uses: ./.github/workflows/self-push-amd.yml
-    with:
-      gpu_flavor: mi250
-    secrets: inherit
+name: Self-hosted runner (AMD mi250 CI caller)
+
+on:
+  workflow_run:
+    workflows: ["Self-hosted runner (push-caller)"]
+    branches: ["main"]
+    types: [completed]
+  push:
+    branches:
+      - run_amd_push_ci_caller*
+    paths:
+      - "src/**"
+      - "tests/**"
+      - ".github/**"
+      - "templates/**"
+      - "utils/**"
+
+jobs:
+  run_amd_ci:
+    name: AMD mi250
+    if: (cancelled() != true) && ((github.event_name == 'workflow_run') || ((github.event_name == 'push') && startsWith(github.ref_name, 'run_amd_push_ci_caller')))
+    uses: ./.github/workflows/self-push-amd.yml
+    with:
+      gpu_flavor: mi250
+    secrets: inherit

.github/workflows/self-push-amd-mi300-caller.yml

@@ -1,10 +1,10 @@
 name: Self-hosted runner (AMD mi300 CI caller)
 
 on:
-  #workflow_run:
-  #  workflows: ["Self-hosted runner (push-caller)"]
-  #  branches: ["main"]
-  #  types: [completed]
+  workflow_run:
+    workflows: ["Self-hosted runner (push-caller)"]
+    branches: ["main"]
+    types: [completed]
   push:
     branches:
       - run_amd_push_ci_caller*

.github/workflows/self-push-amd.yml

@@ -14,6 +14,7 @@ env:
   MKL_NUM_THREADS: 8
   PYTEST_TIMEOUT: 60
   TF_FORCE_GPU_ALLOW_GROWTH: true
+  RUN_PT_TF_CROSS_TESTS: 1
   HF_HUB_READ_TOKEN: ${{ secrets.HF_HUB_READ_TOKEN }}
 
 jobs:
@@ -63,24 +64,23 @@ jobs:
     outputs:
       matrix: ${{ steps.set-matrix.outputs.matrix }}
       test_map: ${{ steps.set-matrix.outputs.test_map }}
-    env:
-      # `CI_BRANCH_PUSH`: The branch name from the push event
-      # `CI_BRANCH_WORKFLOW_RUN`: The name of the branch on which this workflow is triggered by `workflow_run` event
-      # `CI_SHA_PUSH`: The commit SHA from the push event
-      # `CI_SHA_WORKFLOW_RUN`: The commit SHA that triggers this workflow by `workflow_run` event
-      CI_BRANCH_PUSH: ${{ github.event.ref }}
-      CI_BRANCH_WORKFLOW_RUN: ${{ github.event.workflow_run.head_branch }}
-      CI_SHA_PUSH: ${{ github.event.head_commit.id }}
-      CI_SHA_WORKFLOW_RUN: ${{ github.event.workflow_run.head_sha }}
     steps:
       # Necessary to get the correct branch name and commit SHA for `workflow_run` event
       # We also take into account the `push` event (we might want to test some changes in a branch)
       - name: Prepare custom environment variables
         shell: bash
+        # `CI_BRANCH_PUSH`: The branch name from the push event
+        # `CI_BRANCH_WORKFLOW_RUN`: The name of the branch on which this workflow is triggered by `workflow_run` event
         # `CI_BRANCH`: The non-empty branch name from the above two (one and only one of them is empty)
+        # `CI_SHA_PUSH`: The commit SHA from the push event
+        # `CI_SHA_WORKFLOW_RUN`: The commit SHA that triggers this workflow by `workflow_run` event
         # `CI_SHA`: The non-empty commit SHA from the above two (one and only one of them is empty)
         run: |
+          CI_BRANCH_PUSH=${{ github.event.ref }}
           CI_BRANCH_PUSH=${CI_BRANCH_PUSH/'refs/heads/'/''}
+          CI_BRANCH_WORKFLOW_RUN=${{ github.event.workflow_run.head_branch }}
+          CI_SHA_PUSH=${{ github.event.head_commit.id }}
+          CI_SHA_WORKFLOW_RUN=${{ github.event.workflow_run.head_sha }}
           echo $CI_BRANCH_PUSH
           echo $CI_BRANCH_WORKFLOW_RUN
           echo $CI_SHA_PUSH
@@ -159,12 +159,6 @@ jobs:
     container:
       image: huggingface/transformers-pytorch-amd-gpu-push-ci  # <--- We test only for PyTorch for now
       options: --device /dev/kfd --device /dev/dri --env ROCR_VISIBLE_DEVICES --shm-size "16gb" --ipc host -v /mnt/cache/.cache/huggingface:/mnt/cache/
-    env:
-      # For the meaning of these environment variables, see the job `Setup`
-      CI_BRANCH_PUSH: ${{ github.event.ref }}
-      CI_BRANCH_WORKFLOW_RUN: ${{ github.event.workflow_run.head_branch }}
-      CI_SHA_PUSH: ${{ github.event.head_commit.id }}
-      CI_SHA_WORKFLOW_RUN: ${{ github.event.workflow_run.head_sha }}
     steps:
       # Necessary to get the correct branch name and commit SHA for `workflow_run` event
       # We also take into account the `push` event (we might want to test some changes in a branch)
@@ -172,7 +166,11 @@ jobs:
         shell: bash
         # For the meaning of these environment variables, see the job `Setup`
         run: |
+          CI_BRANCH_PUSH=${{ github.event.ref }}
           CI_BRANCH_PUSH=${CI_BRANCH_PUSH/'refs/heads/'/''}
+          CI_BRANCH_WORKFLOW_RUN=${{ github.event.workflow_run.head_branch }}
+          CI_SHA_PUSH=${{ github.event.head_commit.id }}
+          CI_SHA_WORKFLOW_RUN=${{ github.event.workflow_run.head_sha }}
           echo $CI_BRANCH_PUSH
           echo $CI_BRANCH_WORKFLOW_RUN
           echo $CI_SHA_PUSH
@@ -258,12 +256,6 @@ jobs:
 #        run_tests_torch_cuda_extensions_single_gpu,
 #        run_tests_torch_cuda_extensions_multi_gpu
     ]
-    env:
-      # For the meaning of these environment variables, see the job `Setup`
-      CI_BRANCH_PUSH: ${{ github.event.ref }}
-      CI_BRANCH_WORKFLOW_RUN: ${{ github.event.workflow_run.head_branch }}
-      CI_SHA_PUSH: ${{ github.event.head_commit.id }}
-      CI_SHA_WORKFLOW_RUN: ${{ github.event.workflow_run.head_sha }}
     steps:
       - name: Preliminary job status
         shell: bash
@@ -279,7 +271,11 @@ jobs:
         shell: bash
         # For the meaning of these environment variables, see the job `Setup`
         run: |
+          CI_BRANCH_PUSH=${{ github.event.ref }}
           CI_BRANCH_PUSH=${CI_BRANCH_PUSH/'refs/heads/'/''}
+          CI_BRANCH_WORKFLOW_RUN=${{ github.event.workflow_run.head_branch }}
+          CI_SHA_PUSH=${{ github.event.head_commit.id }}
+          CI_SHA_WORKFLOW_RUN=${{ github.event.workflow_run.head_sha }}
           echo $CI_BRANCH_PUSH
           echo $CI_BRANCH_WORKFLOW_RUN
           echo $CI_SHA_PUSH
@@ -328,7 +324,6 @@ jobs:
         # We pass `needs.setup_gpu.outputs.matrix` as the argument. A processing in `notification_service.py` to change
         # `models/bert` to `models_bert` is required, as the artifact names use `_` instead of `/`.
         run: |
-          pip install huggingface_hub
           pip install slack_sdk
           pip show slack_sdk
           python utils/notification_service.py "${{ needs.setup_gpu.outputs.matrix }}"

.github/workflows/self-push-caller.yml

@@ -25,7 +25,7 @@ jobs:
         
         - name: Get changed files
           id: changed-files
-          uses: tj-actions/changed-files@1c8e6069583811afb28f97afeaf8e7da80c6be5c
+          uses: tj-actions/changed-files@v41
         
         - name: Was setup changed 
           id: was_changed
@@ -51,4 +51,4 @@ jobs:
     needs: build-docker-containers
     steps:
       - name: Trigger push CI via workflow_run
-        run: echo "Trigger push CI via workflow_run"
+        run: echo "Trigger push CI via workflow_run"

.github/workflows/self-push.yml

@@ -24,6 +24,7 @@ env:
   MKL_NUM_THREADS: 8
   PYTEST_TIMEOUT: 60
   TF_FORCE_GPU_ALLOW_GROWTH: true
+  RUN_PT_TF_CROSS_TESTS: 1
   CUDA_VISIBLE_DEVICES: 0,1
 
 jobs:
@@ -31,33 +32,31 @@ jobs:
     name: Setup
     strategy:
       matrix:
-        machine_type: [aws-g4dn-2xlarge-cache, aws-g4dn-12xlarge-cache]
-    runs-on:
-      group: '${{ matrix.machine_type }}'
+        machine_type: [single-gpu, multi-gpu]
+    runs-on: ['${{ matrix.machine_type }}', nvidia-gpu, t4, push-ci]
     container:
       image: huggingface/transformers-all-latest-gpu-push-ci
       options: --gpus 0 --shm-size "16gb" --ipc host -v /mnt/cache/.cache/huggingface:/mnt/cache/
     outputs:
       matrix: ${{ steps.set-matrix.outputs.matrix }}
       test_map: ${{ steps.set-matrix.outputs.test_map }}
-    env:
-      # `CI_BRANCH_PUSH`: The branch name from the push event
-      # `CI_BRANCH_WORKFLOW_RUN`: The name of the branch on which this workflow is triggered by `workflow_run` event
-      # `CI_SHA_PUSH`: The commit SHA from the push event
-      # `CI_SHA_WORKFLOW_RUN`: The commit SHA that triggers this workflow by `workflow_run` event
-      CI_BRANCH_PUSH: ${{ github.event.ref }}
-      CI_BRANCH_WORKFLOW_RUN: ${{ github.event.workflow_run.head_branch }}
-      CI_SHA_PUSH: ${{ github.event.head_commit.id }}
-      CI_SHA_WORKFLOW_RUN: ${{ github.event.workflow_run.head_sha }}
     steps:
       # Necessary to get the correct branch name and commit SHA for `workflow_run` event
       # We also take into account the `push` event (we might want to test some changes in a branch)
       - name: Prepare custom environment variables
         shell: bash
+        # `CI_BRANCH_PUSH`: The branch name from the push event
+        # `CI_BRANCH_WORKFLOW_RUN`: The name of the branch on which this workflow is triggered by `workflow_run` event
         # `CI_BRANCH`: The non-empty branch name from the above two (one and only one of them is empty)
+        # `CI_SHA_PUSH`: The commit SHA from the push event
+        # `CI_SHA_WORKFLOW_RUN`: The commit SHA that triggers this workflow by `workflow_run` event
         # `CI_SHA`: The non-empty commit SHA from the above two (one and only one of them is empty)
         run: |
+          CI_BRANCH_PUSH=${{ github.event.ref }}
           CI_BRANCH_PUSH=${CI_BRANCH_PUSH/'refs/heads/'/''}
+          CI_BRANCH_WORKFLOW_RUN=${{ github.event.workflow_run.head_branch }}
+          CI_SHA_PUSH=${{ github.event.head_commit.id }}
+          CI_SHA_WORKFLOW_RUN=${{ github.event.workflow_run.head_sha }}
           echo $CI_BRANCH_PUSH
           echo $CI_BRANCH_WORKFLOW_RUN
           echo $CI_SHA_PUSH
@@ -131,18 +130,11 @@ jobs:
       fail-fast: false
       matrix:
         folders: ${{ fromJson(needs.setup.outputs.matrix) }}
-        machine_type: [aws-g4dn-2xlarge-cache]
-    runs-on:
-      group: '${{ matrix.machine_type }}'
+        machine_type: [single-gpu]
+    runs-on: ['${{ matrix.machine_type }}', nvidia-gpu, t4, push-ci]
     container:
       image: huggingface/transformers-all-latest-gpu-push-ci
       options: --gpus 0 --shm-size "16gb" --ipc host -v /mnt/cache/.cache/huggingface:/mnt/cache/
-    env:
-      # For the meaning of these environment variables, see the job `Setup`
-      CI_BRANCH_PUSH: ${{ github.event.ref }}
-      CI_BRANCH_WORKFLOW_RUN: ${{ github.event.workflow_run.head_branch }}
-      CI_SHA_PUSH: ${{ github.event.head_commit.id }}
-      CI_SHA_WORKFLOW_RUN: ${{ github.event.workflow_run.head_sha }}
     steps:
       # Necessary to get the correct branch name and commit SHA for `workflow_run` event
       # We also take into account the `push` event (we might want to test some changes in a branch)
@@ -150,7 +142,11 @@ jobs:
         shell: bash
         # For the meaning of these environment variables, see the job `Setup`
         run: |
+          CI_BRANCH_PUSH=${{ github.event.ref }}
           CI_BRANCH_PUSH=${CI_BRANCH_PUSH/'refs/heads/'/''}
+          CI_BRANCH_WORKFLOW_RUN=${{ github.event.workflow_run.head_branch }}
+          CI_SHA_PUSH=${{ github.event.head_commit.id }}
+          CI_SHA_WORKFLOW_RUN=${{ github.event.workflow_run.head_sha }}
           echo $CI_BRANCH_PUSH
           echo $CI_BRANCH_WORKFLOW_RUN
           echo $CI_SHA_PUSH
@@ -163,23 +159,6 @@ jobs:
           echo "env.CI_BRANCH = ${{ env.CI_BRANCH }}"
           echo "env.CI_SHA = ${{ env.CI_SHA }}"
 
-      - name: Set `machine_type` for report and artifact names
-        working-directory: /transformers
-        shell: bash
-        run: |
-          echo "${{ matrix.machine_type }}"
-
-          if [ "${{ matrix.machine_type }}" = "aws-g4dn-2xlarge-cache" ]; then
-            machine_type=single-gpu
-          elif [ "${{ matrix.machine_type }}" = "aws-g4dn-12xlarge-cache" ]; then
-            machine_type=multi-gpu
-          else
-            machine_type=${{ matrix.machine_type }}
-          fi
-
-          echo "$machine_type"
-          echo "machine_type=$machine_type" >> $GITHUB_ENV
-
       - name: Update clone using environment variables
         working-directory: /transformers
         run: |
@@ -221,19 +200,19 @@ jobs:
       - name: Run all non-slow selected tests on GPU
         working-directory: /transformers
         run: |
-          python3 -m pytest -n 2 --dist=loadfile -v --make-reports=${{ env.machine_type }}_tests_gpu_${{ matrix.folders }} ${{ fromJson(needs.setup.outputs.test_map)[matrix.folders] }}
+          python3 -m pytest -n 2 --dist=loadfile -v --make-reports=${{ matrix.machine_type }}_tests_gpu_${{ matrix.folders }} ${{ fromJson(needs.setup.outputs.test_map)[matrix.folders] }}
 
       - name: Failure short reports
         if: ${{ failure() }}
         continue-on-error: true
-        run: cat /transformers/reports/${{ env.machine_type }}_tests_gpu_${{ matrix.folders }}/failures_short.txt
+        run: cat /transformers/reports/${{ matrix.machine_type }}_tests_gpu_${{ matrix.folders }}/failures_short.txt
 
-      - name: "Test suite reports artifacts: ${{ env.machine_type }}_run_all_tests_gpu_${{ env.matrix_folders }}_test_reports"
+      - name: "Test suite reports artifacts: ${{ matrix.machine_type }}_run_all_tests_gpu_${{ env.matrix_folders }}_test_reports"
         if: ${{ always() }}
         uses: actions/upload-artifact@v4
         with:
-          name: ${{ env.machine_type }}_run_all_tests_gpu_${{ env.matrix_folders }}_test_reports
-          path: /transformers/reports/${{ env.machine_type }}_tests_gpu_${{ matrix.folders }}
+          name: ${{ matrix.machine_type }}_run_all_tests_gpu_${{ env.matrix_folders }}_test_reports
+          path: /transformers/reports/${{ matrix.machine_type }}_tests_gpu_${{ matrix.folders }}
 
   run_tests_multi_gpu:
     name: Model tests
@@ -244,18 +223,11 @@ jobs:
       fail-fast: false
       matrix:
         folders: ${{ fromJson(needs.setup.outputs.matrix) }}
-        machine_type: [aws-g4dn-12xlarge-cache]
-    runs-on:
-      group: '${{ matrix.machine_type }}'
+        machine_type: [multi-gpu]
+    runs-on: ['${{ matrix.machine_type }}', nvidia-gpu, t4, push-ci]
     container:
       image: huggingface/transformers-all-latest-gpu-push-ci
       options: --gpus all --shm-size "16gb" --ipc host -v /mnt/cache/.cache/huggingface:/mnt/cache/
-    env:
-      # For the meaning of these environment variables, see the job `Setup`
-      CI_BRANCH_PUSH: ${{ github.event.ref }}
-      CI_BRANCH_WORKFLOW_RUN: ${{ github.event.workflow_run.head_branch }}
-      CI_SHA_PUSH: ${{ github.event.head_commit.id }}
-      CI_SHA_WORKFLOW_RUN: ${{ github.event.workflow_run.head_sha }}
     steps:
       # Necessary to get the correct branch name and commit SHA for `workflow_run` event
       # We also take into account the `push` event (we might want to test some changes in a branch)
@@ -263,7 +235,11 @@ jobs:
         shell: bash
         # For the meaning of these environment variables, see the job `Setup`
         run: |
+          CI_BRANCH_PUSH=${{ github.event.ref }}
           CI_BRANCH_PUSH=${CI_BRANCH_PUSH/'refs/heads/'/''}
+          CI_BRANCH_WORKFLOW_RUN=${{ github.event.workflow_run.head_branch }}
+          CI_SHA_PUSH=${{ github.event.head_commit.id }}
+          CI_SHA_WORKFLOW_RUN=${{ github.event.workflow_run.head_sha }}
           echo $CI_BRANCH_PUSH
           echo $CI_BRANCH_WORKFLOW_RUN
           echo $CI_SHA_PUSH
@@ -276,23 +252,6 @@ jobs:
           echo "env.CI_BRANCH = ${{ env.CI_BRANCH }}"
           echo "env.CI_SHA = ${{ env.CI_SHA }}"
 
-      - name: Set `machine_type` for report and artifact names
-        working-directory: /transformers
-        shell: bash
-        run: |
-          echo "${{ matrix.machine_type }}"
-
-          if [ "${{ matrix.machine_type }}" = "aws-g4dn-2xlarge-cache" ]; then
-            machine_type=single-gpu
-          elif [ "${{ matrix.machine_type }}" = "aws-g4dn-12xlarge-cache" ]; then
-            machine_type=multi-gpu
-          else
-            machine_type=${{ matrix.machine_type }}
-          fi
-
-          echo "$machine_type"
-          echo "machine_type=$machine_type" >> $GITHUB_ENV
-
       - name: Update clone using environment variables
         working-directory: /transformers
         run: |
@@ -336,19 +295,19 @@ jobs:
           MKL_SERVICE_FORCE_INTEL: 1
         working-directory: /transformers
         run: |
-          python3 -m pytest -n 2 --dist=loadfile -v --make-reports=${{ env.machine_type }}_tests_gpu_${{ matrix.folders }} ${{ fromJson(needs.setup.outputs.test_map)[matrix.folders] }}
+          python3 -m pytest -n 2 --dist=loadfile -v --make-reports=${{ matrix.machine_type }}_tests_gpu_${{ matrix.folders }} ${{ fromJson(needs.setup.outputs.test_map)[matrix.folders] }}
 
       - name: Failure short reports
         if: ${{ failure() }}
         continue-on-error: true
-        run: cat /transformers/reports/${{ env.machine_type }}_tests_gpu_${{ matrix.folders }}/failures_short.txt
+        run: cat /transformers/reports/${{ matrix.machine_type }}_tests_gpu_${{ matrix.folders }}/failures_short.txt
 
-      - name: "Test suite reports artifacts: ${{ env.machine_type }}_run_all_tests_gpu_${{ env.matrix_folders }}_test_reports"
+      - name: "Test suite reports artifacts: ${{ matrix.machine_type }}_run_all_tests_gpu_${{ env.matrix_folders }}_test_reports"
         if: ${{ always() }}
         uses: actions/upload-artifact@v4
         with:
-          name: ${{ env.machine_type }}_run_all_tests_gpu_${{ env.matrix_folders }}_test_reports
-          path: /transformers/reports/${{ env.machine_type }}_tests_gpu_${{ matrix.folders }}
+          name: ${{ matrix.machine_type }}_run_all_tests_gpu_${{ env.matrix_folders }}_test_reports
+          path: /transformers/reports/${{ matrix.machine_type }}_tests_gpu_${{ matrix.folders }}
 
   run_tests_torch_cuda_extensions_single_gpu:
     name: Torch CUDA extension tests
@@ -357,18 +316,11 @@ jobs:
     strategy:
       fail-fast: false
       matrix:
-        machine_type: [aws-g4dn-2xlarge-cache]
-    runs-on:
-      group: '${{ matrix.machine_type }}'
+        machine_type: [single-gpu]
+    runs-on: ['${{ matrix.machine_type }}', nvidia-gpu, t4, push-ci]
     container:
       image: huggingface/transformers-pytorch-deepspeed-latest-gpu-push-ci
       options: --gpus 0 --shm-size "16gb" --ipc host -v /mnt/cache/.cache/huggingface:/mnt/cache/
-    env:
-      # For the meaning of these environment variables, see the job `Setup`
-      CI_BRANCH_PUSH: ${{ github.event.ref }}
-      CI_BRANCH_WORKFLOW_RUN: ${{ github.event.workflow_run.head_branch }}
-      CI_SHA_PUSH: ${{ github.event.head_commit.id }}
-      CI_SHA_WORKFLOW_RUN: ${{ github.event.workflow_run.head_sha }}
     steps:
       # Necessary to get the correct branch name and commit SHA for `workflow_run` event
       # We also take into account the `push` event (we might want to test some changes in a branch)
@@ -376,7 +328,11 @@ jobs:
         shell: bash
         # For the meaning of these environment variables, see the job `Setup`
         run: |
+          CI_BRANCH_PUSH=${{ github.event.ref }}
           CI_BRANCH_PUSH=${CI_BRANCH_PUSH/'refs/heads/'/''}
+          CI_BRANCH_WORKFLOW_RUN=${{ github.event.workflow_run.head_branch }}
+          CI_SHA_PUSH=${{ github.event.head_commit.id }}
+          CI_SHA_WORKFLOW_RUN=${{ github.event.workflow_run.head_sha }}
           echo $CI_BRANCH_PUSH
           echo $CI_BRANCH_WORKFLOW_RUN
           echo $CI_SHA_PUSH
@@ -389,23 +345,6 @@ jobs:
           echo "env.CI_BRANCH = ${{ env.CI_BRANCH }}"
           echo "env.CI_SHA = ${{ env.CI_SHA }}"
 
-      - name: Set `machine_type` for report and artifact names
-        working-directory: /workspace/transformers
-        shell: bash
-        run: |
-          echo "${{ matrix.machine_type }}"
-
-          if [ "${{ matrix.machine_type }}" = "aws-g4dn-2xlarge-cache" ]; then
-            machine_type=single-gpu
-          elif [ "${{ matrix.machine_type }}" = "aws-g4dn-12xlarge-cache" ]; then
-            machine_type=multi-gpu
-          else
-            machine_type=${{ matrix.machine_type }}
-          fi
-
-          echo "$machine_type"
-          echo "machine_type=$machine_type" >> $GITHUB_ENV
-
       - name: Update clone using environment variables
         working-directory: /workspace/transformers
         run: |
@@ -446,19 +385,19 @@ jobs:
         working-directory: /workspace/transformers
         # TODO: Here we pass all tests in the 2 folders for simplicity. It's better to pass only the identified tests.
         run: |
-          python -m pytest -n 1 --dist=loadfile -v --make-reports=${{ env.machine_type }}_run_torch_cuda_extensions_gpu_test_reports tests/deepspeed tests/extended
+          python -m pytest -n 1 --dist=loadfile -v --make-reports=${{ matrix.machine_type }}_run_torch_cuda_extensions_gpu_test_reports tests/deepspeed tests/extended
 
       - name: Failure short reports
         if: ${{ failure() }}
         continue-on-error: true
-        run: cat /workspace/transformers/reports/${{ env.machine_type }}_run_torch_cuda_extensions_gpu_test_reports/failures_short.txt
+        run: cat /workspace/transformers/reports/${{ matrix.machine_type }}_run_torch_cuda_extensions_gpu_test_reports/failures_short.txt
 
-      - name: "Test suite reports artifacts: ${{ env.machine_type }}_run_torch_cuda_extensions_gpu_test_reports"
+      - name: "Test suite reports artifacts: ${{ matrix.machine_type }}_run_torch_cuda_extensions_gpu_test_reports"
         if: ${{ always() }}
         uses: actions/upload-artifact@v4
         with:
-          name: ${{ env.machine_type }}_run_torch_cuda_extensions_gpu_test_reports
-          path: /workspace/transformers/reports/${{ env.machine_type }}_run_torch_cuda_extensions_gpu_test_reports
+          name: ${{ matrix.machine_type }}_run_torch_cuda_extensions_gpu_test_reports
+          path: /workspace/transformers/reports/${{ matrix.machine_type }}_run_torch_cuda_extensions_gpu_test_reports
 
   run_tests_torch_cuda_extensions_multi_gpu:
     name: Torch CUDA extension tests
@@ -467,18 +406,11 @@ jobs:
     strategy:
       fail-fast: false
       matrix:
-        machine_type: [aws-g4dn-12xlarge-cache]
-    runs-on:
-      group: '${{ matrix.machine_type }}'
+        machine_type: [multi-gpu]
+    runs-on: ['${{ matrix.machine_type }}', nvidia-gpu, t4, push-ci]
     container:
       image: huggingface/transformers-pytorch-deepspeed-latest-gpu-push-ci
       options: --gpus all --shm-size "16gb" --ipc host -v /mnt/cache/.cache/huggingface:/mnt/cache/
-    env:
-      # For the meaning of these environment variables, see the job `Setup`
-      CI_BRANCH_PUSH: ${{ github.event.ref }}
-      CI_BRANCH_WORKFLOW_RUN: ${{ github.event.workflow_run.head_branch }}
-      CI_SHA_PUSH: ${{ github.event.head_commit.id }}
-      CI_SHA_WORKFLOW_RUN: ${{ github.event.workflow_run.head_sha }}
     steps:
       # Necessary to get the correct branch name and commit SHA for `workflow_run` event
       # We also take into account the `push` event (we might want to test some changes in a branch)
@@ -486,7 +418,11 @@ jobs:
         shell: bash
         # For the meaning of these environment variables, see the job `Setup`
         run: |
+          CI_BRANCH_PUSH=${{ github.event.ref }}
           CI_BRANCH_PUSH=${CI_BRANCH_PUSH/'refs/heads/'/''}
+          CI_BRANCH_WORKFLOW_RUN=${{ github.event.workflow_run.head_branch }}
+          CI_SHA_PUSH=${{ github.event.head_commit.id }}
+          CI_SHA_WORKFLOW_RUN=${{ github.event.workflow_run.head_sha }}
           echo $CI_BRANCH_PUSH
           echo $CI_BRANCH_WORKFLOW_RUN
           echo $CI_SHA_PUSH
@@ -499,23 +435,6 @@ jobs:
           echo "env.CI_BRANCH = ${{ env.CI_BRANCH }}"
           echo "env.CI_SHA = ${{ env.CI_SHA }}"
 
-      - name: Set `machine_type` for report and artifact names
-        working-directory: /workspace/transformers
-        shell: bash
-        run: |
-          echo "${{ matrix.machine_type }}"
-
-          if [ "${{ matrix.machine_type }}" = "aws-g4dn-2xlarge-cache" ]; then
-            machine_type=single-gpu
-          elif [ "${{ matrix.machine_type }}" = "aws-g4dn-12xlarge-cache" ]; then
-            machine_type=multi-gpu
-          else
-            machine_type=${{ matrix.machine_type }}
-          fi
-
-          echo "$machine_type"
-          echo "machine_type=$machine_type" >> $GITHUB_ENV
-
       - name: Update clone using environment variables
         working-directory: /workspace/transformers
         run: |
@@ -556,19 +475,19 @@ jobs:
         working-directory: /workspace/transformers
         # TODO: Here we pass all tests in the 2 folders for simplicity. It's better to pass only the identified tests.
         run: |
-          python -m pytest -n 1 --dist=loadfile -v --make-reports=${{ env.machine_type }}_run_torch_cuda_extensions_gpu_test_reports tests/deepspeed tests/extended
+          python -m pytest -n 1 --dist=loadfile -v --make-reports=${{ matrix.machine_type }}_run_torch_cuda_extensions_gpu_test_reports tests/deepspeed tests/extended
 
       - name: Failure short reports
         if: ${{ failure() }}
         continue-on-error: true
-        run: cat /workspace/transformers/reports/${{ env.machine_type }}_run_torch_cuda_extensions_gpu_test_reports/failures_short.txt
+        run: cat /workspace/transformers/reports/${{ matrix.machine_type }}_run_torch_cuda_extensions_gpu_test_reports/failures_short.txt
 
-      - name: "Test suite reports artifacts: ${{ env.machine_type }}_run_torch_cuda_extensions_gpu_test_reports"
+      - name: "Test suite reports artifacts: ${{ matrix.machine_type }}_run_torch_cuda_extensions_gpu_test_reports"
         if: ${{ always() }}
         uses: actions/upload-artifact@v4
         with:
-          name: ${{ env.machine_type }}_run_torch_cuda_extensions_gpu_test_reports
-          path: /workspace/transformers/reports/${{ env.machine_type }}_run_torch_cuda_extensions_gpu_test_reports
+          name: ${{ matrix.machine_type }}_run_torch_cuda_extensions_gpu_test_reports
+          path: /workspace/transformers/reports/${{ matrix.machine_type }}_run_torch_cuda_extensions_gpu_test_reports
 
   send_results:
     name: Send results to webhook
@@ -581,12 +500,6 @@ jobs:
         run_tests_torch_cuda_extensions_single_gpu,
         run_tests_torch_cuda_extensions_multi_gpu
     ]
-    env:
-      # For the meaning of these environment variables, see the job `Setup`
-      CI_BRANCH_PUSH: ${{ github.event.ref }}
-      CI_BRANCH_WORKFLOW_RUN: ${{ github.event.workflow_run.head_branch }}
-      CI_SHA_PUSH: ${{ github.event.head_commit.id }}
-      CI_SHA_WORKFLOW_RUN: ${{ github.event.workflow_run.head_sha }}
     steps:
       - name: Preliminary job status
         shell: bash
@@ -600,7 +513,11 @@ jobs:
         shell: bash
         # For the meaning of these environment variables, see the job `Setup`
         run: |
+          CI_BRANCH_PUSH=${{ github.event.ref }}
           CI_BRANCH_PUSH=${CI_BRANCH_PUSH/'refs/heads/'/''}
+          CI_BRANCH_WORKFLOW_RUN=${{ github.event.workflow_run.head_branch }}
+          CI_SHA_PUSH=${{ github.event.head_commit.id }}
+          CI_SHA_WORKFLOW_RUN=${{ github.event.workflow_run.head_sha }}
           echo $CI_BRANCH_PUSH
           echo $CI_BRANCH_WORKFLOW_RUN
           echo $CI_SHA_PUSH
@@ -646,7 +563,6 @@ jobs:
         # We pass `needs.setup.outputs.matrix` as the argument. A processing in `notification_service.py` to change
         # `models/bert` to `models_bert` is required, as the artifact names use `_` instead of `/`.
         run: |
-          pip install huggingface_hub
           pip install slack_sdk
           pip show slack_sdk
           python utils/notification_service.py "${{ needs.setup.outputs.matrix }}"

.github/workflows/self-scheduled-amd-mi210-caller.yml

@@ -0,0 +1,20 @@
+name: Self-hosted runner (AMD mi210 scheduled CI caller)
+
+on:
+  workflow_run:
+    workflows: ["Self-hosted runner (AMD scheduled CI caller)"]
+    branches: ["main"]
+    types: [completed]
+  push:
+    branches:
+      - run_amd_scheduled_ci_caller*
+
+jobs:
+  run_amd_ci:
+    name: AMD mi210
+    if: (cancelled() != true) && ((github.event_name == 'workflow_run') || ((github.event_name == 'push') && startsWith(github.ref_name, 'run_amd_scheduled_ci_caller')))
+    uses: ./.github/workflows/self-scheduled-amd.yml
+    with:
+      gpu_flavor: mi210
+      slack_report_channel: "#transformers-ci-daily-amd"
+    secrets: inherit

.github/workflows/self-scheduled-amd-mi250-caller.yml

@@ -1,59 +1,20 @@
-name: Self-hosted runner (AMD mi250 scheduled CI caller)
-
-on:
-  workflow_run:
-    workflows: ["Self-hosted runner (AMD scheduled CI caller)"]
-    branches: ["main"]
-    types: [completed]
-  push:
-    branches:
-      - run_amd_scheduled_ci_caller*
-
-jobs:
-  model-ci:
-    name: Model CI
-    uses: huggingface/hf-workflows/.github/workflows/transformers_amd_ci_scheduled.yaml@main
-    with:
-      job: run_models_gpu
-      slack_report_channel: "#transformers-ci-daily-amd"
-      runner: mi250
-      docker: huggingface/transformers-pytorch-amd-gpu
-      ci_event: Scheduled CI (AMD) - mi250
-      report_repo_id: optimum-amd/transformers_daily_ci
-    secrets: inherit
-
-  torch-pipeline:
-    name: Torch pipeline CI
-    uses: huggingface/hf-workflows/.github/workflows/transformers_amd_ci_scheduled.yaml@main
-    with:
-      job: run_pipelines_torch_gpu
-      slack_report_channel: "#transformers-ci-daily-amd"
-      runner: mi250
-      docker: huggingface/transformers-pytorch-amd-gpu
-      ci_event: Scheduled CI (AMD) - mi250
-      report_repo_id: optimum-amd/transformers_daily_ci
-    secrets: inherit
-
-  example-ci:
-    name: Example CI
-    uses: huggingface/hf-workflows/.github/workflows/transformers_amd_ci_scheduled.yaml@main
-    with:
-      job: run_examples_gpu
-      slack_report_channel: "#transformers-ci-daily-amd"
-      runner: mi250
-      docker: huggingface/transformers-pytorch-amd-gpu
-      ci_event: Scheduled CI (AMD) - mi250
-      report_repo_id: optimum-amd/transformers_daily_ci
-    secrets: inherit
-
-  deepspeed-ci:
-    name: DeepSpeed CI
-    uses: huggingface/hf-workflows/.github/workflows/transformers_amd_ci_scheduled.yaml@main
-    with:
-      job: run_torch_cuda_extensions_gpu
-      slack_report_channel: "#transformers-ci-daily-amd"
-      runner: mi250
-      docker: huggingface/transformers-pytorch-deepspeed-amd-gpu
-      ci_event: Scheduled CI (AMD) - mi250
-      report_repo_id: optimum-amd/transformers_daily_ci
-    secrets: inherit
+name: Self-hosted runner (AMD mi250 scheduled CI caller)
+
+on:
+  workflow_run:
+    workflows: ["Self-hosted runner (AMD scheduled CI caller)"]
+    branches: ["main"]
+    types: [completed]
+  push:
+    branches:
+      - run_amd_scheduled_ci_caller*
+
+jobs:
+  run_amd_ci:
+    name: AMD mi250
+    if: (cancelled() != true) && ((github.event_name == 'workflow_run') || ((github.event_name == 'push') && startsWith(github.ref_name, 'run_amd_scheduled_ci_caller')))
+    uses: ./.github/workflows/self-scheduled-amd.yml
+    with:
+      gpu_flavor: mi250
+      slack_report_channel: "#transformers-ci-daily-amd"
+    secrets: inherit

.github/workflows/self-scheduled-amd-mi300-caller.yml

@@ -1,8 +1,4 @@
-name: Self-hosted runner scale set (AMD mi300 scheduled CI caller)
-
-# Note: For every job in this workflow, the name of the runner scale set is finalized in the runner yaml i.e. huggingface/hf-workflows/.github/workflows/transformers_amd_ci_scheduled_arc_scale_set.yaml
-# For example, 1gpu scale set: amd-mi300-ci-1gpu
-#              2gpu scale set: amd-mi300-ci-2gpu
+name: Self-hosted runner (AMD mi300 scheduled CI caller)
 
 on:
   workflow_run:
@@ -14,50 +10,12 @@ on:
       - run_amd_scheduled_ci_caller*
 
 jobs:
-  model-ci:
-    name: Model CI
-    uses: huggingface/hf-workflows/.github/workflows/transformers_amd_ci_scheduled_arc_scale_set.yaml@main
+  run_amd_ci:
+    name: AMD mi300
+    needs: build-docker-containers
+    if: (cancelled() != true) && ((github.event_name == 'workflow_run') || ((github.event_name == 'push') && (startsWith(github.ref_name, 'run_amd_push_ci_caller') || startsWith(github.ref_name, 'mi300-ci'))))
+    uses: ./.github/workflows/self-scheduled-amd.yml
     with:
-      job: run_models_gpu
-      slack_report_channel: "#amd-hf-ci"
-      runner_scale_set: amd-mi300-ci
-      docker: huggingface/transformers-pytorch-amd-gpu
-      ci_event: Scheduled CI (AMD) - mi300
-      report_repo_id: optimum-amd/transformers_daily_ci
-    secrets: inherit
-
-  torch-pipeline:
-    name: Torch pipeline CI
-    uses: huggingface/hf-workflows/.github/workflows/transformers_amd_ci_scheduled_arc_scale_set.yaml@main
-    with:
-      job: run_pipelines_torch_gpu
-      slack_report_channel: "#amd-hf-ci"
-      runner_scale_set: amd-mi300-ci
-      docker: huggingface/transformers-pytorch-amd-gpu
-      ci_event: Scheduled CI (AMD) - mi300
-      report_repo_id: optimum-amd/transformers_daily_ci
-    secrets: inherit
-
-  example-ci:
-    name: Example CI
-    uses: huggingface/hf-workflows/.github/workflows/transformers_amd_ci_scheduled_arc_scale_set.yaml@main
-    with:
-      job: run_examples_gpu
-      slack_report_channel: "#amd-hf-ci"
-      runner_scale_set: amd-mi300-ci
-      docker: huggingface/transformers-pytorch-amd-gpu
-      ci_event: Scheduled CI (AMD) - mi300
-      report_repo_id: optimum-amd/transformers_daily_ci
-    secrets: inherit
-
-  deepspeed-ci:
-    name: DeepSpeed CI
-    uses: huggingface/hf-workflows/.github/workflows/transformers_amd_ci_scheduled_arc_scale_set.yaml@main
-    with:
-      job: run_torch_cuda_extensions_gpu
-      slack_report_channel: "#amd-hf-ci"
-      runner_scale_set: amd-mi300-ci
-      docker: huggingface/transformers-pytorch-deepspeed-amd-gpu
-      ci_event: Scheduled CI (AMD) - mi300
-      report_repo_id: optimum-amd/transformers_daily_ci
+      gpu_flavor: mi300
+      slack_report_channel: "#transformers-ci-daily-amd"
     secrets: inherit

.github/workflows/self-scheduled-amd.yml

@@ -0,0 +1,519 @@
+name: Self-hosted runner (scheduled-amd)
+
+# Note: For the AMD CI, we rely on a caller workflow and on the workflow_call event to trigger the
+# CI in order to run it on both MI210 and MI250, without having to use matrix here which pushes
+# us towards the limit of allowed jobs on GitHub Actions.
+on:
+  workflow_call:
+    inputs:
+      gpu_flavor:
+        required: true
+        type: string
+
+env:
+  HF_HOME: /mnt/cache
+  TRANSFORMERS_IS_CI: yes
+  OMP_NUM_THREADS: 8
+  MKL_NUM_THREADS: 8
+  RUN_SLOW: yes
+  HF_HUB_READ_TOKEN: ${{ secrets.HF_HUB_READ_TOKEN }}
+  SIGOPT_API_TOKEN: ${{ secrets.SIGOPT_API_TOKEN }}
+
+
+# Important note: each job (run_tests_single_gpu, run_tests_multi_gpu, run_examples_gpu, run_pipelines_torch_gpu) requires all the previous jobs before running.
+# This is done so that we avoid parallelizing the scheduled tests, to leave available
+# runners for the push CI that is running on the same machine.
+jobs:
+  check_runner_status:
+    name: Check Runner Status
+    runs-on: ubuntu-22.04
+    steps:
+      - name: Checkout transformers
+        uses: actions/checkout@v4
+        with:
+          fetch-depth: 2
+
+      - name: Check Runner Status
+        run: python utils/check_self_hosted_runner.py --target_runners hf-amd-mi210-ci-1gpu-1,hf-amd-mi250-ci-1gpu-1,hf-amd-mi300-ci-1gpu-1 --token ${{ secrets.ACCESS_REPO_INFO_TOKEN }}
+
+  check_runners:
+    name: Check Runners
+    needs: check_runner_status
+    strategy:
+      matrix:
+        machine_type: [single-gpu, multi-gpu]
+    runs-on: [self-hosted, amd-gpu, '${{ matrix.machine_type }}', '${{ inputs.gpu_flavor }}']
+    container:
+      image: huggingface/transformers-pytorch-amd-gpu
+      options: --device /dev/kfd --device /dev/dri --env ROCR_VISIBLE_DEVICES --shm-size "16gb" --ipc host -v /mnt/cache/.cache/huggingface:/mnt/cache/
+    steps:
+      - name: ROCM-SMI
+        run: |
+          rocm-smi
+      - name: ROCM-INFO
+        run: |
+          rocminfo  | grep "Agent" -A 14
+      - name: Show ROCR environment
+        run: |
+          echo "ROCR: $ROCR_VISIBLE_DEVICES"
+
+  setup:
+    name: Setup
+    needs: check_runners
+    strategy:
+      matrix:
+        machine_type: [single-gpu, multi-gpu]
+    runs-on: [self-hosted, amd-gpu, '${{ matrix.machine_type }}', '${{ inputs.gpu_flavor }}']
+    container:
+      image: huggingface/transformers-pytorch-amd-gpu
+      options: --device /dev/kfd --device /dev/dri --env ROCR_VISIBLE_DEVICES --shm-size "16gb" --ipc host -v /mnt/cache/.cache/huggingface:/mnt/cache/
+    outputs:
+      matrix: ${{ steps.set-matrix.outputs.matrix }}
+    steps:
+      - name: Update clone
+        working-directory: /transformers
+        run: |
+          git fetch && git checkout ${{ github.sha }}
+
+      - name: Cleanup
+        working-directory: /transformers
+        run: |
+          rm -rf tests/__pycache__
+          rm -rf tests/models/__pycache__
+          rm -rf reports
+
+      - name: Show installed libraries and their versions
+        working-directory: /transformers
+        run: pip freeze
+
+      - id: set-matrix
+        name: Identify models to test
+        working-directory: /transformers/tests
+        run: |
+          echo "matrix=$(python3 -c 'import os; tests = os.getcwd(); model_tests = os.listdir(os.path.join(tests, "models")); d1 = sorted(list(filter(os.path.isdir, os.listdir(tests)))); d2 = sorted(list(filter(os.path.isdir, [f"models/{x}" for x in model_tests]))); d1.remove("models"); d = d2 + d1; print(d)')" >> $GITHUB_OUTPUT
+
+      - name: ROCM-SMI
+        run: |
+          rocm-smi
+
+      - name: ROCM-INFO
+        run: |
+          rocminfo  | grep "Agent" -A 14
+      - name: Show ROCR environment
+        run: |
+          echo "ROCR: $ROCR_VISIBLE_DEVICES"
+
+      - name: Environment
+        working-directory: /transformers
+        run: |
+          python3 utils/print_env.py
+
+  run_models_gpu_single_gpu:
+    name: Single GPU tests
+    strategy:
+      max-parallel: 1  # For now, not to parallelize. Can change later if it works well.
+      fail-fast: false
+      matrix:
+        folders: ${{ fromJson(needs.setup.outputs.matrix) }}
+        machine_type: [single-gpu]
+    runs-on: [self-hosted, amd-gpu, '${{ matrix.machine_type }}', '${{ inputs.gpu_flavor }}']
+    container:
+      image: huggingface/transformers-pytorch-amd-gpu
+      options: --device /dev/kfd --device /dev/dri --env ROCR_VISIBLE_DEVICES --shm-size "16gb" --ipc host -v /mnt/cache/.cache/huggingface:/mnt/cache/
+    needs: setup
+    steps:
+      - name: Echo folder ${{ matrix.folders }}
+        shell: bash
+        # For folders like `models/bert`, set an env. var. (`matrix_folders`) to `models_bert`, which will be used to
+        # set the artifact folder names (because the character `/` is not allowed).
+        run: |
+          echo "${{ matrix.folders }}"
+          matrix_folders=${{ matrix.folders }}
+          matrix_folders=${matrix_folders/'models/'/'models_'}
+          echo "$matrix_folders"
+          echo "matrix_folders=$matrix_folders" >> $GITHUB_ENV
+
+      - name: Update clone
+        working-directory: /transformers
+        run: git fetch && git checkout ${{ github.sha }}
+
+      - name: Reinstall transformers in edit mode (remove the one installed during docker image build)
+        working-directory: /transformers
+        run: python3 -m pip uninstall -y transformers && python3 -m pip install -e .
+
+      - name: ROCM-SMI
+        run: |
+          rocm-smi
+      - name: ROCM-INFO
+        run: |
+          rocminfo  | grep "Agent" -A 14
+      - name: Show ROCR environment
+        run: |
+          echo "ROCR: $ROCR_VISIBLE_DEVICES"
+
+      - name: Environment
+        working-directory: /transformers
+        run: |
+          python3 utils/print_env.py
+
+      - name: Show installed libraries and their versions
+        working-directory: /transformers
+        run: pip freeze
+
+      - name: Run all tests on GPU
+        working-directory: /transformers
+        run: python3 -m pytest -v --make-reports=${{ matrix.machine_type }}_run_models_gpu_${{ matrix.folders }}_test_reports tests/${{ matrix.folders }} -m "not not_device_test"
+
+      - name: Failure short reports
+        if: ${{ failure() }}
+        continue-on-error: true
+        run: cat /transformers/reports/${{ matrix.machine_type }}_run_models_gpu_${{ matrix.folders }}_test_reports/failures_short.txt
+
+      - name: "Test suite reports artifacts: ${{ matrix.machine_type }}_run_models_gpu_${{ env.matrix_folders }}_test_reports"
+        if: ${{ always() }}
+        uses: actions/upload-artifact@v4
+        with:
+          name: ${{ matrix.machine_type }}_run_models_gpu_${{ env.matrix_folders }}_test_reports
+          path: /transformers/reports/${{ matrix.machine_type }}_run_models_gpu_${{ matrix.folders }}_test_reports
+
+  run_models_gpu_multi_gpu:
+    name: Multi GPU tests
+    strategy:
+      max-parallel: 1
+      fail-fast: false
+      matrix:
+        folders: ${{ fromJson(needs.setup.outputs.matrix) }}
+        machine_type: [multi-gpu]
+    runs-on: [self-hosted, amd-gpu, '${{ matrix.machine_type }}', '${{ inputs.gpu_flavor }}']
+    container:
+      image: huggingface/transformers-pytorch-amd-gpu
+      options: --device /dev/kfd --device /dev/dri --env ROCR_VISIBLE_DEVICES --shm-size "16gb" --ipc host -v /mnt/cache/.cache/huggingface:/mnt/cache/
+    needs: setup
+    steps:
+      - name: Echo folder ${{ matrix.folders }}
+        shell: bash
+        # For folders like `models/bert`, set an env. var. (`matrix_folders`) to `models_bert`, which will be used to
+        # set the artifact folder names (because the character `/` is not allowed).
+        run: |
+          echo "${{ matrix.folders }}"
+          matrix_folders=${{ matrix.folders }}
+          matrix_folders=${matrix_folders/'models/'/'models_'}
+          echo "$matrix_folders"
+          echo "matrix_folders=$matrix_folders" >> $GITHUB_ENV
+
+      - name: Update clone
+        working-directory: /transformers
+        run: git fetch && git checkout ${{ github.sha }}
+
+      - name: Reinstall transformers in edit mode (remove the one installed during docker image build)
+        working-directory: /transformers
+        run: python3 -m pip uninstall -y transformers && python3 -m pip install -e .
+
+      - name: ROCM-SMI
+        run: |
+          rocm-smi
+      - name: ROCM-INFO
+        run: |
+          rocminfo  | grep "Agent" -A 14
+      - name: Show ROCR environment
+        run: |
+          echo "ROCR: $ROCR_VISIBLE_DEVICES"
+
+      - name: Environment
+        working-directory: /transformers
+        run: |
+          python3 utils/print_env.py
+
+      - name: Show installed libraries and their versions
+        working-directory: /transformers
+        run: pip freeze
+
+      - name: Run all tests on GPU
+        working-directory: /transformers
+        run: python3 -m pytest -v --make-reports=${{ matrix.machine_type }}_run_models_gpu_${{ matrix.folders }}_test_reports tests/${{ matrix.folders }} -m "not not_device_test"
+
+      - name: Failure short reports
+        if: ${{ failure() }}
+        continue-on-error: true
+        run: cat /transformers/reports/${{ matrix.machine_type }}_run_models_gpu_${{ matrix.folders }}_test_reports/failures_short.txt
+
+      - name: "Test suite reports artifacts: ${{ matrix.machine_type }}_run_models_gpu_${{ env.matrix_folders }}_test_reports"
+        if: ${{ always() }}
+        uses: actions/upload-artifact@v4
+        with:
+          name: ${{ matrix.machine_type }}_run_models_gpu_${{ env.matrix_folders }}_test_reports
+          path: /transformers/reports/${{ matrix.machine_type }}_run_models_gpu_${{ matrix.folders }}_test_reports
+
+  run_examples_gpu:
+    name: Examples tests
+    strategy:
+      fail-fast: false
+      matrix:
+        machine_type: [single-gpu]
+    runs-on: [self-hosted, amd-gpu, '${{ matrix.machine_type }}', '${{ inputs.gpu_flavor }}']
+    container:
+      image: huggingface/transformers-pytorch-amd-gpu
+      options: --device /dev/kfd --device /dev/dri --env ROCR_VISIBLE_DEVICES --shm-size "16gb" --ipc host -v /mnt/cache/.cache/huggingface:/mnt/cache/
+    needs: setup
+    steps:
+      - name: Update clone
+        working-directory: /transformers
+        run: git fetch && git checkout ${{ github.sha }}
+
+      - name: Reinstall transformers in edit mode (remove the one installed during docker image build)
+        working-directory: /transformers
+        run: python3 -m pip uninstall -y transformers && python3 -m pip install -e .
+
+      - name: ROCM-SMI
+        run: |
+          rocm-smi
+      - name: ROCM-INFO
+        run: |
+          rocminfo  | grep "Agent" -A 14
+      - name: Show ROCR environment
+        run: |
+          echo "ROCR: $ROCR_VISIBLE_DEVICES"
+
+      - name: Environment
+        working-directory: /transformers
+        run: |
+          python3 utils/print_env.py
+
+      - name: Show installed libraries and their versions
+        working-directory: /transformers
+        run: pip freeze
+
+      - name: Run examples tests on GPU
+        working-directory: /transformers
+        run: |
+          pip install -r examples/pytorch/_tests_requirements.txt
+          python3 -m pytest -v --make-reports=${{ matrix.machine_type }}_run_examples_gpu_test_reports examples/pytorch -m "not not_device_test"
+
+      - name: Failure short reports
+        if: ${{ failure() }}
+        continue-on-error: true
+        run: cat /transformers/reports/${{ matrix.machine_type }}_run_examples_gpu_test_reports/failures_short.txt
+
+      - name: "Test suite reports artifacts: ${{ matrix.machine_type }}_run_examples_gpu_test_reports"
+        if: ${{ always() }}
+        uses: actions/upload-artifact@v4
+        with:
+          name: ${{ matrix.machine_type }}_run_examples_gpu_test_reports
+          path: /transformers/reports/${{ matrix.machine_type }}_run_examples_gpu_test_reports
+
+  run_pipelines_torch_gpu:
+    name: PyTorch pipelines tests
+    strategy:
+      fail-fast: false
+      matrix:
+        machine_type: [single-gpu, multi-gpu]
+    runs-on: [self-hosted, amd-gpu, '${{ matrix.machine_type }}', '${{ inputs.gpu_flavor }}']
+    container:
+      image: huggingface/transformers-pytorch-amd-gpu
+      options: --device /dev/kfd --device /dev/dri --env ROCR_VISIBLE_DEVICES --shm-size "16gb" --ipc host -v /mnt/cache/.cache/huggingface:/mnt/cache/
+    needs: setup
+    steps:
+      - name: Update clone
+        working-directory: /transformers
+        run: git fetch && git checkout ${{ github.sha }}
+
+      - name: Reinstall transformers in edit mode (remove the one installed during docker image build)
+        working-directory: /transformers
+        run: python3 -m pip uninstall -y transformers && python3 -m pip install -e .
+
+      - name: ROCM-SMI
+        run: |
+          rocm-smi
+      - name: ROCM-INFO
+        run: |
+          rocminfo  | grep "Agent" -A 14
+      - name: Show ROCR environment
+        run: |
+          echo "ROCR: $ROCR_VISIBLE_DEVICES"
+
+      - name: Environment
+        working-directory: /transformers
+        run: |
+          python3 utils/print_env.py
+
+      - name: Show installed libraries and their versions
+        working-directory: /transformers
+        run: pip freeze
+
+      - name: Run all pipeline tests on GPU
+        working-directory: /transformers
+        run: |
+          python3 -m pytest -n 1 -v --dist=loadfile --make-reports=${{ matrix.machine_type }}_run_pipelines_torch_gpu_test_reports tests/pipelines -m "not not_device_test"
+
+      - name: Failure short reports
+        if: ${{ failure() }}
+        continue-on-error: true
+        run: cat /transformers/reports/${{ matrix.machine_type }}_run_pipelines_torch_gpu_test_reports/failures_short.txt
+
+      - name: "Test suite reports artifacts: ${{ matrix.machine_type }}_run_pipelines_torch_gpu_test_reports"
+        if: ${{ always() }}
+        uses: actions/upload-artifact@v4
+        with:
+          name: ${{ matrix.machine_type }}_run_pipelines_torch_gpu_test_reports
+          path: /transformers/reports/${{ matrix.machine_type }}_run_pipelines_torch_gpu_test_reports
+
+  run_torch_cuda_extensions_gpu:
+    name: Torch ROCm deepspeed tests
+    strategy:
+      fail-fast: false
+      matrix:
+        machine_type: [single-gpu, multi-gpu]
+
+    runs-on: [self-hosted, amd-gpu, '${{ matrix.machine_type }}', '${{ inputs.gpu_flavor }}']
+    needs: setup
+    container:
+      image: huggingface/transformers-pytorch-deepspeed-amd-gpu
+      options: --device /dev/kfd --device /dev/dri --env ROCR_VISIBLE_DEVICES --shm-size "16gb" --ipc host -v /mnt/cache/.cache/huggingface:/mnt/cache/
+    steps:
+      - name: Update clone
+        working-directory: /transformers
+        run: git fetch && git checkout ${{ github.sha }}
+
+      - name: Reinstall transformers in edit mode (remove the one installed during docker image build)
+        working-directory: /transformers
+        run: python3 -m pip uninstall -y transformers && python3 -m pip install -e .
+
+      - name: ROCM-SMI
+        run: |
+          rocm-smi
+      - name: ROCM-INFO
+        run: |
+          rocminfo  | grep "Agent" -A 14
+
+      - name: Show ROCR environment
+        run: |
+          echo "ROCR: $ROCR_VISIBLE_DEVICES"
+
+      - name: Environment
+        working-directory: /transformers
+        run: |
+          python3 utils/print_env.py
+
+      - name: Show installed libraries and their versions
+        working-directory: /transformers
+        run: pip freeze
+
+      - name: Run all tests on GPU
+        working-directory: /transformers
+        run: python3 -m pytest -v --make-reports=${{ matrix.machine_type }}_run_torch_cuda_extensions_gpu_test_reports tests/deepspeed tests/extended -m "not not_device_test"
+
+      - name: Failure short reports
+        if: ${{ failure() }}
+        continue-on-error: true
+        run: cat /transformers/reports/${{ matrix.machine_type }}_run_torch_cuda_extensions_gpu_test_reports/failures_short.txt
+
+      - name: "Test suite reports artifacts: ${{ matrix.machine_type }}_run_torch_cuda_extensions_gpu_test_reports"
+        if: ${{ always() }}
+        uses: actions/upload-artifact@v4
+        with:
+          name: ${{ matrix.machine_type }}_run_torch_cuda_extensions_gpu_test_reports
+          path: /transformers/reports/${{ matrix.machine_type }}_run_torch_cuda_extensions_gpu_test_reports
+
+  run_extract_warnings:
+    name: Extract warnings in CI artifacts
+    runs-on: ubuntu-22.04
+    if: always()
+    needs: [
+      check_runner_status,
+      check_runners,
+      setup,
+      run_models_gpu_single_gpu,
+      run_models_gpu_multi_gpu,
+      run_examples_gpu,
+      run_pipelines_torch_gpu,
+      run_torch_cuda_extensions_gpu
+    ]
+    steps:
+      - name: Checkout transformers
+        uses: actions/checkout@v4
+        with:
+          fetch-depth: 2
+
+      - name: Install transformers
+        run: pip install transformers
+
+      - name: Show installed libraries and their versions
+        run: pip freeze
+
+      - name: Create output directory
+        run: mkdir warnings_in_ci
+
+      - uses: actions/download-artifact@v4
+        with:
+          path: warnings_in_ci
+
+      - name: Show artifacts
+        run: echo "$(python3 -c 'import os; d = os.listdir(); print(d)')"
+        working-directory: warnings_in_ci
+
+      - name: Extract warnings in CI artifacts
+        run: |
+          python3 utils/extract_warnings.py --workflow_run_id ${{ github.run_id }} --output_dir warnings_in_ci --token ${{ secrets.ACCESS_REPO_INFO_TOKEN }} --from_gh
+          echo "$(python3 -c 'import os; import json; fp = open("warnings_in_ci/selected_warnings.json"); d = json.load(fp); d = "\n".join(d) ;print(d)')"
+
+      - name: Upload artifact
+        if: ${{ always() }}
+        uses: actions/upload-artifact@v4
+        with:
+          name: warnings_in_ci
+          path: warnings_in_ci/selected_warnings.json
+
+  send_results:
+    name: Send results to webhook
+    runs-on: ubuntu-22.04
+    if: always()
+    needs: [
+      check_runner_status,
+      check_runners,
+      setup,
+      run_models_gpu_single_gpu,
+      run_models_gpu_multi_gpu,
+      run_examples_gpu,
+      run_pipelines_torch_gpu,
+      run_torch_cuda_extensions_gpu,
+      run_extract_warnings
+    ]
+    steps:
+      - name: Preliminary job status
+        shell: bash
+        # For the meaning of these environment variables, see the job `Setup`
+        run: |
+          echo "Runner availability: ${{ needs.check_runner_status.result }}"
+          echo "Runner status: ${{ needs.check_runners.result }}"
+          echo "Setup status: ${{ needs.setup.result }}"
+
+      - uses: actions/checkout@v4
+      - uses: actions/download-artifact@v4
+      - name: Send message to Slack
+        env:
+          CI_SLACK_BOT_TOKEN: ${{ secrets.CI_SLACK_BOT_TOKEN }}
+          CI_SLACK_CHANNEL_ID_DAILY_AMD: ${{ secrets.CI_SLACK_CHANNEL_ID_DAILY_AMD }}
+          CI_SLACK_CHANNEL_DUMMY_TESTS: ${{ secrets.CI_SLACK_CHANNEL_DUMMY_TESTS }}
+          CI_SLACK_REPORT_CHANNEL_ID: ${{ secrets.CI_SLACK_CHANNEL_ID_DAILY_AMD }}
+          ACCESS_REPO_INFO_TOKEN: ${{ secrets.ACCESS_REPO_INFO_TOKEN }}
+          CI_EVENT: Scheduled CI (AMD) - ${{ inputs.gpu_flavor }}
+          CI_SHA: ${{ github.sha }}
+          CI_WORKFLOW_REF: ${{ github.workflow_ref }}
+          RUNNER_STATUS: ${{ needs.check_runner_status.result }}
+          RUNNER_ENV_STATUS: ${{ needs.check_runners.result }}
+          SETUP_STATUS: ${{ needs.setup.result }}
+        # We pass `needs.setup.outputs.matrix` as the argument. A processing in `notification_service.py` to change
+        # `models/bert` to `models_bert` is required, as the artifact names use `_` instead of `/`.
+        run: |
+          sudo apt-get install -y curl
+          pip install slack_sdk
+          pip show slack_sdk
+          python utils/notification_service.py "${{ needs.setup.outputs.matrix }}"
+
+      # Upload complete failure tables, as they might be big and only truncated versions could be sent to Slack.
+      - name: Failure table artifacts
+        if: ${{ always() }}
+        uses: actions/upload-artifact@v4
+        with:
+          name: test_failure_tables
+          path: test_failure_tables

.github/workflows/self-scheduled-caller.yml

@@ -8,52 +8,17 @@ on:
   push:
     branches:
       - run_scheduled_ci*
-  workflow_dispatch:
-    inputs:
-      prev_workflow_run_id:
-        description: 'previous workflow run id to compare'
-        type: string
-        required: false
-        default: ""
-      other_workflow_run_id:
-        description: 'other workflow run id to compare'
-        type: string
-        required: false
-        default: ""
-
-
-# Used for `push` to easily modify the target workflow runs to compare against
-env:
-    prev_workflow_run_id: ""
-    other_workflow_run_id: ""
-
 
 jobs:
-  setup:
-    name: Setup
-    runs-on: ubuntu-22.04
-    steps:
-      - name: Setup
-        run: |
-          mkdir "setup_values"
-          echo "${{ inputs.prev_workflow_run_id || env.prev_workflow_run_id }}" > "setup_values/prev_workflow_run_id.txt"
-          echo "${{ inputs.other_workflow_run_id || env.other_workflow_run_id }}" > "setup_values/other_workflow_run_id.txt"
-
-      - name: Upload artifacts
-        uses: actions/upload-artifact@v4
-        with:
-          name: setup_values
-          path: setup_values
-
   model-ci:
     name: Model CI
     uses: ./.github/workflows/self-scheduled.yml
     with:
       job: run_models_gpu
       slack_report_channel: "#transformers-ci-daily-models"
+      runner: daily-ci
       docker: huggingface/transformers-all-latest-gpu
       ci_event: Daily CI
-      report_repo_id: hf-internal-testing/transformers_daily_ci
     secrets: inherit
 
   torch-pipeline:
@@ -62,9 +27,20 @@ jobs:
     with:
       job: run_pipelines_torch_gpu
       slack_report_channel: "#transformers-ci-daily-pipeline-torch"
+      runner: daily-ci
       docker: huggingface/transformers-pytorch-gpu
       ci_event: Daily CI
-      report_repo_id: hf-internal-testing/transformers_daily_ci
+    secrets: inherit
+
+  tf-pipeline:
+    name: TF pipeline CI
+    uses: ./.github/workflows/self-scheduled.yml
+    with:
+      job: run_pipelines_tf_gpu
+      slack_report_channel: "#transformers-ci-daily-pipeline-tf"
+      runner: daily-ci
+      docker: huggingface/transformers-tensorflow-gpu
+      ci_event: Daily CI
     secrets: inherit
 
   example-ci:
@@ -73,20 +49,9 @@ jobs:
     with:
       job: run_examples_gpu
       slack_report_channel: "#transformers-ci-daily-examples"
+      runner: daily-ci
       docker: huggingface/transformers-all-latest-gpu
       ci_event: Daily CI
-      report_repo_id: hf-internal-testing/transformers_daily_ci
-    secrets: inherit
-
-  trainer-fsdp-ci:
-    name: Trainer/FSDP CI
-    uses: ./.github/workflows/self-scheduled.yml
-    with:
-      job: run_trainer_and_fsdp_gpu
-      slack_report_channel: "#transformers-ci-daily-training"
-      docker: huggingface/transformers-all-latest-gpu
-      ci_event: Daily CI
-      report_repo_id: hf-internal-testing/transformers_daily_ci
     secrets: inherit
 
   deepspeed-ci:
@@ -94,11 +59,11 @@ jobs:
     uses: ./.github/workflows/self-scheduled.yml
     with:
       job: run_torch_cuda_extensions_gpu
-      slack_report_channel: "#transformers-ci-daily-training"
+      slack_report_channel: "#transformers-ci-daily-deepspeed"
+      runner: daily-ci
       docker: huggingface/transformers-pytorch-deepspeed-latest-gpu
       ci_event: Daily CI
       working-directory-prefix: /workspace
-      report_repo_id: hf-internal-testing/transformers_daily_ci
     secrets: inherit
 
   quantization-ci:
@@ -107,7 +72,7 @@ jobs:
     with:
       job: run_quantization_torch_gpu
       slack_report_channel: "#transformers-ci-daily-quantization"
+      runner: daily-ci
       docker: huggingface/transformers-quantization-latest-gpu
       ci_event: Daily CI
-      report_repo_id: hf-internal-testing/transformers_daily_ci
     secrets: inherit

.github/workflows/self-scheduled-intel-gaudi.yml

@@ -1,345 +0,0 @@
-name: Self-hosted runner (scheduled-intel-gaudi)
-
-on:
-  workflow_call:
-    inputs:
-      job:
-        required: true
-        type: string
-      slack_report_channel:
-        required: true
-        type: string
-      runner_scale_set:
-        required: true
-        type: string
-      ci_event:
-        required: true
-        type: string
-      report_repo_id:
-        required: true
-        type: string
-
-env:
-  NUM_SLICES: 2
-  RUN_SLOW: yes
-  PT_HPU_LAZY_MODE: 0
-  TRANSFORMERS_IS_CI: yes
-  PT_ENABLE_INT64_SUPPORT: 1
-  HF_HUB_READ_TOKEN: ${{ secrets.HF_HUB_READ_TOKEN }}
-  SIGOPT_API_TOKEN: ${{ secrets.SIGOPT_API_TOKEN }}
-  HF_HOME: /mnt/cache/.cache/huggingface
-
-jobs:
-  setup:
-    if: contains(fromJSON('["run_models_gpu", "run_trainer_and_fsdp_gpu"]'), inputs.job)
-    name: Setup
-    runs-on: ubuntu-latest
-    outputs:
-      slice_ids: ${{ steps.set-matrix.outputs.slice_ids }}
-      folder_slices: ${{ steps.set-matrix.outputs.folder_slices }}
-      quantization_matrix: ${{ steps.set-matrix.outputs.quantization_matrix }}
-    steps:
-      - name: Checkout
-        uses: actions/checkout@v4
-        with:
-          fetch-depth: 0
-
-      - name: Set up Python
-        uses: actions/setup-python@v5
-        with:
-          python-version: "3.10"
-
-      - id: set-matrix
-        if: contains(fromJSON('["run_models_gpu", "run_trainer_and_fsdp_gpu"]'), inputs.job)
-        name: Identify models to test
-        working-directory: tests
-        run: |
-          if [ "${{ inputs.job }}" = "run_models_gpu" ]; then
-            echo "folder_slices=$(python3 ../utils/split_model_tests.py --num_splits ${{ env.NUM_SLICES }})" >> $GITHUB_OUTPUT
-            echo "slice_ids=$(python3 -c 'd = list(range(${{ env.NUM_SLICES }})); print(d)')" >> $GITHUB_OUTPUT
-          elif [ "${{ inputs.job }}" = "run_trainer_and_fsdp_gpu" ]; then
-            echo "folder_slices=[['trainer'], ['fsdp']]" >> $GITHUB_OUTPUT
-            echo "slice_ids=[0, 1]" >> $GITHUB_OUTPUT
-          fi
-
-      - id: set-matrix-quantization
-        if: ${{ inputs.job == 'run_quantization_torch_gpu' }}
-        name: Identify quantization method to test
-        working-directory: tests
-        run: |
-          echo "quantization_matrix=$(python3 -c 'import os; tests = os.getcwd(); quantization_tests = os.listdir(os.path.join(tests, "quantization")); d = sorted(list(filter(os.path.isdir, [f"quantization/{x}" for x in quantization_tests]))) ;  print(d)')" >> $GITHUB_OUTPUT
-
-  run_models_gpu:
-    if: ${{ inputs.job == 'run_models_gpu' }}
-    name: " "
-    needs: setup
-    strategy:
-      fail-fast: false
-      matrix:
-        machine_type: [1gaudi, 2gaudi]
-        slice_id: ${{ fromJSON(needs.setup.outputs.slice_ids) }}
-    uses: ./.github/workflows/model_jobs_intel_gaudi.yml
-    with:
-      slice_id: ${{ matrix.slice_id }}
-      machine_type: ${{ matrix.machine_type }}
-      folder_slices: ${{ needs.setup.outputs.folder_slices }}
-      runner: ${{ inputs.runner_scale_set }}-${{ matrix.machine_type }}
-      report_name_prefix: run_models_gpu
-
-    secrets: inherit
-
-  run_trainer_and_fsdp_gpu:
-    if: ${{ inputs.job == 'run_trainer_and_fsdp_gpu' }}
-    name: " "
-    needs: setup
-    strategy:
-      fail-fast: false
-      matrix:
-        machine_type: [1gaudi, 2gaudi]
-        slice_id: ${{ fromJSON(needs.setup.outputs.slice_ids) }}
-    uses: ./.github/workflows/model_jobs_intel_gaudi.yml
-    with:
-      slice_id: ${{ matrix.slice_id }}
-      machine_type: ${{ matrix.machine_type }}
-      folder_slices: ${{ needs.setup.outputs.folder_slices }}
-      runner: ${{ inputs.runner_scale_set }}-${{ matrix.machine_type }}
-      report_name_prefix: run_trainer_and_fsdp_gpu
-
-    secrets: inherit
-
-  run_pipelines_gpu:
-    if: ${{ inputs.job == 'run_pipelines_gpu' }}
-    name: Pipelines
-    strategy:
-      fail-fast: false
-      matrix:
-        machine_type: [1gaudi, 2gaudi]
-    runs-on:
-      group: ${{ inputs.runner_scale_set }}-${{ matrix.machine_type }}
-    container:
-      image: vault.habana.ai/gaudi-docker/1.21.1/ubuntu22.04/habanalabs/pytorch-installer-2.6.0:latest
-      options: --runtime=habana
-        -v /mnt/cache/.cache/huggingface:/mnt/cache/.cache/huggingface
-        --env OMPI_MCA_btl_vader_single_copy_mechanism=none
-        --env HABANA_VISIBLE_DEVICES
-        --env HABANA_VISIBLE_MODULES
-        --cap-add=sys_nice
-        --shm-size=64G
-    steps:
-      - name: Checkout
-        uses: actions/checkout@v4
-        with:
-          fetch-depth: 0
-
-      - name: Install dependencies
-        run: |
-          pip install -e .[testing,torch] "numpy<2.0.0" scipy scikit-learn librosa soundfile
-
-      - name: HL-SMI
-        run: |
-          hl-smi
-          echo "HABANA_VISIBLE_DEVICES=${HABANA_VISIBLE_DEVICES}"
-          echo "HABANA_VISIBLE_MODULES=${HABANA_VISIBLE_MODULES}"
-
-      - name: Environment
-        run: python3 utils/print_env.py
-
-      - name: Show installed libraries and their versions
-        run: pip freeze
-
-      - name: Set `machine_type` for report and artifact names
-        shell: bash
-        run: |
-          if [ "${{ matrix.machine_type }}" = "1gaudi" ]; then
-            machine_type=single-gpu
-          elif [ "${{ matrix.machine_type }}" = "2gaudi" ]; then
-            machine_type=multi-gpu
-          else
-            machine_type=${{ matrix.machine_type }}
-          fi
-          echo "machine_type=$machine_type" >> $GITHUB_ENV
-
-      - name: Run all pipeline tests on Intel Gaudi
-        run: |
-          python3 -m pytest -v --make-reports=${{ env.machine_type }}_run_pipelines_gpu_test_reports tests/pipelines -m "not not_device_test"
-
-      - name: Failure short reports
-        if: ${{ failure() }}
-        continue-on-error: true
-        run: |
-          cat reports/${{ env.machine_type }}_run_pipelines_gpu_test_reports/failures_short.txt
-
-      - name: "Test suite reports artifacts: ${{ env.machine_type }}_run_pipelines_gpu_test_reports"
-        if: ${{ always() }}
-        uses: actions/upload-artifact@v4
-        with:
-          name: ${{ env.machine_type }}_run_pipelines_gpu_test_reports
-          path: reports/${{ env.machine_type }}_run_pipelines_gpu_test_reports
-
-  run_examples_gpu:
-    if: ${{ inputs.job == 'run_examples_gpu' }}
-    name: Examples directory
-    strategy:
-      fail-fast: false
-      matrix:
-        machine_type: [1gaudi]
-    runs-on:
-      group: ${{ inputs.runner_scale_set }}-${{ matrix.machine_type }}
-    container:
-      image: vault.habana.ai/gaudi-docker/1.21.1/ubuntu22.04/habanalabs/pytorch-installer-2.6.0:latest
-      options: --runtime=habana
-        -v /mnt/cache/.cache/huggingface:/mnt/cache/.cache/huggingface
-        --env OMPI_MCA_btl_vader_single_copy_mechanism=none
-        --env HABANA_VISIBLE_DEVICES
-        --env HABANA_VISIBLE_MODULES
-        --cap-add=sys_nice
-        --shm-size=64G
-    steps:
-      - name: Checkout
-        uses: actions/checkout@v4
-        with:
-          fetch-depth: 0
-
-      - name: Install dependencies
-        run: |
-          pip install -e .[testing,torch] "numpy<2.0.0" scipy scikit-learn librosa soundfile
-
-      - name: HL-SMI
-        run: |
-          hl-smi
-          echo "HABANA_VISIBLE_DEVICES=${HABANA_VISIBLE_DEVICES}"
-          echo "HABANA_VISIBLE_MODULES=${HABANA_VISIBLE_MODULES}"
-
-      - name: Environment
-        run: |
-          python3 utils/print_env.py
-
-      - name: Show installed libraries and their versions
-        run: |
-          pip freeze
-
-      - name: Set `machine_type` for report and artifact names
-        shell: bash
-        run: |
-          if [ "${{ matrix.machine_type }}" = "1gaudi" ]; then
-            machine_type=single-gpu
-          elif [ "${{ matrix.machine_type }}" = "2gaudi" ]; then
-            machine_type=multi-gpu
-          else
-            machine_type=${{ matrix.machine_type }}
-          fi
-          echo "machine_type=$machine_type" >> $GITHUB_ENV
-
-      - name: Run examples tests on Intel Gaudi
-        run: |
-          pip install -r examples/pytorch/_tests_requirements.txt
-          python3 -m pytest -v --make-reports=${{ env.machine_type }}_run_examples_gpu_test_reports examples/pytorch -m "not not_device_test"
-
-      - name: Failure short reports
-        if: ${{ failure() }}
-        continue-on-error: true
-        run: |
-          cat reports/${{ env.machine_type }}_run_examples_gpu_test_reports/failures_short.txt
-
-      - name: "Test suite reports artifacts: ${{ env.machine_type }}_run_examples_gpu_test_reports"
-        if: ${{ always() }}
-        uses: actions/upload-artifact@v4
-        with:
-          name: ${{ env.machine_type }}_run_examples_gpu_test_reports
-          path: reports/${{ env.machine_type }}_run_examples_gpu_test_reports
-
-  run_deepspeed_gpu:
-    if: ${{ inputs.job == 'run_deepspeed_gpu' }}
-    name: Intel Gaudi deepspeed tests
-    strategy:
-      fail-fast: false
-      matrix:
-        machine_type: [1gaudi, 2gaudi]
-    runs-on:
-      group: ${{ inputs.runner_scale_set }}-${{ matrix.machine_type }}
-    container:
-      image: vault.habana.ai/gaudi-docker/1.21.1/ubuntu22.04/habanalabs/pytorch-installer-2.6.0:latest
-      options: --runtime=habana
-        -v /mnt/cache/.cache/huggingface:/mnt/cache/.cache/huggingface
-        --env OMPI_MCA_btl_vader_single_copy_mechanism=none
-        --env HABANA_VISIBLE_DEVICES
-        --env HABANA_VISIBLE_MODULES
-        --cap-add=sys_nice
-        --shm-size=64G
-    steps:
-      - name: Checkout
-        uses: actions/checkout@v4
-        with:
-          fetch-depth: 0
-
-      - name: Install dependencies
-        run: |
-          pip install -e .[testing,torch] "numpy<2.0.0" scipy scikit-learn librosa soundfile
-          pip install git+https://github.com/HabanaAI/DeepSpeed.git@1.20.0
-
-      - name: HL-SMI
-        run: |
-          hl-smi
-          echo "HABANA_VISIBLE_DEVICES=${HABANA_VISIBLE_DEVICES}"
-          echo "HABANA_VISIBLE_MODULES=${HABANA_VISIBLE_MODULES}"
-
-      - name: Environment
-        run: |
-          python3 utils/print_env.py
-
-      - name: Show installed libraries and their versions
-        run: |
-          pip freeze
-
-      - name: Set `machine_type` for report and artifact names
-        shell: bash
-        run: |
-          if [ "${{ matrix.machine_type }}" = "1gaudi" ]; then
-            machine_type=single-gpu
-          elif [ "${{ matrix.machine_type }}" = "2gaudi" ]; then
-            machine_type=multi-gpu
-          else
-            machine_type=${{ matrix.machine_type }}
-          fi
-          echo "machine_type=$machine_type" >> $GITHUB_ENV
-
-      - name: Run all deepspeed tests on intel Gaudi
-        run: |
-          python3 -m pytest -v --make-reports=${{ env.machine_type }}_run_deepspeed_gpu_test_reports tests/deepspeed -m "not not_device_test"
-
-      - name: Failure short reports
-        if: ${{ failure() }}
-        continue-on-error: true
-        run: |
-          cat reports/${{ env.machine_type }}_run_deepspeed_gpu_test_reports/failures_short.txt
-
-      - name: "Test suite reports artifacts: ${{ env.machine_type }}_run_deepspeed_gpu_test_reports"
-        if: ${{ always() }}
-        uses: actions/upload-artifact@v4
-        with:
-          name: ${{ env.machine_type }}_run_deepspeed_gpu_test_reports
-          path: reports/${{ env.machine_type }}_run_deepspeed_gpu_test_reports
-
-  send_results:
-    name: Slack Report
-    needs:
-      [
-        setup,
-        run_models_gpu,
-        run_examples_gpu,
-        run_pipelines_gpu,
-        run_deepspeed_gpu,
-        run_trainer_and_fsdp_gpu,
-      ]
-    if: ${{ always() }}
-    uses: ./.github/workflows/slack-report.yml
-    with:
-      job: ${{ inputs.job }}
-      setup_status: ${{ needs.setup.result }}
-      slack_report_channel: ${{ inputs.slack_report_channel }}
-      quantization_matrix: ${{ needs.setup.outputs.quantization_matrix }}
-      folder_slices: ${{ needs.setup.outputs.folder_slices }}
-      report_repo_id: ${{ inputs.report_repo_id }}
-      ci_event: ${{ inputs.ci_event }}
-
-    secrets: inherit

.github/workflows/self-scheduled-intel-gaudi3-caller.yml

@@ -1,67 +0,0 @@
-name: Self-hosted runner (Intel Gaudi3 scheduled CI caller)
-
-on:
-  repository_dispatch:
-  workflow_dispatch:
-  schedule:
-    - cron: "17 2 * * *"
-
-jobs:
-  model-ci:
-    name: Model CI
-    uses: ./.github/workflows/self-scheduled-intel-gaudi.yml
-    with:
-      job: run_models_gpu
-      ci_event: Scheduled CI (Intel) - Gaudi3
-      runner_scale_set: itac-bm-emr-gaudi3-dell
-      slack_report_channel: "#transformers-ci-daily-intel-gaudi3"
-      report_repo_id: optimum-intel/transformers_daily_ci_intel_gaudi3
-
-    secrets: inherit
-
-  pipeline-ci:
-    name: Pipeline CI
-    uses: ./.github/workflows/self-scheduled-intel-gaudi.yml
-    with:
-      job: run_pipelines_gpu
-      ci_event: Scheduled CI (Intel) - Gaudi3
-      runner_scale_set: itac-bm-emr-gaudi3-dell
-      slack_report_channel: "#transformers-ci-daily-intel-gaudi3"
-      report_repo_id: optimum-intel/transformers_daily_ci_intel_gaudi3
-
-    secrets: inherit
-
-  example-ci:
-    name: Example CI
-    uses: ./.github/workflows/self-scheduled-intel-gaudi.yml
-    with:
-      job: run_examples_gpu
-      ci_event: Scheduled CI (Intel) - Gaudi3
-      runner_scale_set: itac-bm-emr-gaudi3-dell
-      slack_report_channel: "#transformers-ci-daily-intel-gaudi3"
-      report_repo_id: optimum-intel/transformers_daily_ci_intel_gaudi3
-
-    secrets: inherit
-
-  deepspeed-ci:
-    name: DeepSpeed CI
-    uses: ./.github/workflows/self-scheduled-intel-gaudi.yml
-    with:
-      job: run_deepspeed_gpu
-      ci_event: Scheduled CI (Intel) - Gaudi3
-      runner_scale_set: itac-bm-emr-gaudi3-dell
-      slack_report_channel: "#transformers-ci-daily-intel-gaudi3"
-      report_repo_id: optimum-intel/transformers_daily_ci_intel_gaudi3
-
-    secrets: inherit
-
-  trainer-fsdp-ci:
-    name: Trainer/FSDP CI
-    uses: ./.github/workflows/self-scheduled-intel-gaudi.yml
-    with:
-      job: run_trainer_and_fsdp_gpu
-      ci_event: Scheduled CI (Intel) - Gaudi3
-      runner_scale_set: itac-bm-emr-gaudi3-dell
-      slack_report_channel: "#transformers-ci-daily-intel-gaudi3"
-      report_repo_id: optimum-intel/transformers_daily_ci_intel_gaudi3
-    secrets: inherit

.github/workflows/self-scheduled.yml

@@ -15,6 +15,9 @@ on:
       slack_report_channel:
         required: true
         type: string
+      runner:
+        required: true
+        type: string
       docker:
         required: true
         type: string
@@ -25,10 +28,6 @@ on:
         default: ''
         required: false
         type: string
-      report_repo_id:
-        required: true
-        type: string
-
 
 env:
   HF_HOME: /mnt/cache
@@ -41,25 +40,24 @@ env:
   HF_HUB_READ_TOKEN: ${{ secrets.HF_HUB_READ_TOKEN }}
   SIGOPT_API_TOKEN: ${{ secrets.SIGOPT_API_TOKEN }}
   TF_FORCE_GPU_ALLOW_GROWTH: true
+  RUN_PT_TF_CROSS_TESTS: 1
   CUDA_VISIBLE_DEVICES: 0,1
   NUM_SLICES: 2
 
 jobs:
   setup:
-    if: contains(fromJSON('["run_models_gpu", "run_trainer_and_fsdp_gpu", "run_quantization_torch_gpu"]'), inputs.job)
+    if: contains(fromJSON('["run_models_gpu", "run_quantization_torch_gpu"]'), inputs.job)
     name: Setup
     strategy:
       matrix:
-        machine_type: [aws-g4dn-4xlarge-cache, aws-g4dn-12xlarge-cache]
-    runs-on:
-      group: '${{ matrix.machine_type }}'
+        machine_type: [single-gpu, multi-gpu]
+    runs-on: ['${{ matrix.machine_type }}', nvidia-gpu, t4, '${{ inputs.runner }}']
     container:
       image: huggingface/transformers-all-latest-gpu
       options: --gpus 0 --shm-size "16gb" --ipc host -v /mnt/cache/.cache/huggingface:/mnt/cache/
     outputs:
       folder_slices: ${{ steps.set-matrix.outputs.folder_slices }}
       slice_ids: ${{ steps.set-matrix.outputs.slice_ids }}
-      runner_map: ${{ steps.set-matrix.outputs.runner_map }}
       quantization_matrix: ${{ steps.set-matrix-quantization.outputs.quantization_matrix }}
     steps:
       - name: Update clone
@@ -79,19 +77,13 @@ jobs:
         run: pip freeze
 
       - id: set-matrix
-        if: contains(fromJSON('["run_models_gpu", "run_trainer_and_fsdp_gpu"]'), inputs.job)
+        if: ${{ inputs.job == 'run_models_gpu' }}
         name: Identify models to test
         working-directory: /transformers/tests
         run: |
-          if [ "${{ inputs.job }}" = "run_models_gpu" ]; then
-            echo "folder_slices=$(python3 ../utils/split_model_tests.py --num_splits ${{ env.NUM_SLICES }})" >> $GITHUB_OUTPUT
-            echo "slice_ids=$(python3 -c 'd = list(range(${{ env.NUM_SLICES }})); print(d)')" >> $GITHUB_OUTPUT
-            echo "runner_map=$(python3 ../utils/get_runner_map.py)" >> $GITHUB_OUTPUT
-          elif [ "${{ inputs.job }}" = "run_trainer_and_fsdp_gpu" ]; then
-            echo "folder_slices=[['trainer'], ['fsdp']]" >> $GITHUB_OUTPUT
-            echo "slice_ids=[0, 1]" >> $GITHUB_OUTPUT
-          fi
-
+          echo "folder_slices=$(python3 ../utils/split_model_tests.py --num_splits ${{ env.NUM_SLICES }})" >> $GITHUB_OUTPUT
+          echo "slice_ids=$(python3 -c 'd = list(range(${{ env.NUM_SLICES }})); print(d)')" >> $GITHUB_OUTPUT
+      
       - id: set-matrix-quantization
         if: ${{ inputs.job == 'run_quantization_torch_gpu' }}
         name: Identify quantization method to test
@@ -117,37 +109,18 @@ jobs:
       folder_slices: ${{ needs.setup.outputs.folder_slices }}
       machine_type: ${{ matrix.machine_type }}
       slice_id: ${{ matrix.slice_id }}
-      runner_map: ${{ needs.setup.outputs.runner_map }}
+      runner: ${{ inputs.runner }}
       docker: ${{ inputs.docker }}
     secrets: inherit
 
-  run_trainer_and_fsdp_gpu:
-    if: ${{ inputs.job == 'run_trainer_and_fsdp_gpu' }}
-    name: " "
-    needs: setup
-    strategy:
-      fail-fast: false
-      matrix:
-        machine_type: [aws-g4dn-4xlarge-cache, aws-g4dn-12xlarge-cache]
-        slice_id: [0, 1]
-    uses: ./.github/workflows/model_jobs.yml
-    with:
-      folder_slices: ${{ needs.setup.outputs.folder_slices }}
-      machine_type: ${{ matrix.machine_type }}
-      slice_id: ${{ matrix.slice_id }}
-      docker: ${{ inputs.docker }}
-      report_name_prefix: run_trainer_and_fsdp_gpu
-    secrets: inherit
-
   run_pipelines_torch_gpu:
     if: ${{ inputs.job == 'run_pipelines_torch_gpu' }}
     name: PyTorch pipelines
     strategy:
       fail-fast: false
       matrix:
-        machine_type: [aws-g4dn-4xlarge-cache, aws-g4dn-12xlarge-cache]
-    runs-on:
-      group: '${{ matrix.machine_type }}'
+        machine_type: [single-gpu, multi-gpu]
+    runs-on: ['${{ matrix.machine_type }}', nvidia-gpu, t4, '${{ inputs.runner }}']
     container:
       image: huggingface/transformers-pytorch-gpu
       options: --gpus all --shm-size "16gb" --ipc host -v /mnt/cache/.cache/huggingface:/mnt/cache/
@@ -173,39 +146,73 @@ jobs:
         working-directory: /transformers
         run: pip freeze
 
-      - name: Set `machine_type` for report and artifact names
-        working-directory: /transformers
-        shell: bash
-        run: |
-          echo "${{ matrix.machine_type }}"
-
-          if [ "${{ matrix.machine_type }}" = "aws-g4dn-4xlarge-cache" ]; then
-            machine_type=single-gpu
-          elif [ "${{ matrix.machine_type }}" = "aws-g4dn-12xlarge-cache" ]; then
-            machine_type=multi-gpu
-          else
-            machine_type=${{ matrix.machine_type }}
-          fi
-
-          echo "$machine_type"
-          echo "machine_type=$machine_type" >> $GITHUB_ENV
-
       - name: Run all pipeline tests on GPU
         working-directory: /transformers
         run: |
-          python3 -m pytest -n 1 -v --dist=loadfile --make-reports=${{ env.machine_type }}_run_pipelines_torch_gpu_test_reports tests/pipelines
+          python3 -m pytest -n 1 -v --dist=loadfile --make-reports=${{ matrix.machine_type }}_run_pipelines_torch_gpu_test_reports tests/pipelines
 
       - name: Failure short reports
         if: ${{ failure() }}
         continue-on-error: true
-        run: cat /transformers/reports/${{ env.machine_type }}_run_pipelines_torch_gpu_test_reports/failures_short.txt
+        run: cat /transformers/reports/${{ matrix.machine_type }}_run_pipelines_torch_gpu_test_reports/failures_short.txt
 
-      - name: "Test suite reports artifacts: ${{ env.machine_type }}_run_pipelines_torch_gpu_test_reports"
+      - name: "Test suite reports artifacts: ${{ matrix.machine_type }}_run_pipelines_torch_gpu_test_reports"
         if: ${{ always() }}
         uses: actions/upload-artifact@v4
         with:
-          name: ${{ env.machine_type }}_run_pipelines_torch_gpu_test_reports
-          path: /transformers/reports/${{ env.machine_type }}_run_pipelines_torch_gpu_test_reports
+          name: ${{ matrix.machine_type }}_run_pipelines_torch_gpu_test_reports
+          path: /transformers/reports/${{ matrix.machine_type }}_run_pipelines_torch_gpu_test_reports
+
+  run_pipelines_tf_gpu:
+    if: ${{ inputs.job == 'run_pipelines_tf_gpu' }}
+    name: TensorFlow pipelines
+    strategy:
+      fail-fast: false
+      matrix:
+        machine_type: [single-gpu, multi-gpu]
+    runs-on: ['${{ matrix.machine_type }}', nvidia-gpu, t4, '${{ inputs.runner }}']
+    container:
+      image: huggingface/transformers-tensorflow-gpu
+      options: --gpus all --shm-size "16gb" --ipc host -v /mnt/cache/.cache/huggingface:/mnt/cache/
+    steps:
+      - name: Update clone
+        working-directory: /transformers
+        run: |
+          git fetch && git checkout ${{ github.sha }}
+
+      - name: Reinstall transformers in edit mode (remove the one installed during docker image build)
+        working-directory: /transformers
+        run: python3 -m pip uninstall -y transformers && python3 -m pip install -e .
+
+      - name: NVIDIA-SMI
+        run: |
+          nvidia-smi
+
+      - name: Environment
+        working-directory: /transformers
+        run: |
+          python3 utils/print_env.py
+
+      - name: Show installed libraries and their versions
+        working-directory: /transformers
+        run: pip freeze
+
+      - name: Run all pipeline tests on GPU
+        working-directory: /transformers
+        run: |
+          python3 -m pytest -n 1 -v --dist=loadfile --make-reports=${{ matrix.machine_type }}_run_pipelines_tf_gpu_test_reports tests/pipelines
+
+      - name: Failure short reports
+        if: ${{ always() }}
+        run: |
+          cat /transformers/reports/${{ matrix.machine_type }}_run_pipelines_tf_gpu_test_reports/failures_short.txt
+
+      - name: "Test suite reports artifacts: ${{ matrix.machine_type }}_run_pipelines_tf_gpu_test_reports"
+        if: ${{ always() }}
+        uses: actions/upload-artifact@v4
+        with:
+          name: ${{ matrix.machine_type }}_run_pipelines_tf_gpu_test_reports
+          path: /transformers/reports/${{ matrix.machine_type }}_run_pipelines_tf_gpu_test_reports
 
   run_examples_gpu:
     if: ${{ inputs.job == 'run_examples_gpu' }}
@@ -213,9 +220,8 @@ jobs:
     strategy:
       fail-fast: false
       matrix:
-        machine_type: [aws-g4dn-4xlarge-cache]
-    runs-on:
-      group: '${{ matrix.machine_type }}'
+        machine_type: [single-gpu]
+    runs-on: ['${{ matrix.machine_type }}', nvidia-gpu, t4, '${{ inputs.runner }}']
     container:
       image: huggingface/transformers-all-latest-gpu
       options: --gpus 0 --shm-size "16gb" --ipc host -v /mnt/cache/.cache/huggingface:/mnt/cache/
@@ -241,40 +247,23 @@ jobs:
         working-directory: /transformers
         run: pip freeze
 
-      - name: Set `machine_type` for report and artifact names
-        working-directory: /transformers
-        shell: bash
-        run: |
-          echo "${{ matrix.machine_type }}"
-
-          if [ "${{ matrix.machine_type }}" = "aws-g4dn-4xlarge-cache" ]; then
-            machine_type=single-gpu
-          elif [ "${{ matrix.machine_type }}" = "aws-g4dn-12xlarge-cache" ]; then
-            machine_type=multi-gpu
-          else
-            machine_type=${{ matrix.machine_type }}
-          fi
-
-          echo "$machine_type"
-          echo "machine_type=$machine_type" >> $GITHUB_ENV
-
       - name: Run examples tests on GPU
         working-directory: /transformers
         run: |
           pip install -r examples/pytorch/_tests_requirements.txt
-          python3 -m pytest -v --make-reports=${{ env.machine_type }}_run_examples_gpu_test_reports examples/pytorch
+          python3 -m pytest -v --make-reports=${{ matrix.machine_type }}_run_examples_gpu_test_reports examples/pytorch
 
       - name: Failure short reports
         if: ${{ failure() }}
         continue-on-error: true
-        run: cat /transformers/reports/${{ env.machine_type }}_run_examples_gpu_test_reports/failures_short.txt
+        run: cat /transformers/reports/${{ matrix.machine_type }}_run_examples_gpu_test_reports/failures_short.txt
 
-      - name: "Test suite reports artifacts: ${{ env.machine_type }}_run_examples_gpu_test_reports"
+      - name: "Test suite reports artifacts: ${{ matrix.machine_type }}_run_examples_gpu_test_reports"
         if: ${{ always() }}
         uses: actions/upload-artifact@v4
         with:
-          name: ${{ env.machine_type }}_run_examples_gpu_test_reports
-          path: /transformers/reports/${{ env.machine_type }}_run_examples_gpu_test_reports
+          name: ${{ matrix.machine_type }}_run_examples_gpu_test_reports
+          path: /transformers/reports/${{ matrix.machine_type }}_run_examples_gpu_test_reports
 
   run_torch_cuda_extensions_gpu:
     if: ${{ inputs.job == 'run_torch_cuda_extensions_gpu' }}
@@ -282,9 +271,8 @@ jobs:
     strategy:
       fail-fast: false
       matrix:
-        machine_type: [aws-g4dn-4xlarge-cache, aws-g4dn-12xlarge-cache]
-    runs-on:
-      group: '${{ matrix.machine_type }}'
+        machine_type: [single-gpu, multi-gpu]
+    runs-on: ['${{ matrix.machine_type }}', nvidia-gpu, t4, '${{ inputs.runner }}']
     container:
       image: ${{ inputs.docker }}
       options: --gpus all --shm-size "16gb" --ipc host -v /mnt/cache/.cache/huggingface:/mnt/cache/
@@ -322,7 +310,7 @@ jobs:
         run: |
           python3 -m pip uninstall -y deepspeed
           rm -rf DeepSpeed
-          git clone https://github.com/deepspeedai/DeepSpeed && cd DeepSpeed && rm -rf build
+          git clone https://github.com/microsoft/DeepSpeed && cd DeepSpeed && rm -rf build
           DS_BUILD_CPU_ADAM=1 DS_BUILD_FUSED_ADAM=1 python3 -m pip install . --global-option="build_ext" --global-option="-j8" --no-cache -v --disable-pip-version-check
 
       - name: NVIDIA-SMI
@@ -338,39 +326,22 @@ jobs:
         working-directory: ${{ inputs.working-directory-prefix }}/transformers
         run: pip freeze
 
-      - name: Set `machine_type` for report and artifact names
-        working-directory: ${{ inputs.working-directory-prefix }}/transformers
-        shell: bash
-        run: |
-          echo "${{ matrix.machine_type }}"
-
-          if [ "${{ matrix.machine_type }}" = "aws-g4dn-4xlarge-cache" ]; then
-            machine_type=single-gpu
-          elif [ "${{ matrix.machine_type }}" = "aws-g4dn-12xlarge-cache" ]; then
-            machine_type=multi-gpu
-          else
-            machine_type=${{ matrix.machine_type }}
-          fi
-
-          echo "$machine_type"
-          echo "machine_type=$machine_type" >> $GITHUB_ENV
-
       - name: Run all tests on GPU
         working-directory: ${{ inputs.working-directory-prefix }}/transformers
         run: |
-          python3 -m pytest -v --make-reports=${{ env.machine_type }}_run_torch_cuda_extensions_gpu_test_reports tests/deepspeed tests/extended
+          python3 -m pytest -v --make-reports=${{ matrix.machine_type }}_run_torch_cuda_extensions_gpu_test_reports tests/deepspeed tests/extended
 
       - name: Failure short reports
         if: ${{ failure() }}
         continue-on-error: true
-        run: cat ${{ inputs.working-directory-prefix }}/transformers/reports/${{ env.machine_type }}_run_torch_cuda_extensions_gpu_test_reports/failures_short.txt
+        run: cat ${{ inputs.working-directory-prefix }}/transformers/reports/${{ matrix.machine_type }}_run_torch_cuda_extensions_gpu_test_reports/failures_short.txt
 
-      - name: "Test suite reports artifacts: ${{ env.machine_type }}_run_torch_cuda_extensions_gpu_test_reports"
+      - name: "Test suite reports artifacts: ${{ matrix.machine_type }}_run_torch_cuda_extensions_gpu_test_reports"
         if: ${{ always() }}
         uses: actions/upload-artifact@v4
         with:
-          name: ${{ env.machine_type }}_run_torch_cuda_extensions_gpu_test_reports
-          path: ${{ inputs.working-directory-prefix }}/transformers/reports/${{ env.machine_type }}_run_torch_cuda_extensions_gpu_test_reports
+          name: ${{ matrix.machine_type }}_run_torch_cuda_extensions_gpu_test_reports
+          path: ${{ inputs.working-directory-prefix }}/transformers/reports/${{ matrix.machine_type }}_run_torch_cuda_extensions_gpu_test_reports
 
   run_quantization_torch_gpu:
     if: ${{ inputs.job == 'run_quantization_torch_gpu' }}
@@ -381,9 +352,8 @@ jobs:
       fail-fast: false
       matrix:
         folders: ${{ fromJson(needs.setup.outputs.quantization_matrix) }}
-        machine_type: [aws-g4dn-4xlarge-cache, aws-g4dn-12xlarge-cache]
-    runs-on:
-      group: '${{ matrix.machine_type }}'
+        machine_type: [single-gpu, multi-gpu]
+    runs-on: ['${{ matrix.machine_type }}', nvidia-gpu, t4, '${{ inputs.runner }}']
     container:
       image: huggingface/transformers-quantization-latest-gpu
       options: --gpus all --shm-size "16gb" --ipc host -v /mnt/cache/.cache/huggingface:/mnt/cache/
@@ -418,39 +388,22 @@ jobs:
         working-directory: /transformers
         run: pip freeze
 
-      - name: Set `machine_type` for report and artifact names
-        working-directory: /transformers
-        shell: bash
-        run: |
-          echo "${{ matrix.machine_type }}"
-
-          if [ "${{ matrix.machine_type }}" = "aws-g4dn-4xlarge-cache" ]; then
-            machine_type=single-gpu
-          elif [ "${{ matrix.machine_type }}" = "aws-g4dn-12xlarge-cache" ]; then
-            machine_type=multi-gpu
-          else
-            machine_type=${{ matrix.machine_type }}
-          fi
-
-          echo "$machine_type"
-          echo "machine_type=$machine_type" >> $GITHUB_ENV
-
       - name: Run quantization tests on GPU
         working-directory: /transformers
         run: |
-          python3 -m pytest -v --make-reports=${{ env.machine_type }}_run_quantization_torch_gpu_${{ matrix.folders }}_test_reports tests/${{ matrix.folders }}
+          python3 -m pytest -v --make-reports=${{ matrix.machine_type }}_run_quantization_torch_gpu_${{ matrix.folders }}_test_reports tests/${{ matrix.folders }}
 
       - name: Failure short reports
         if: ${{ failure() }}
         continue-on-error: true
-        run: cat /transformers/reports/${{ env.machine_type }}_run_quantization_torch_gpu_${{ matrix.folders }}_test_reports/failures_short.txt
+        run: cat /transformers/reports/${{ matrix.machine_type }}_run_quantization_torch_gpu_${{ matrix.folders }}_test_reports/failures_short.txt
 
-      - name: "Test suite reports artifacts: ${{ env.machine_type }}_run_quantization_torch_gpu_${{ env.matrix_folders }}_test_reports"
+      - name: "Test suite reports artifacts: ${{ matrix.machine_type }}_run_quantization_torch_gpu_${{ env.matrix_folders }}_test_reports"
         if: ${{ always() }}
         uses: actions/upload-artifact@v4
         with:
-          name: ${{ env.machine_type }}_run_quantization_torch_gpu_${{ env.matrix_folders }}_test_reports
-          path: /transformers/reports/${{ env.machine_type }}_run_quantization_torch_gpu_${{ matrix.folders }}_test_reports
+          name: ${{ matrix.machine_type }}_run_quantization_torch_gpu_${{ env.matrix_folders }}_test_reports
+          path: /transformers/reports/${{ matrix.machine_type }}_run_quantization_torch_gpu_${{ matrix.folders }}_test_reports
 
   run_extract_warnings:
     # Let's only do this for the job `run_models_gpu` to simplify the (already complex) logic.
@@ -498,8 +451,8 @@ jobs:
     needs: [
       setup,
       run_models_gpu,
-      run_trainer_and_fsdp_gpu,
       run_pipelines_torch_gpu,
+      run_pipelines_tf_gpu,
       run_examples_gpu,
       run_torch_cuda_extensions_gpu,
       run_quantization_torch_gpu,
@@ -516,21 +469,5 @@ jobs:
       folder_slices: ${{ needs.setup.outputs.folder_slices }}
       quantization_matrix: ${{ needs.setup.outputs.quantization_matrix }}
       ci_event: ${{ inputs.ci_event }}
-      report_repo_id: ${{ inputs.report_repo_id }}
-
-    secrets: inherit
-
-  check_new_failures:
-    if: ${{ always() && inputs.ci_event == 'Daily CI' && needs.send_results.result == 'success' }}
-    name: Check new failures
-    needs: send_results
-    uses: ./.github/workflows/check_failed_tests.yml
-    with:
-      docker: ${{ inputs.docker }}
-      start_sha: ${{ github.sha }}
-      job: ${{ inputs.job }}
-      slack_report_channel: ${{ inputs.slack_report_channel }}
-      ci_event: ${{ inputs.ci_event }}
-      report_repo_id: ${{ inputs.report_repo_id }}
 
     secrets: inherit

.github/workflows/slack-report.yml

@@ -21,9 +21,6 @@ on:
       ci_event:
         required: true
         type: string
-      report_repo_id:
-        required: true
-        type: string
 
 env:
   TRANSFORMERS_CI_RESULTS_UPLOAD_TOKEN: ${{ secrets.TRANSFORMERS_CI_RESULTS_UPLOAD_TOKEN }}
@@ -42,23 +39,8 @@ jobs:
 
       - uses: actions/checkout@v4
       - uses: actions/download-artifact@v4
-
-      - name: Prepare some setup values
-        run: |
-          if [ -f setup_values/prev_workflow_run_id.txt ]; then
-            echo "PREV_WORKFLOW_RUN_ID=$(cat setup_values/prev_workflow_run_id.txt)" >> $GITHUB_ENV
-          else
-            echo "PREV_WORKFLOW_RUN_ID=" >> $GITHUB_ENV
-          fi
-
-          if [ -f setup_values/other_workflow_run_id.txt ]; then
-            echo "OTHER_WORKFLOW_RUN_ID=$(cat setup_values/other_workflow_run_id.txt)" >> $GITHUB_ENV
-          else
-            echo "OTHER_WORKFLOW_RUN_ID=" >> $GITHUB_ENV
-          fi
-
       - name: Send message to Slack
-        shell: bash
+        if: ${{ inputs.job != 'run_quantization_torch_gpu' }}
         env:
           CI_SLACK_BOT_TOKEN: ${{ secrets.CI_SLACK_BOT_TOKEN }}
           CI_SLACK_CHANNEL_ID: ${{ secrets.CI_SLACK_CHANNEL_ID }}
@@ -68,22 +50,19 @@ jobs:
           ACCESS_REPO_INFO_TOKEN: ${{ secrets.ACCESS_REPO_INFO_TOKEN }}
           CI_EVENT: ${{ inputs.ci_event }}
           CI_SHA: ${{ github.sha }}
+          CI_WORKFLOW_REF: ${{ github.workflow_ref }}
           CI_TEST_JOB: ${{ inputs.job }}
           SETUP_STATUS: ${{ inputs.setup_status }}
-          REPORT_REPO_ID: ${{ inputs.report_repo_id }}
         # We pass `needs.setup.outputs.matrix` as the argument. A processing in `notification_service.py` to change
         # `models/bert` to `models_bert` is required, as the artifact names use `_` instead of `/`.
         # For a job that doesn't depend on (i.e. `needs`) `setup`, the value for `inputs.folder_slices` would be an
         # empty string, and the called script still get one argument (which is the emtpy string).
         run: |
+          sudo apt-get install -y curl
           pip install huggingface_hub
           pip install slack_sdk
           pip show slack_sdk
-          if [ "${{ inputs.quantization_matrix }}" != "" ]; then
-            python utils/notification_service.py "${{ inputs.quantization_matrix }}"
-          else
-            python utils/notification_service.py "${{ inputs.folder_slices }}"
-          fi          
+          python utils/notification_service.py "${{ inputs.folder_slices }}"
 
       # Upload complete failure tables, as they might be big and only truncated versions could be sent to Slack.
       - name: Failure table artifacts
@@ -91,3 +70,32 @@ jobs:
         with:
           name: ci_results_${{ inputs.job }}
           path: ci_results_${{ inputs.job }}
+      
+      - uses: actions/checkout@v4
+      - uses: actions/download-artifact@v4
+      - name: Send message to Slack for quantization workflow
+        if: ${{ inputs.job == 'run_quantization_torch_gpu' }}
+        env:
+          CI_SLACK_BOT_TOKEN: ${{ secrets.CI_SLACK_BOT_TOKEN }}
+          ACCESS_REPO_INFO_TOKEN: ${{ secrets.ACCESS_REPO_INFO_TOKEN }}
+          SLACK_REPORT_CHANNEL: ${{ inputs.slack_report_channel }}
+          CI_EVENT: ${{ inputs.ci_event }}
+          CI_SHA: ${{ github.sha }}
+          CI_TEST_JOB: ${{ inputs.job }}
+          SETUP_STATUS: ${{ inputs.setup_status }}
+        # We pass `needs.setup.outputs.quantization_matrix` as the argument. A processing in `notification_service_quantization.py` to change
+        # `quantization/bnb` to `quantization_bnb` is required, as the artifact names use `_` instead of `/`.
+        run: |
+          sudo apt-get install -y curl
+          pip install huggingface_hub
+          pip install slack_sdk
+          pip show slack_sdk
+          python utils/notification_service_quantization.py "${{ inputs.quantization_matrix }}" 
+
+      # Upload complete failure tables, as they might be big and only truncated versions could be sent to Slack.
+      - name: Failure table artifacts
+        if: ${{ inputs.job == 'run_quantization_torch_gpu' }}
+        uses: actions/upload-artifact@v4
+        with:
+          name: ci_results_${{ inputs.job }}
+          path: ci_results_${{ inputs.job }}

.github/workflows/ssh-runner.yml

@@ -5,7 +5,7 @@ on:
     inputs:
       runner_type:
         description: 'Type of runner to test (a10 or t4)'
-        required: true
+        required: true 
       docker_image:
         description: 'Name of the Docker image'
         required: true
@@ -15,48 +15,20 @@ on:
 
 env:
   HF_HUB_READ_TOKEN: ${{ secrets.HF_HUB_READ_TOKEN }}
-  HF_HOME: /mnt/cache
-  TRANSFORMERS_IS_CI: yes
-  OMP_NUM_THREADS: 8
-  MKL_NUM_THREADS: 8
-  RUN_SLOW: yes # For gated repositories, we still need to agree to share information on the Hub repo. page in order to get access. # This token is created under the bot `hf-transformers-bot`.
-  SIGOPT_API_TOKEN: ${{ secrets.SIGOPT_API_TOKEN }}
-  TF_FORCE_GPU_ALLOW_GROWTH: true
+  HF_HOME: /mnt/cache 
+  TRANSFORMERS_IS_CI: yes 
+  OMP_NUM_THREADS: 8 
+  MKL_NUM_THREADS: 8 
+  RUN_SLOW: yes # For gated repositories, we still need to agree to share information on the Hub repo. page in order to get access. # This token is created under the bot `hf-transformers-bot`. 
+  SIGOPT_API_TOKEN: ${{ secrets.SIGOPT_API_TOKEN }} 
+  TF_FORCE_GPU_ALLOW_GROWTH: true 
   CUDA_VISIBLE_DEVICES: 0,1
+  RUN_PT_TF_CROSS_TESTS: 1
 
 jobs:
-  get_runner:
-    name: "Get runner to use"
-    runs-on: ubuntu-22.04
-    outputs:
-      RUNNER: ${{ steps.set_runner.outputs.RUNNER }}
-    steps:
-      - name: Get runner to use
-        shell: bash
-        run: |
-          if [[ "${{ github.event.inputs.num_gpus }}" == "single" && "${{ github.event.inputs.runner_type }}" == "t4" ]]; then
-            echo "RUNNER=aws-g4dn-4xlarge-cache" >> $GITHUB_ENV
-          elif [[ "${{ github.event.inputs.num_gpus }}" == "multi" && "${{ github.event.inputs.runner_type }}" == "t4" ]]; then
-            echo "RUNNER=aws-g4dn-12xlarge-cache" >> $GITHUB_ENV
-          elif [[ "${{ github.event.inputs.num_gpus }}" == "single" && "${{ github.event.inputs.runner_type }}" == "a10" ]]; then
-            echo "RUNNER=aws-g5-4xlarge-cache" >> $GITHUB_ENV
-          elif [[ "${{ github.event.inputs.num_gpus }}" == "multi" && "${{ github.event.inputs.runner_type }}" == "a10" ]]; then
-            echo "RUNNER=aws-g5-12xlarge-cache" >> $GITHUB_ENV
-          else
-            echo "RUNNER=" >> $GITHUB_ENV
-          fi
-
-      - name: Set runner to use
-        id: set_runner
-        run: |
-          echo ${{ env.RUNNER }}
-          echo "RUNNER=${{ env.RUNNER }}" >> $GITHUB_OUTPUT
-
   ssh_runner:
     name: "SSH"
-    needs: get_runner
-    runs-on:
-      group: ${{ needs.get_runner.outputs.RUNNER }}
+    runs-on: ["${{ github.event.inputs.num_gpus }}-gpu", nvidia-gpu, "${{ github.event.inputs.runner_type }}", ci]
     container:
       image: ${{ github.event.inputs.docker_image }}
       options: --gpus all --privileged --ipc host -v /mnt/cache/.cache/huggingface:/mnt/cache/
@@ -77,37 +49,15 @@ jobs:
       - name: Show installed libraries and their versions
         working-directory: /transformers
         run: pip freeze
-
+      
       - name: NVIDIA-SMI
         run: |
           nvidia-smi
-
-      - name: Store Slack infos
-        #because the SSH can be enabled dynamically if the workflow failed, so we need to store slack infos to be able to retrieve them during the waitforssh step
-        shell: bash
-        run: |
-          echo "${{ github.actor }}"
-          github_actor=${{ github.actor }}
-          github_actor=${github_actor/'-'/'_'}
-          echo "$github_actor"
-          echo "github_actor=$github_actor" >> $GITHUB_ENV
-
-      - name: Store Slack infos
-        #because the SSH can be enabled dynamically if the workflow failed, so we need to store slack infos to be able to retrieve them during the waitforssh step
-        shell: bash
-        run: |
-          echo "${{ env.github_actor }}"
-          if [ "${{ secrets[format('{0}_{1}', env.github_actor, 'SLACK_ID')] }}" != "" ]; then
-            echo "SLACKCHANNEL=${{ secrets[format('{0}_{1}', env.github_actor, 'SLACK_ID')] }}" >> $GITHUB_ENV
-          else
-            echo "SLACKCHANNEL=${{ secrets.SLACK_CIFEEDBACK_CHANNEL }}" >> $GITHUB_ENV
-          fi
-
+      
       - name: Tailscale # In order to be able to SSH when a test fails
         uses: huggingface/tailscale-action@main
         with:
           authkey: ${{ secrets.TAILSCALE_SSH_AUTHKEY }}
-          slackChannel: ${{ env.SLACKCHANNEL }}
+          slackChannel: ${{ secrets.SLACK_CIFEEDBACK_CHANNEL }}
           slackToken: ${{ secrets.SLACK_CIFEEDBACK_BOT_TOKEN }}
           waitForSSH: true
-          sshTimeout: 15m

.github/workflows/stale.yml

@@ -9,15 +9,13 @@ jobs:
     name: Close Stale Issues
     if: github.repository == 'huggingface/transformers'
     runs-on: ubuntu-22.04
-    permissions:
-      issues: write
     env:
       GITHUB_TOKEN: ${{ secrets.GITHUB_TOKEN }}
     steps:
     - uses: actions/checkout@v4
 
     - name: Setup Python
-      uses: actions/setup-python@v5
+      uses: actions/setup-python@v4
       with:
         python-version: 3.8
 

.github/workflows/trufflehog.yml

@@ -16,5 +16,3 @@ jobs:
           fetch-depth: 0
       - name: Secret Scanning
         uses: trufflesecurity/trufflehog@main
-        with:
-          extra_args: --results=verified,unknown

.github/workflows/update_metdata.yml

@@ -19,7 +19,7 @@ jobs:
       - name: Setup environment
         run: |
           pip install --upgrade pip
-          pip install datasets pandas
+          pip install datasets pandas==2.0.3
           pip install .[torch,tf,flax]
 
       - name: Update metadata

AGENTS.md

@@ -1,39 +0,0 @@
-# AGENTS.md Guide for Hugging Face Transformers
-
-This AGENTS.md file provides guidance for code agents working with this codebase.
-
-## Core Project Structure
-
-- `/src/transformers`: This contains the core source code for the library
-  - `/models`: Code for individual models. Models inherit from base classes in the root `/src/transformers` directory.
-- `/tests`: This contains the core test classes for the library. These are usually inherited rather than directly run.
-  - `/models`: Tests for individual models. Model tests inherit from common tests in the root `/tests` directory.
-- `/docs`: This contains the documentation for the library, including guides, tutorials, and API references.
-
-## Coding Conventions for Hugging Face Transformers
-
-- PRs should be as brief as possible. Bugfix PRs in particular can often be only one or two lines long, and do not need large comments, docstrings or new functions in this case. Aim to minimize the size of the diff.
-- When writing tests, they should be added to an existing file. The only exception is for PRs to add a new model, when a new test directory should be created for that model.
-- Code style is enforced in the CI. You can install the style tools with `pip install -e .[quality]`. You can then run `make fixup` to apply style and consistency fixes to your code.
-
-## Copying and inheritance
-
-Many models in the codebase have similar code, but it is not shared by inheritance because we want each model file to be self-contained.
-We use two mechanisms to keep this code in sync:
-
-- "Copied from" syntax. Functions or entire classes can have a comment at the top like this: `# Copied from transformers.models.llama.modeling_llama.rotate_half` or `# Copied from transformers.models.t5.modeling_t5.T5LayerNorm with T5->MT5`
-  These comments are actively checked by the style tools, and copies will automatically be updated when the base code is updated. If you need to update a copied function, you should
-  either update the base function and use `make fixup` to propagate the change to all copies, or simply remove the `# Copied from` comment if that is inappropriate.
-- "Modular" files. These files briefly define models by composing them using inheritance from other models. They are not meant to be used directly. Instead, the style tools
-  automatically generate a complete modeling file, like `modeling_bert.py`, from the modular file like `modular_bert.py`. If a model has a modular file, the modeling file
-  should never be edited directly! Instead, changes should be made in the modular file, and then you should run `make fixup` to update the modeling file automatically.
-
-When adding new models, you should prefer `modular` style.
-
-## Testing
-
-After making changes, you should usually run `make fixup` to ensure any copies and modular files are updated, and then test all affected models. This includes both
-the model you made the changes in and any other models that were updated by `make fixup`. Tests can be run with `pytest tests/models/[name]/test_modeling_[name].py`
-If your changes affect code in other classes like tokenizers or processors, you should run those tests instead, like `test_processing_[name].py` or `test_tokenization_[name].py`.
-
-In order to run tests, you may need to install dependencies. You can do this with `pip install -e .[testing]`. You will probably also need to `pip install torch accelerate` if your environment does not already have them.

CONTRIBUTING.md

@@ -78,7 +78,7 @@ Once you've confirmed the bug hasn't already been reported, please include the f
 To get the OS and software versions automatically, run the following command:
 
 ```bash
-transformers env
+transformers-cli env
 ```
 
 You can also run the same command from the root of the repository:
@@ -132,7 +132,7 @@ You will need basic `git` proficiency to contribute to
 manual. Type `git --help` in a shell and enjoy! If you prefer books, [Pro
 Git](https://git-scm.com/book/en/v2) is a very good reference.
 
-You'll need **[Python 3.9](https://github.com/huggingface/transformers/blob/main/setup.py#L449)** or above to contribute to 🤗 Transformers. Follow the steps below to start contributing:
+You'll need **[Python 3.8](https://github.com/huggingface/transformers/blob/main/setup.py#L449)** or above to contribute to 🤗 Transformers. Follow the steps below to start contributing:
 
 1. Fork the [repository](https://github.com/huggingface/transformers) by
    clicking on the **[Fork](https://github.com/huggingface/transformers/fork)** button on the repository's page. This creates a copy of the code
@@ -221,10 +221,10 @@ You'll need **[Python 3.9](https://github.com/huggingface/transformers/blob/main
    [Checks on a Pull Request](https://huggingface.co/docs/transformers/pr_checks) guide.
 
    If you're modifying documents under the `docs/source` directory, make sure the documentation can still be built. This check will also run in the CI when you open a pull request. To run a local check
-   make sure you install the [documentation builder](https://github.com/huggingface/doc-builder).
+   make sure you install the documentation builder:
 
    ```bash
-   pip install hf-doc-builder
+   pip install ".[docs]"
    ```
 
    Run the following command from the root of the repository:
@@ -343,6 +343,8 @@ RUN_SLOW=yes python -m pytest -n auto --dist=loadfile -s -v ./examples/pytorch/t
 
 Like the slow tests, there are other environment variables available which are not enabled by default during testing:
 - `RUN_CUSTOM_TOKENIZERS`: Enables tests for custom tokenizers.
+- `RUN_PT_FLAX_CROSS_TESTS`: Enables tests for PyTorch + Flax integration.
+- `RUN_PT_TF_CROSS_TESTS`: Enables tests for TensorFlow + PyTorch integration.
 
 More environment variables and additional information can be found in the [testing_utils.py](https://github.com/huggingface/transformers/blob/main/src/transformers/testing_utils.py).
 

ISSUES.md

@@ -26,7 +26,7 @@ There are two main venues to receive support: [the forums](https://discuss.huggi
 
 [The user forums](https://discuss.huggingface.co/) are supported by the wide community of the library users and backed up by developers when needed.
 
-If you have a difficulty with deploying this library or some questions, or you'd like to discuss a new feature, please first consider discussing those things at the forums. Only when you feel your subject matter has been crystallized and you still need support from the library developers do proceed to file an [issue](https://github.com/huggingface/transformers/issues).
+If you have a difficulty with deploying this library or some questions, or you'd like to discuss a new feature, please first consider discussing those things at the forums. Only when you feel your subject matter has been crystalized and you still need support from the library developers do proceed to file an [issue](https://github.com/huggingface/transformers/issues).
 
 In particular all "Please explain" questions or objectively very user-specific feature requests belong to the forums. Here are some example of such questions:
 
@@ -263,9 +263,9 @@ You are not required to read the following guidelines before opening an issue. H
     But if you're replying to a comment that happened some comments back it's always a good practice to quote just the relevant lines you're replying it. The `>` is used for quoting, or you can always use the menu to do so. For example your editor box will look like:
 
     ```
-    > How big is your GPU cluster?
+    > How big is your gpu cluster?
 
-    Our cluster is made of 256 GPUs.
+    Our cluster is made of 256 gpus.
     ```
 
     If you are addressing multiple comments, quote the relevant parts of each before your answer. Some people use the same comment to do multiple replies, others separate them into separate comments. Either way works. The latter approach helps for linking to a specific comment.

Makefile

@@ -8,19 +8,13 @@ check_dirs := examples tests src utils
 exclude_folders :=  ""
 
 modified_only_fixup:
-	@current_branch=$$(git branch --show-current); \
-	if [ "$$current_branch" = "main" ]; then \
-		echo "On main branch, running 'style' target instead..."; \
-		$(MAKE) style; \
+	$(eval modified_py_files := $(shell python utils/get_modified_files.py $(check_dirs)))
+	@if test -n "$(modified_py_files)"; then \
+		echo "Checking/fixing $(modified_py_files)"; \
+		ruff check $(modified_py_files) --fix --exclude $(exclude_folders); \
+		ruff format $(modified_py_files) --exclude $(exclude_folders);\
 	else \
-		modified_py_files=$$(python utils/get_modified_files.py $(check_dirs)); \
-		if [ -n "$$modified_py_files" ]; then \
-			echo "Checking/fixing files: $${modified_py_files}"; \
-			ruff check $${modified_py_files} --fix --exclude $(exclude_folders); \
-			ruff format $${modified_py_files} --exclude $(exclude_folders); \
-		else \
-			echo "No library .py files were modified"; \
-		fi; \
+		echo "No library .py files were modified"; \
 	fi
 
 # Update src/transformers/dependency_versions_table.py
@@ -42,16 +36,16 @@ autogenerate_code: deps_table_update
 
 repo-consistency:
 	python utils/check_copies.py
-	python utils/check_modular_conversion.py
+	python utils/check_table.py
 	python utils/check_dummies.py
 	python utils/check_repo.py
 	python utils/check_inits.py
-	python utils/check_pipeline_typing.py
 	python utils/check_config_docstrings.py
 	python utils/check_config_attributes.py
 	python utils/check_doctest_list.py
 	python utils/update_metadata.py --check-only
 	python utils/check_docstrings.py
+	python utils/check_support_list.py
 
 # this target runs checks on all files
 
@@ -59,6 +53,7 @@ quality:
 	@python -c "from transformers import *" || (echo '🚨 import failed, this means you introduced unprotected imports! 🚨'; exit 1)
 	ruff check $(check_dirs) setup.py conftest.py
 	ruff format --check $(check_dirs) setup.py conftest.py
+	python utils/custom_init_isort.py --check_only
 	python utils/sort_auto_mappings.py --check_only
 	python utils/check_doc_toc.py
 	python utils/check_docstrings.py --check_all
@@ -67,6 +62,7 @@ quality:
 # Format source code automatically and check is there are any problems left that need manual fixing
 
 extra_style_checks:
+	python utils/custom_init_isort.py
 	python utils/sort_auto_mappings.py
 	python utils/check_doc_toc.py --fix_and_overwrite
 
@@ -86,9 +82,8 @@ fixup: modified_only_fixup extra_style_checks autogenerate_code repo-consistency
 
 fix-copies:
 	python utils/check_copies.py --fix_and_overwrite
-	python utils/check_modular_conversion.py --fix_and_overwrite
+	python utils/check_table.py --fix_and_overwrite
 	python utils/check_dummies.py --fix_and_overwrite
-	python utils/check_pipeline_typing.py --fix_and_overwrite
 	python utils/check_doctest_list.py --fix_and_overwrite
 	python utils/check_docstrings.py --fix_and_overwrite
 

README.md

@@ -25,7 +25,6 @@ limitations under the License.
 </p>
 
 <p align="center">
-    <a href="https://huggingface.com/models"><img alt="Checkpoints on Hub" src="https://img.shields.io/endpoint?url=https://huggingface.co/api/shields/models&color=brightgreen"></a>
     <a href="https://circleci.com/gh/huggingface/transformers"><img alt="Build" src="https://img.shields.io/circleci/build/github/huggingface/transformers/main"></a>
     <a href="https://github.com/huggingface/transformers/blob/main/LICENSE"><img alt="GitHub" src="https://img.shields.io/github/license/huggingface/transformers.svg?color=blue"></a>
     <a href="https://huggingface.co/docs/transformers/index"><img alt="Documentation" src="https://img.shields.io/website/http/huggingface.co/docs/transformers/index.svg?down_color=red&down_message=offline&up_message=online"></a>
@@ -49,274 +48,259 @@ limitations under the License.
         <a href="https://github.com/huggingface/transformers/blob/main/i18n/README_fr.md">Français</a> |
         <a href="https://github.com/huggingface/transformers/blob/main/i18n/README_de.md">Deutsch</a> |
         <a href="https://github.com/huggingface/transformers/blob/main/i18n/README_vi.md">Tiếng Việt</a> |
-        <a href="https://github.com/huggingface/transformers/blob/main/i18n/README_ar.md">العربية</a> |
-        <a href="https://github.com/huggingface/transformers/blob/main/i18n/README_ur.md">اردو</a> |
     </p>
 </h4>
 
 <h3 align="center">
-    <p>State-of-the-art pretrained models for inference and training</p>
+    <p>State-of-the-art Machine Learning for JAX, PyTorch and TensorFlow</p>
 </h3>
 
 <h3 align="center">
-    <img src="https://huggingface.co/datasets/huggingface/documentation-images/resolve/main/transformers/transformers_as_a_model_definition.png"/>
+    <a href="https://hf.co/course"><img src="https://huggingface.co/datasets/huggingface/documentation-images/resolve/main/course_banner.png"></a>
 </h3>
 
+🤗 Transformers provides thousands of pretrained models to perform tasks on different modalities such as text, vision, and audio.
 
-Transformers acts as the model-definition framework for state-of-the-art machine learning models in text, computer 
-vision, audio, video, and multimodal model, for both inference and training. 
+These models can be applied on:
 
-It centralizes the model definition so that this definition is agreed upon across the ecosystem. `transformers` is the 
-pivot across frameworks: if a model definition is supported, it will be compatible with the majority of training 
-frameworks (Axolotl, Unsloth, DeepSpeed, FSDP, PyTorch-Lightning, ...), inference engines (vLLM, SGLang, TGI, ...),
-and adjacent modeling libraries (llama.cpp, mlx, ...) which leverage the model definition from `transformers`.
+* 📝 Text, for tasks like text classification, information extraction, question answering, summarization, translation, and text generation, in over 100 languages.
+* 🖼️ Images, for tasks like image classification, object detection, and segmentation.
+* 🗣️ Audio, for tasks like speech recognition and audio classification.
 
-We pledge to help support new state-of-the-art models and democratize their usage by having their model definition be
-simple, customizable, and efficient.
+Transformer models can also perform tasks on **several modalities combined**, such as table question answering, optical character recognition, information extraction from scanned documents, video classification, and visual question answering.
 
-There are over 1M+ Transformers [model checkpoints](https://huggingface.co/models?library=transformers&sort=trending) on the [Hugging Face Hub](https://huggingface.com/models) you can use.
+🤗 Transformers provides APIs to quickly download and use those pretrained models on a given text, fine-tune them on your own datasets and then share them with the community on our [model hub](https://huggingface.co/models). At the same time, each python module defining an architecture is fully standalone and can be modified to enable quick research experiments.
 
-Explore the [Hub](https://huggingface.com/) today to find a model and use Transformers to help you get started right away.
+🤗 Transformers is backed by the three most popular deep learning libraries — [Jax](https://jax.readthedocs.io/en/latest/), [PyTorch](https://pytorch.org/) and [TensorFlow](https://www.tensorflow.org/) — with a seamless integration between them. It's straightforward to train your models with one before loading them for inference with the other.
 
-## Installation
+## Online demos
 
-Transformers works with Python 3.9+ [PyTorch](https://pytorch.org/get-started/locally/) 2.1+, [TensorFlow](https://www.tensorflow.org/install/pip) 2.6+, and [Flax](https://flax.readthedocs.io/en/latest/) 0.4.1+.
+You can test most of our models directly on their pages from the [model hub](https://huggingface.co/models). We also offer [private model hosting, versioning, & an inference API](https://huggingface.co/pricing) for public and private models.
 
-Create and activate a virtual environment with [venv](https://docs.python.org/3/library/venv.html) or [uv](https://docs.astral.sh/uv/), a fast Rust-based Python package and project manager.
+Here are a few examples:
 
-```py
-# venv
-python -m venv .my-env
-source .my-env/bin/activate
-# uv
-uv venv .my-env
-source .my-env/bin/activate
+In Natural Language Processing:
+- [Masked word completion with BERT](https://huggingface.co/google-bert/bert-base-uncased?text=Paris+is+the+%5BMASK%5D+of+France)
+- [Named Entity Recognition with Electra](https://huggingface.co/dbmdz/electra-large-discriminator-finetuned-conll03-english?text=My+name+is+Sarah+and+I+live+in+London+city)
+- [Text generation with Mistral](https://huggingface.co/mistralai/Mistral-7B-Instruct-v0.2)
+- [Natural Language Inference with RoBERTa](https://huggingface.co/FacebookAI/roberta-large-mnli?text=The+dog+was+lost.+Nobody+lost+any+animal)
+- [Summarization with BART](https://huggingface.co/facebook/bart-large-cnn?text=The+tower+is+324+metres+%281%2C063+ft%29+tall%2C+about+the+same+height+as+an+81-storey+building%2C+and+the+tallest+structure+in+Paris.+Its+base+is+square%2C+measuring+125+metres+%28410+ft%29+on+each+side.+During+its+construction%2C+the+Eiffel+Tower+surpassed+the+Washington+Monument+to+become+the+tallest+man-made+structure+in+the+world%2C+a+title+it+held+for+41+years+until+the+Chrysler+Building+in+New+York+City+was+finished+in+1930.+It+was+the+first+structure+to+reach+a+height+of+300+metres.+Due+to+the+addition+of+a+broadcasting+aerial+at+the+top+of+the+tower+in+1957%2C+it+is+now+taller+than+the+Chrysler+Building+by+5.2+metres+%2817+ft%29.+Excluding+transmitters%2C+the+Eiffel+Tower+is+the+second+tallest+free-standing+structure+in+France+after+the+Millau+Viaduct)
+- [Question answering with DistilBERT](https://huggingface.co/distilbert/distilbert-base-uncased-distilled-squad?text=Which+name+is+also+used+to+describe+the+Amazon+rainforest+in+English%3F&context=The+Amazon+rainforest+%28Portuguese%3A+Floresta+Amaz%C3%B4nica+or+Amaz%C3%B4nia%3B+Spanish%3A+Selva+Amaz%C3%B3nica%2C+Amazon%C3%ADa+or+usually+Amazonia%3B+French%3A+For%C3%AAt+amazonienne%3B+Dutch%3A+Amazoneregenwoud%29%2C+also+known+in+English+as+Amazonia+or+the+Amazon+Jungle%2C+is+a+moist+broadleaf+forest+that+covers+most+of+the+Amazon+basin+of+South+America.+This+basin+encompasses+7%2C000%2C000+square+kilometres+%282%2C700%2C000+sq+mi%29%2C+of+which+5%2C500%2C000+square+kilometres+%282%2C100%2C000+sq+mi%29+are+covered+by+the+rainforest.+This+region+includes+territory+belonging+to+nine+nations.+The+majority+of+the+forest+is+contained+within+Brazil%2C+with+60%25+of+the+rainforest%2C+followed+by+Peru+with+13%25%2C+Colombia+with+10%25%2C+and+with+minor+amounts+in+Venezuela%2C+Ecuador%2C+Bolivia%2C+Guyana%2C+Suriname+and+French+Guiana.+States+or+departments+in+four+nations+contain+%22Amazonas%22+in+their+names.+The+Amazon+represents+over+half+of+the+planet%27s+remaining+rainforests%2C+and+comprises+the+largest+and+most+biodiverse+tract+of+tropical+rainforest+in+the+world%2C+with+an+estimated+390+billion+individual+trees+divided+into+16%2C000+species)
+- [Translation with T5](https://huggingface.co/google-t5/t5-base?text=My+name+is+Wolfgang+and+I+live+in+Berlin)
+
+In Computer Vision:
+- [Image classification with ViT](https://huggingface.co/google/vit-base-patch16-224)
+- [Object Detection with DETR](https://huggingface.co/facebook/detr-resnet-50)
+- [Semantic Segmentation with SegFormer](https://huggingface.co/nvidia/segformer-b0-finetuned-ade-512-512)
+- [Panoptic Segmentation with Mask2Former](https://huggingface.co/facebook/mask2former-swin-large-coco-panoptic)
+- [Depth Estimation with Depth Anything](https://huggingface.co/docs/transformers/main/model_doc/depth_anything)
+- [Video Classification with VideoMAE](https://huggingface.co/docs/transformers/model_doc/videomae)
+- [Universal Segmentation with OneFormer](https://huggingface.co/shi-labs/oneformer_ade20k_dinat_large)
+
+In Audio:
+- [Automatic Speech Recognition with Whisper](https://huggingface.co/openai/whisper-large-v3)
+- [Keyword Spotting with Wav2Vec2](https://huggingface.co/superb/wav2vec2-base-superb-ks)
+- [Audio Classification with Audio Spectrogram Transformer](https://huggingface.co/MIT/ast-finetuned-audioset-10-10-0.4593)
+
+In Multimodal tasks:
+- [Table Question Answering with TAPAS](https://huggingface.co/google/tapas-base-finetuned-wtq)
+- [Visual Question Answering with ViLT](https://huggingface.co/dandelin/vilt-b32-finetuned-vqa)
+- [Image captioning with LLaVa](https://huggingface.co/llava-hf/llava-1.5-7b-hf)
+- [Zero-shot Image Classification with SigLIP](https://huggingface.co/google/siglip-so400m-patch14-384)
+- [Document Question Answering with LayoutLM](https://huggingface.co/impira/layoutlm-document-qa)
+- [Zero-shot Video Classification with X-CLIP](https://huggingface.co/docs/transformers/model_doc/xclip)
+- [Zero-shot Object Detection with OWLv2](https://huggingface.co/docs/transformers/en/model_doc/owlv2)
+- [Zero-shot Image Segmentation with CLIPSeg](https://huggingface.co/docs/transformers/model_doc/clipseg)
+- [Automatic Mask Generation with SAM](https://huggingface.co/docs/transformers/model_doc/sam)
+
+
+## 100 projects using Transformers
+
+Transformers is more than a toolkit to use pretrained models: it's a community of projects built around it and the
+Hugging Face Hub. We want Transformers to enable developers, researchers, students, professors, engineers, and anyone
+else to build their dream projects.
+
+In order to celebrate the 100,000 stars of transformers, we have decided to put the spotlight on the
+community, and we have created the [awesome-transformers](./awesome-transformers.md) page which lists 100
+incredible projects built in the vicinity of transformers.
+
+If you own or use a project that you believe should be part of the list, please open a PR to add it!
+
+## If you are looking for custom support from the Hugging Face team
+
+<a target="_blank" href="https://huggingface.co/support">
+    <img alt="HuggingFace Expert Acceleration Program" src="https://cdn-media.huggingface.co/marketing/transformers/new-support-improved.png" style="max-width: 600px; border: 1px solid #eee; border-radius: 4px; box-shadow: 0 1px 2px 0 rgba(0, 0, 0, 0.05);">
+</a><br>
+
+## Quick tour
+
+To immediately use a model on a given input (text, image, audio, ...), we provide the `pipeline` API. Pipelines group together a pretrained model with the preprocessing that was used during that model's training. Here is how to quickly use a pipeline to classify positive versus negative texts:
+
+```python
+>>> from transformers import pipeline
+
+# Allocate a pipeline for sentiment-analysis
+>>> classifier = pipeline('sentiment-analysis')
+>>> classifier('We are very happy to introduce pipeline to the transformers repository.')
+[{'label': 'POSITIVE', 'score': 0.9996980428695679}]
 ```
 
-Install Transformers in your virtual environment.
+The second line of code downloads and caches the pretrained model used by the pipeline, while the third evaluates it on the given text. Here, the answer is "positive" with a confidence of 99.97%.
 
-```py
-# pip
-pip install "transformers[torch]"
+Many tasks have a pre-trained `pipeline` ready to go, in NLP but also in computer vision and speech. For example, we can easily extract detected objects in an image:
 
-# uv
-uv pip install "transformers[torch]"
+``` python
+>>> import requests
+>>> from PIL import Image
+>>> from transformers import pipeline
+
+# Download an image with cute cats
+>>> url = "https://huggingface.co/datasets/huggingface/documentation-images/resolve/main/coco_sample.png"
+>>> image_data = requests.get(url, stream=True).raw
+>>> image = Image.open(image_data)
+
+# Allocate a pipeline for object detection
+>>> object_detector = pipeline('object-detection')
+>>> object_detector(image)
+[{'score': 0.9982201457023621,
+  'label': 'remote',
+  'box': {'xmin': 40, 'ymin': 70, 'xmax': 175, 'ymax': 117}},
+ {'score': 0.9960021376609802,
+  'label': 'remote',
+  'box': {'xmin': 333, 'ymin': 72, 'xmax': 368, 'ymax': 187}},
+ {'score': 0.9954745173454285,
+  'label': 'couch',
+  'box': {'xmin': 0, 'ymin': 1, 'xmax': 639, 'ymax': 473}},
+ {'score': 0.9988006353378296,
+  'label': 'cat',
+  'box': {'xmin': 13, 'ymin': 52, 'xmax': 314, 'ymax': 470}},
+ {'score': 0.9986783862113953,
+  'label': 'cat',
+  'box': {'xmin': 345, 'ymin': 23, 'xmax': 640, 'ymax': 368}}]
 ```
 
-Install Transformers from source if you want the latest changes in the library or are interested in contributing. However, the *latest* version may not be stable. Feel free to open an [issue](https://github.com/huggingface/transformers/issues) if you encounter an error.
-
-```shell
-git clone https://github.com/huggingface/transformers.git
-cd transformers
-
-# pip
-pip install .[torch]
-
-# uv
-uv pip install .[torch]
-```
-
-## Quickstart
-
-Get started with Transformers right away with the [Pipeline](https://huggingface.co/docs/transformers/pipeline_tutorial) API. The `Pipeline` is a high-level inference class that supports text, audio, vision, and multimodal tasks. It handles preprocessing the input and returns the appropriate output.
-
-Instantiate a pipeline and specify model to use for text generation. The model is downloaded and cached so you can easily reuse it again. Finally, pass some text to prompt the model.
-
-```py
-from transformers import pipeline
-
-pipeline = pipeline(task="text-generation", model="Qwen/Qwen2.5-1.5B")
-pipeline("the secret to baking a really good cake is ")
-[{'generated_text': 'the secret to baking a really good cake is 1) to use the right ingredients and 2) to follow the recipe exactly. the recipe for the cake is as follows: 1 cup of sugar, 1 cup of flour, 1 cup of milk, 1 cup of butter, 1 cup of eggs, 1 cup of chocolate chips. if you want to make 2 cakes, how much sugar do you need? To make 2 cakes, you will need 2 cups of sugar.'}]
-```
-
-To chat with a model, the usage pattern is the same. The only difference is you need to construct a chat history (the input to `Pipeline`) between you and the system.
-
-> [!TIP]
-> You can also chat with a model directly from the command line.
-> ```shell
-> transformers chat Qwen/Qwen2.5-0.5B-Instruct
-> ```
-
-```py
-import torch
-from transformers import pipeline
-
-chat = [
-    {"role": "system", "content": "You are a sassy, wise-cracking robot as imagined by Hollywood circa 1986."},
-    {"role": "user", "content": "Hey, can you tell me any fun things to do in New York?"}
-]
-
-pipeline = pipeline(task="text-generation", model="meta-llama/Meta-Llama-3-8B-Instruct", torch_dtype=torch.bfloat16, device_map="auto")
-response = pipeline(chat, max_new_tokens=512)
-print(response[0]["generated_text"][-1]["content"])
-```
-
-Expand the examples below to see how `Pipeline` works for different modalities and tasks.
-
-<details>
-<summary>Automatic speech recognition</summary>
-
-```py
-from transformers import pipeline
-
-pipeline = pipeline(task="automatic-speech-recognition", model="openai/whisper-large-v3")
-pipeline("https://huggingface.co/datasets/Narsil/asr_dummy/resolve/main/mlk.flac")
-{'text': ' I have a dream that one day this nation will rise up and live out the true meaning of its creed.'}
-```
-
-</details>
-
-<details>
-<summary>Image classification</summary>
+Here, we get a list of objects detected in the image, with a box surrounding the object and a confidence score. Here is the original image on the left, with the predictions displayed on the right:
 
 <h3 align="center">
-    <a><img src="https://huggingface.co/datasets/Narsil/image_dummy/raw/main/parrots.png"></a>
+    <a><img src="https://huggingface.co/datasets/huggingface/documentation-images/resolve/main/coco_sample.png" width="400"></a>
+    <a><img src="https://huggingface.co/datasets/huggingface/documentation-images/resolve/main/coco_sample_post_processed.png" width="400"></a>
 </h3>
 
-```py
-from transformers import pipeline
+You can learn more about the tasks supported by the `pipeline` API in [this tutorial](https://huggingface.co/docs/transformers/task_summary).
 
-pipeline = pipeline(task="image-classification", model="facebook/dinov2-small-imagenet1k-1-layer")
-pipeline("https://huggingface.co/datasets/Narsil/image_dummy/raw/main/parrots.png")
-[{'label': 'macaw', 'score': 0.997848391532898},
- {'label': 'sulphur-crested cockatoo, Kakatoe galerita, Cacatua galerita',
-  'score': 0.0016551691805943847},
- {'label': 'lorikeet', 'score': 0.00018523589824326336},
- {'label': 'African grey, African gray, Psittacus erithacus',
-  'score': 7.85409429227002e-05},
- {'label': 'quail', 'score': 5.502637941390276e-05}]
+In addition to `pipeline`, to download and use any of the pretrained models on your given task, all it takes is three lines of code. Here is the PyTorch version:
+```python
+>>> from transformers import AutoTokenizer, AutoModel
+
+>>> tokenizer = AutoTokenizer.from_pretrained("google-bert/bert-base-uncased")
+>>> model = AutoModel.from_pretrained("google-bert/bert-base-uncased")
+
+>>> inputs = tokenizer("Hello world!", return_tensors="pt")
+>>> outputs = model(**inputs)
 ```
 
-</details>
+And here is the equivalent code for TensorFlow:
+```python
+>>> from transformers import AutoTokenizer, TFAutoModel
 
-<details>
-<summary>Visual question answering</summary>
+>>> tokenizer = AutoTokenizer.from_pretrained("google-bert/bert-base-uncased")
+>>> model = TFAutoModel.from_pretrained("google-bert/bert-base-uncased")
 
-
-<h3 align="center">
-    <a><img src="https://huggingface.co/datasets/huggingface/documentation-images/resolve/main/transformers/tasks/idefics-few-shot.jpg"></a>
-</h3>
-
-```py
-from transformers import pipeline
-
-pipeline = pipeline(task="visual-question-answering", model="Salesforce/blip-vqa-base")
-pipeline(
-    image="https://huggingface.co/datasets/huggingface/documentation-images/resolve/main/transformers/tasks/idefics-few-shot.jpg",
-    question="What is in the image?",
-)
-[{'answer': 'statue of liberty'}]
+>>> inputs = tokenizer("Hello world!", return_tensors="tf")
+>>> outputs = model(**inputs)
 ```
 
-</details>
+The tokenizer is responsible for all the preprocessing the pretrained model expects and can be called directly on a single string (as in the above examples) or a list. It will output a dictionary that you can use in downstream code or simply directly pass to your model using the ** argument unpacking operator.
 
-## Why should I use Transformers?
+The model itself is a regular [Pytorch `nn.Module`](https://pytorch.org/docs/stable/nn.html#torch.nn.Module) or a [TensorFlow `tf.keras.Model`](https://www.tensorflow.org/api_docs/python/tf/keras/Model) (depending on your backend) which you can use as usual. [This tutorial](https://huggingface.co/docs/transformers/training) explains how to integrate such a model into a classic PyTorch or TensorFlow training loop, or how to use our `Trainer` API to quickly fine-tune on a new dataset.
+
+## Why should I use transformers?
 
 1. Easy-to-use state-of-the-art models:
-    - High performance on natural language understanding & generation, computer vision, audio, video, and multimodal tasks.
-    - Low barrier to entry for researchers, engineers, and developers.
+    - High performance on natural language understanding & generation, computer vision, and audio tasks.
+    - Low barrier to entry for educators and practitioners.
     - Few user-facing abstractions with just three classes to learn.
     - A unified API for using all our pretrained models.
 
 1. Lower compute costs, smaller carbon footprint:
-    - Share trained models instead of training from scratch.
-    - Reduce compute time and production costs.
-    - Dozens of model architectures with 1M+ pretrained checkpoints across all modalities.
+    - Researchers can share trained models instead of always retraining.
+    - Practitioners can reduce compute time and production costs.
+    - Dozens of architectures with over 400,000 pretrained models across all modalities.
 
-1. Choose the right framework for every part of a models lifetime:
+1. Choose the right framework for every part of a model's lifetime:
     - Train state-of-the-art models in 3 lines of code.
-    - Move a single model between PyTorch/JAX/TF2.0 frameworks at will.
-    - Pick the right framework for training, evaluation, and production.
+    - Move a single model between TF2.0/PyTorch/JAX frameworks at will.
+    - Seamlessly pick the right framework for training, evaluation, and production.
 
 1. Easily customize a model or an example to your needs:
     - We provide examples for each architecture to reproduce the results published by its original authors.
     - Model internals are exposed as consistently as possible.
     - Model files can be used independently of the library for quick experiments.
 
-<a target="_blank" href="https://huggingface.co/enterprise">
-    <img alt="Hugging Face Enterprise Hub" src="https://github.com/user-attachments/assets/247fb16d-d251-4583-96c4-d3d76dda4925">
-</a><br>
-
-## Why shouldn't I use Transformers?
+## Why shouldn't I use transformers?
 
 - This library is not a modular toolbox of building blocks for neural nets. The code in the model files is not refactored with additional abstractions on purpose, so that researchers can quickly iterate on each of the models without diving into additional abstractions/files.
-- The training API is optimized to work with PyTorch models provided by Transformers. For generic machine learning loops, you should use another library like [Accelerate](https://huggingface.co/docs/accelerate).
-- The [example scripts]((https://github.com/huggingface/transformers/tree/main/examples)) are only *examples*. They may not necessarily work out-of-the-box on your specific use case and you'll need to adapt the code for it to work.
+- The training API is not intended to work on any model but is optimized to work with the models provided by the library. For generic machine learning loops, you should use another library (possibly, [Accelerate](https://huggingface.co/docs/accelerate)).
+- While we strive to present as many use cases as possible, the scripts in our [examples folder](https://github.com/huggingface/transformers/tree/main/examples) are just that: examples. It is expected that they won't work out-of-the-box on your specific problem and that you will be required to change a few lines of code to adapt them to your needs.
 
-## 100 projects using Transformers
+## Installation
 
-Transformers is more than a toolkit to use pretrained models, it's a community of projects built around it and the
-Hugging Face Hub. We want Transformers to enable developers, researchers, students, professors, engineers, and anyone
-else to build their dream projects.
+### With pip
 
-In order to celebrate Transformers 100,000 stars, we wanted to put the spotlight on the
-community with the [awesome-transformers](./awesome-transformers.md) page which lists 100
-incredible projects built with Transformers.
+This repository is tested on Python 3.8+, Flax 0.4.1+, PyTorch 1.11+, and TensorFlow 2.6+.
 
-If you own or use a project that you believe should be part of the list, please open a PR to add it!
+You should install 🤗 Transformers in a [virtual environment](https://docs.python.org/3/library/venv.html). If you're unfamiliar with Python virtual environments, check out the [user guide](https://packaging.python.org/guides/installing-using-pip-and-virtual-environments/).
 
-## Example models
+First, create a virtual environment with the version of Python you're going to use and activate it.
 
-You can test most of our models directly on their [Hub model pages](https://huggingface.co/models).
+Then, you will need to install at least one of Flax, PyTorch, or TensorFlow.
+Please refer to [TensorFlow installation page](https://www.tensorflow.org/install/), [PyTorch installation page](https://pytorch.org/get-started/locally/#start-locally) and/or [Flax](https://github.com/google/flax#quick-install) and [Jax](https://github.com/google/jax#installation) installation pages regarding the specific installation command for your platform.
 
-Expand each modality below to see a few example models for various use cases.
+When one of those backends has been installed, 🤗 Transformers can be installed using pip as follows:
 
-<details>
-<summary>Audio</summary>
+```bash
+pip install transformers
+```
 
-- Audio classification with [Whisper](https://huggingface.co/openai/whisper-large-v3-turbo)
-- Automatic speech recognition with [Moonshine](https://huggingface.co/UsefulSensors/moonshine)
-- Keyword spotting with [Wav2Vec2](https://huggingface.co/superb/wav2vec2-base-superb-ks)
-- Speech to speech generation with [Moshi](https://huggingface.co/kyutai/moshiko-pytorch-bf16)
-- Text to audio with [MusicGen](https://huggingface.co/facebook/musicgen-large)
-- Text to speech with [Bark](https://huggingface.co/suno/bark)
+If you'd like to play with the examples or need the bleeding edge of the code and can't wait for a new release, you must [install the library from source](https://huggingface.co/docs/transformers/installation#installing-from-source).
 
-</details>
+### With conda
 
-<details>
-<summary>Computer vision</summary>
+🤗 Transformers can be installed using conda as follows:
 
-- Automatic mask generation with [SAM](https://huggingface.co/facebook/sam-vit-base)
-- Depth estimation with [DepthPro](https://huggingface.co/apple/DepthPro-hf)
-- Image classification with [DINO v2](https://huggingface.co/facebook/dinov2-base)
-- Keypoint detection with [SuperGlue](https://huggingface.co/magic-leap-community/superglue_outdoor)
-- Keypoint matching with [SuperGlue](https://huggingface.co/magic-leap-community/superglue)
-- Object detection with [RT-DETRv2](https://huggingface.co/PekingU/rtdetr_v2_r50vd)
-- Pose Estimation with [VitPose](https://huggingface.co/usyd-community/vitpose-base-simple)
-- Universal segmentation with [OneFormer](https://huggingface.co/shi-labs/oneformer_ade20k_swin_large)
-- Video classification with [VideoMAE](https://huggingface.co/MCG-NJU/videomae-large)
+```shell script
+conda install conda-forge::transformers
+```
 
-</details>
+> **_NOTE:_** Installing `transformers` from the `huggingface` channel is deprecated.
 
-<details>
-<summary>Multimodal</summary>
+Follow the installation pages of Flax, PyTorch or TensorFlow to see how to install them with conda.
 
-- Audio or text to text with [Qwen2-Audio](https://huggingface.co/Qwen/Qwen2-Audio-7B)
-- Document question answering with [LayoutLMv3](https://huggingface.co/microsoft/layoutlmv3-base)
-- Image or text to text with [Qwen-VL](https://huggingface.co/Qwen/Qwen2.5-VL-3B-Instruct)
-- Image captioning [BLIP-2](https://huggingface.co/Salesforce/blip2-opt-2.7b)
-- OCR-based document understanding with [GOT-OCR2](https://huggingface.co/stepfun-ai/GOT-OCR-2.0-hf)
-- Table question answering with [TAPAS](https://huggingface.co/google/tapas-base)
-- Unified multimodal understanding and generation with [Emu3](https://huggingface.co/BAAI/Emu3-Gen)
-- Vision to text with [Llava-OneVision](https://huggingface.co/llava-hf/llava-onevision-qwen2-0.5b-ov-hf)
-- Visual question answering with [Llava](https://huggingface.co/llava-hf/llava-1.5-7b-hf)
-- Visual referring expression segmentation with [Kosmos-2](https://huggingface.co/microsoft/kosmos-2-patch14-224)
+> **_NOTE:_**  On Windows, you may be prompted to activate Developer Mode in order to benefit from caching. If this is not an option for you, please let us know in [this issue](https://github.com/huggingface/huggingface_hub/issues/1062).
 
-</details>
+## Model architectures
 
-<details>
-<summary>NLP</summary>
+**[All the model checkpoints](https://huggingface.co/models)** provided by 🤗 Transformers are seamlessly integrated from the huggingface.co [model hub](https://huggingface.co/models), where they are uploaded directly by [users](https://huggingface.co/users) and [organizations](https://huggingface.co/organizations).
 
-- Masked word completion with [ModernBERT](https://huggingface.co/answerdotai/ModernBERT-base)
-- Named entity recognition with [Gemma](https://huggingface.co/google/gemma-2-2b)
-- Question answering with [Mixtral](https://huggingface.co/mistralai/Mixtral-8x7B-v0.1)
-- Summarization with [BART](https://huggingface.co/facebook/bart-large-cnn)
-- Translation with [T5](https://huggingface.co/google-t5/t5-base)
-- Text generation with [Llama](https://huggingface.co/meta-llama/Llama-3.2-1B)
-- Text classification with [Qwen](https://huggingface.co/Qwen/Qwen2.5-0.5B)
+Current number of checkpoints: ![](https://img.shields.io/endpoint?url=https://huggingface.co/api/shields/models&color=brightgreen)
 
-</details>
+🤗 Transformers currently provides the following architectures: see [here](https://huggingface.co/docs/transformers/model_summary) for a high-level summary of each them.
+
+To check if each model has an implementation in Flax, PyTorch or TensorFlow, or has an associated tokenizer backed by the 🤗 Tokenizers library, refer to [this table](https://huggingface.co/docs/transformers/index#supported-frameworks).
+
+These implementations have been tested on several datasets (see the example scripts) and should match the performance of the original implementations. You can find more details on performance in the Examples section of the [documentation](https://github.com/huggingface/transformers/tree/main/examples).
+
+
+## Learn more
+
+| Section | Description |
+|-|-|
+| [Documentation](https://huggingface.co/docs/transformers/) | Full API documentation and tutorials |
+| [Task summary](https://huggingface.co/docs/transformers/task_summary) | Tasks supported by 🤗 Transformers |
+| [Preprocessing tutorial](https://huggingface.co/docs/transformers/preprocessing) | Using the `Tokenizer` class to prepare data for the models |
+| [Training and fine-tuning](https://huggingface.co/docs/transformers/training) | Using the models provided by 🤗 Transformers in a PyTorch/TensorFlow training loop and the `Trainer` API |
+| [Quick tour: Fine-tuning/usage scripts](https://github.com/huggingface/transformers/tree/main/examples) | Example scripts for fine-tuning models on a wide range of tasks |
+| [Model sharing and uploading](https://huggingface.co/docs/transformers/model_sharing) | Upload and share your fine-tuned models with the community |
 
 ## Citation
 

SECURITY.md

@@ -27,6 +27,14 @@ These models require the `trust_remote_code=True` parameter to be set when using
 the content of the modeling files when using this argument. We recommend setting a revision in order to ensure you
 protect yourself from updates on the repository.
 
+#### Tools
+
+Through the `Agent` framework, remote tools can be downloaded to be used by the Agent. You're to specify these tools
+yourself, but please keep in mind that their code will be run on your machine if the Agent chooses to run them.
+
+Please inspect the code of the tools before passing them to the Agent to protect your runtime and local setup.
+
 ## Reporting a Vulnerability
 
-Feel free to submit vulnerability reports to [security@huggingface.co](mailto:security@huggingface.co), where someone from the HF security team will review and recommend next steps. If reporting a vulnerability specific to open source, please note [Huntr](https://huntr.com) is a vulnerability disclosure program for open source software.
+🤗 Please feel free to submit vulnerability reports to our private bug bounty program at https://hackerone.com/hugging_face. You'll need to request access to the program by emailing security@huggingface.co.
+Note that you'll need to be invited to our program, so send us a quick email at security@huggingface.co if you've found a vulnerability.

awesome-transformers.md

@@ -15,7 +15,7 @@ to add it.
 
 Keywords: Open-source, LLaMa, GPT-J, instruction, assistant
 
-## [recommenders](https://github.com/recommenders-team/recommenders)
+## [recommenders](https://github.com/microsoft/recommenders)
 
 This repository contains examples and best practices for building recommendation systems, provided as Jupyter notebooks. It goes over several aspects required to build efficient recommendation systems: data preparation, modeling, evaluation, model selection & optimization, as well as operationalization
 
@@ -29,7 +29,7 @@ Keywords: inpainting, SD, Stable Diffusion
 
 ## [flair](https://github.com/flairNLP/flair)
 
-FLAIR is a powerful PyTorch NLP framework, covering several important tasks: NER, sentiment-analysis, part-of-speech tagging, text and document embeddings, among other things.
+FLAIR is a powerful PyTorch NLP framework, convering several important tasks: NER, sentiment-analysis, part-of-speech tagging, text and document embeddings, among other things.
 
 Keywords: NLP, text embedding, document embedding, biomedical, NER, PoS, sentiment-analysis
 
@@ -39,15 +39,15 @@ MindsDB is a low-code ML platform, which automates and integrates several ML fra
 
 Keywords: Database, low-code, AI table
 
-## [langchain](https://github.com/langchain-ai/langchain)
+## [langchain](https://github.com/hwchase17/langchain)
 
-[langchain](https://github.com/langchain-ai/langchain) is aimed at assisting in the development of apps merging both LLMs and other sources of knowledge. The library allows chaining calls to applications, creating a sequence across many tools.
+[langchain](https://github.com/hwchase17/langchain) is aimed at assisting in the development of apps merging both LLMs and other sources of knowledge. The library allows chaining calls to applications, creating a sequence across many tools.
 
 Keywords: LLMs, Large Language Models, Agents, Chains
 
-## [LlamaIndex](https://github.com/run-llama/llama_index)
+## [LlamaIndex](https://github.com/jerryjliu/llama_index)
 
-[LlamaIndex](https://github.com/run-llama/llama_index) is a project that provides a central interface to connect your LLM's with external data. It provides various kinds of indices and retrieval mechanisms to perform different LLM tasks and obtain knowledge-augmented results.
+[LlamaIndex](https://github.com/jerryjliu/llama_index) is a project that provides a central interface to connect your LLM's with external data. It provides various kinds of indices and retreival mechanisms to perform different LLM tasks and obtain knowledge-augmented results.
 
 Keywords: LLMs, Large Language Models, Data Retrieval, Indices, Knowledge Augmentation 
 
@@ -146,9 +146,9 @@ Keywords: Framework, simplicity, NLP
 
 Keywords: LLM, Agents, HF Hub
 
-## [transformers.js](https://github.com/huggingface/transformers.js/)
+## [transformers.js](https://xenova.github.io/transformers.js/)
 
-[transformers.js](https://github.com/huggingface/transformers.js/) is a JavaScript library targeted at running models from transformers directly within the browser.
+[transformers.js](https://xenova.github.io/transformers.js/) is a JavaScript library targeted at running models from transformers directly within the browser.
 
 Keywords: Transformers, JavaScript, browser
 
@@ -437,7 +437,7 @@ Keywords: DALL-E, Russian
 
 Keywords: Knowledge Extraction, Knowledge Graphs
 
-## [Nebuly](https://github.com/nebuly-ai/optimate)
+## [Nebuly](https://github.com/nebuly-ai/nebuly)
 
 Nebuly is the next-generation platform to monitor and optimize your AI costs in one place. The platform connects to all your AI cost sources (compute, API providers, AI software licenses, etc) and centralizes them in one place to give you full visibility on a model basis. The platform also provides optimization recommendations and a co-pilot model that can guide during the optimization process. The platform builds on top of the open-source tools allowing you to optimize the different steps of your AI stack to squeeze out the best possible cost performances.
 

benchmark/README.md

@@ -1,49 +0,0 @@
-# Benchmarks
-
-You might want to add new benchmarks.
-
-You will need to define a python function named `run_benchmark` in your python file and the file must be located in this `benchmark/` directory.
-
-The expected function signature is the following:
-
-```py
-def run_benchmark(logger: Logger, branch: str, commit_id: str, commit_msg: str, num_tokens_to_generate=100):
-```
-
-## Writing metrics to the database
-
-`MetricsRecorder` is thread-safe, in the sense of the python [`Thread`](https://docs.python.org/3/library/threading.html#threading.Thread). This means you can start a background thread to do the readings on the device measurements while not blocking the main thread to execute the model measurements.
-
-cf [`llama.py`](./llama.py) to see an example of this in practice.
-
-```py
-from benchmarks_entrypoint import MetricsRecorder
-import psycopg2
-
-def run_benchmark(logger: Logger, branch: str, commit_id: str, commit_msg: str, num_tokens_to_generate=100):
-  metrics_recorder = MetricsRecorder(psycopg2.connect("dbname=metrics"), logger, branch, commit_id, commit_msg)
-  benchmark_id = metrics_recorder.initialise_benchmark({"gpu_name": gpu_name, "model_id": model_id})
-    # To collect device measurements
-    metrics_recorder.collect_device_measurements(
-        benchmark_id, cpu_util, mem_megabytes, gpu_util, gpu_mem_megabytes
-    )
-    # To collect your model measurements
-    metrics_recorder.collect_model_measurements(
-        benchmark_id,
-        {
-            "model_load_time": model_load_time,
-            "first_eager_forward_pass_time_secs": first_eager_fwd_pass_time,
-            "second_eager_forward_pass_time_secs": second_eager_fwd_pass_time,
-            "first_eager_generate_time_secs": first_eager_generate_time,
-            "second_eager_generate_time_secs": second_eager_generate_time,
-            "time_to_first_token_secs": time_to_first_token,
-            "time_to_second_token_secs": time_to_second_token,
-            "time_to_third_token_secs": time_to_third_token,
-            "time_to_next_token_mean_secs": mean_time_to_next_token,
-            "first_compile_generate_time_secs": first_compile_generate_time,
-            "second_compile_generate_time_secs": second_compile_generate_time,
-            "third_compile_generate_time_secs": third_compile_generate_time,
-            "fourth_compile_generate_time_secs": fourth_compile_generate_time,
-        },
-    )
-```

benchmark/benchmark.py

@@ -90,7 +90,7 @@ def summarize(run_dir, metrics, expand_metrics=False):
 
         model = benchmark.config.backend["model"]
 
-        # This looks like `benchmark.input_shapes.batch_size=1,benchmark.input_shapes.sequence_length=5`.
+        # Ths looks like `benchmark.input_shapes.batch_size=1,benchmark.input_shapes.sequence_length=5`.
         # (we rely on the usage of hydra's `${hydra.job.override_dirname}`.)
         benchmark_name = re.sub(f"backend.model={model},*", "", report_dir)
         benchmark_name = str(Path(benchmark_name).parts[-1])
@@ -101,7 +101,7 @@ def summarize(run_dir, metrics, expand_metrics=False):
         # post-processing of report: show a few selected/important metric
         for metric in metrics:
             keys = metric.split(".")
-            value = report.to_dict()
+            value = report
             current = metrics_values
             for key in keys:
                 # Avoid KeyError when a user's specified metric has typo.

benchmark/benchmarks_entrypoint.py

@@ -1,152 +0,0 @@
-import argparse
-import importlib.util
-import logging
-import os
-import sys
-from typing import Dict, Tuple
-
-from psycopg2.extensions import register_adapter
-from psycopg2.extras import Json
-
-
-register_adapter(dict, Json)
-
-
-class ImportModuleException(Exception):
-    pass
-
-
-class MetricsRecorder:
-    def __init__(
-        self, connection, logger: logging.Logger, repository: str, branch: str, commit_id: str, commit_msg: str
-    ):
-        self.conn = connection
-        self.conn.autocommit = True
-        self.logger = logger
-        self.repository = repository
-        self.branch = branch
-        self.commit_id = commit_id
-        self.commit_msg = commit_msg
-
-    def initialise_benchmark(self, metadata: dict[str, str]) -> int:
-        """
-        Creates a new benchmark, returns the benchmark id
-        """
-        # gpu_name: str, model_id: str
-        with self.conn.cursor() as cur:
-            cur.execute(
-                "INSERT INTO benchmarks (repository, branch, commit_id, commit_message, metadata) VALUES (%s, %s, %s, %s, %s) RETURNING benchmark_id",
-                (self.repository, self.branch, self.commit_id, self.commit_msg, metadata),
-            )
-            benchmark_id = cur.fetchone()[0]
-            logger.debug(f"initialised benchmark #{benchmark_id}")
-            return benchmark_id
-
-    def collect_device_measurements(self, benchmark_id: int, cpu_util, mem_megabytes, gpu_util, gpu_mem_megabytes):
-        """
-        Collect device metrics, such as CPU & GPU usage. These are "static", as in you cannot pass arbitrary arguments to the function.
-        """
-        with self.conn.cursor() as cur:
-            cur.execute(
-                "INSERT INTO device_measurements (benchmark_id, cpu_util, mem_megabytes, gpu_util, gpu_mem_megabytes) VALUES (%s, %s, %s, %s, %s)",
-                (benchmark_id, cpu_util, mem_megabytes, gpu_util, gpu_mem_megabytes),
-            )
-        self.logger.debug(
-            f"inserted device measurements for benchmark #{benchmark_id} [CPU util: {cpu_util}, mem MBs: {mem_megabytes}, GPU util: {gpu_util}, GPU mem MBs: {gpu_mem_megabytes}]"
-        )
-
-    def collect_model_measurements(self, benchmark_id: int, measurements: dict[str, float]):
-        with self.conn.cursor() as cur:
-            cur.execute(
-                """
-                INSERT INTO model_measurements (
-                    benchmark_id,
-                    measurements
-                ) VALUES (%s, %s)
-                """,
-                (
-                    benchmark_id,
-                    measurements,
-                ),
-            )
-        self.logger.debug(f"inserted model measurements for benchmark #{benchmark_id}: {measurements}")
-
-    def close(self):
-        self.conn.close()
-
-
-logger = logging.getLogger(__name__)
-logger.setLevel(logging.INFO)
-
-handler = logging.StreamHandler(sys.stdout)
-handler.setLevel(logging.INFO)
-formatter = logging.Formatter("[%(levelname)s - %(asctime)s] %(message)s")
-handler.setFormatter(formatter)
-logger.addHandler(handler)
-
-
-def parse_arguments() -> tuple[str, str, str, str]:
-    """
-    Parse command line arguments for the benchmarking CLI.
-    """
-    parser = argparse.ArgumentParser(description="CLI for benchmarking the huggingface/transformers.")
-
-    parser.add_argument(
-        "repository",
-        type=str,
-        help="The repository name on which the benchmarking is performed.",
-    )
-
-    parser.add_argument(
-        "branch",
-        type=str,
-        help="The branch name on which the benchmarking is performed.",
-    )
-
-    parser.add_argument(
-        "commit_id",
-        type=str,
-        help="The commit hash on which the benchmarking is performed.",
-    )
-
-    parser.add_argument(
-        "commit_msg",
-        type=str,
-        help="The commit message associated with the commit, truncated to 70 characters.",
-    )
-
-    args = parser.parse_args()
-
-    return args.repository, args.branch, args.commit_id, args.commit_msg
-
-
-def import_from_path(module_name, file_path):
-    try:
-        spec = importlib.util.spec_from_file_location(module_name, file_path)
-        module = importlib.util.module_from_spec(spec)
-        sys.modules[module_name] = module
-        spec.loader.exec_module(module)
-        return module
-    except Exception as e:
-        raise ImportModuleException(f"failed to load python module: {e}")
-
-
-if __name__ == "__main__":
-    benchmarks_folder_path = os.path.dirname(os.path.realpath(__file__))
-
-    repository, branch, commit_id, commit_msg = parse_arguments()
-
-    for entry in os.scandir(benchmarks_folder_path):
-        try:
-            if not entry.name.endswith(".py"):
-                continue
-            if entry.path == __file__:
-                continue
-            logger.debug(f"loading: {entry.name}")
-            module = import_from_path(entry.name.split(".")[0], entry.path)
-            logger.info(f"running benchmarks in: {entry.name}")
-            module.run_benchmark(logger, repository, branch, commit_id, commit_msg)
-        except ImportModuleException as e:
-            logger.error(e)
-        except Exception as e:
-            logger.error(f"error running benchmarks for {entry.name}: {e}")

benchmark/default.yml

@@ -1,10 +0,0 @@
-apiVersion: 1
-
-providers:
-  - name: 'Transformers Benchmarks'
-    orgId: 1
-    type: file
-    updateIntervalSeconds: 10
-    allowUiUpdates: true
-    options:
-      path: /etc/grafana/dashboards

benchmark/grafana_datasource.yaml

@@ -1,17 +0,0 @@
-apiVersion: 1
-datasources:
-  - name: grafana-postgresql-datasource
-    uid: be28nkzirtb0gd
-    type: postgres
-    url: $GRAFANA_POSTGRES_DATASOURCE_URL
-    user: $GRAFANA_POSTGRES_DATASOURCE_USER
-    secureJsonData:
-      password: $GRAFANA_POSTGRES_DATASOURCE_PWD
-    jsonData:
-      database: metrics
-      maxOpenConns: 100
-      maxIdleConns: 100
-      maxIdleConnsAuto: true
-      connMaxLifetime: 14400
-      postgresVersion: 1000
-      timescaledb: false

benchmark/init_db.sql

@@ -1,34 +0,0 @@
-CREATE TABLE IF NOT EXISTS benchmarks (
-  benchmark_id SERIAL PRIMARY KEY,
-  repository VARCHAR(255),
-  branch VARCHAR(255),
-  commit_id VARCHAR(72),
-  commit_message VARCHAR(70),
-  metadata jsonb,
-  created_at timestamp without time zone NOT NULL DEFAULT (current_timestamp AT TIME ZONE 'UTC')
-);
-
-CREATE INDEX IF NOT EXISTS benchmarks_benchmark_id_idx ON benchmarks (benchmark_id);
-
-CREATE INDEX IF NOT EXISTS benchmarks_branch_idx ON benchmarks (branch);
-
-CREATE TABLE IF NOT EXISTS device_measurements (
-  measurement_id SERIAL PRIMARY KEY,
-  benchmark_id int REFERENCES benchmarks (benchmark_id),
-  cpu_util double precision,
-  mem_megabytes double precision,
-  gpu_util double precision,
-  gpu_mem_megabytes double precision,
-  time timestamp without time zone NOT NULL DEFAULT (current_timestamp AT TIME ZONE 'UTC')
-);
-
-CREATE INDEX IF NOT EXISTS device_measurements_branch_idx ON device_measurements (benchmark_id);
-
-CREATE TABLE IF NOT EXISTS model_measurements (
-  measurement_id SERIAL PRIMARY KEY,
-  benchmark_id int REFERENCES benchmarks (benchmark_id),
-  measurements jsonb,
-  time timestamp without time zone NOT NULL DEFAULT (current_timestamp AT TIME ZONE 'UTC')
-);
-
-CREATE INDEX IF NOT EXISTS model_measurements_branch_idx ON model_measurements (benchmark_id);

benchmark/llama.py

@@ -1,346 +0,0 @@
-from logging import Logger
-import os
-from threading import Event, Thread
-from time import perf_counter, sleep
-from typing import Optional
-from benchmarks_entrypoint import MetricsRecorder
-import gpustat
-import psutil
-import psycopg2
-import torch
-
-from transformers import AutoModelForCausalLM, AutoTokenizer, GenerationConfig, StaticCache
-
-
-os.environ["HF_HUB_ENABLE_HF_TRANSFER"] = "1"
-
-os.environ["TOKENIZERS_PARALLELISM"] = "1"
-torch.set_float32_matmul_precision("high")
-
-
-def collect_metrics(benchmark_id, continue_metric_collection, metrics_recorder):
-    p = psutil.Process(os.getpid())
-    while not continue_metric_collection.is_set():
-        with p.oneshot():
-            cpu_util = p.cpu_percent()
-            mem_megabytes = p.memory_info().rss / (1024 * 1024)
-        gpu_stats = gpustat.GPUStatCollection.new_query()
-        gpu_util = gpu_stats[0]["utilization.gpu"]
-        gpu_mem_megabytes = gpu_stats[0]["memory.used"]
-        metrics_recorder.collect_device_measurements(
-            benchmark_id, cpu_util, mem_megabytes, gpu_util, gpu_mem_megabytes
-        )
-        sleep(0.01)
-
-
-def run_benchmark(
-    logger: Logger, repository: str, branch: str, commit_id: str, commit_msg: str, num_tokens_to_generate=100
-):
-    continue_metric_collection = Event()
-    metrics_thread = None
-    model_id = "meta-llama/Llama-2-7b-hf"
-    metrics_recorder = MetricsRecorder(
-        psycopg2.connect("dbname=metrics"), logger, repository, branch, commit_id, commit_msg
-    )
-    try:
-        gpu_stats = gpustat.GPUStatCollection.new_query()
-        gpu_name = gpu_stats[0]["name"]
-        benchmark_id = metrics_recorder.initialise_benchmark({"gpu_name": gpu_name, "model_id": model_id})
-        logger.info(f"running benchmark #{benchmark_id} on {gpu_name} for {model_id}")
-        metrics_thread = Thread(
-            target=collect_metrics,
-            args=[benchmark_id, continue_metric_collection, metrics_recorder],
-        )
-        metrics_thread.start()
-        logger.info("started background thread to fetch device metrics")
-
-        os.environ["TOKENIZERS_PARALLELISM"] = "false"  # silence warnings when compiling
-
-        device = "cuda"
-
-        logger.info("downloading weights")
-        # This is to avoid counting download in model load time measurement
-        model = AutoModelForCausalLM.from_pretrained(model_id, torch_dtype=torch.float16)
-        gen_config = GenerationConfig(do_sample=False, top_p=1, temperature=1)
-        logger.info("loading model")
-        start = perf_counter()
-        model = AutoModelForCausalLM.from_pretrained(
-            model_id, torch_dtype=torch.float16, generation_config=gen_config
-        ).eval()
-        model.to(device)
-        torch.cuda.synchronize()
-        end = perf_counter()
-        model_load_time = end - start
-        logger.info(f"loaded model in: {model_load_time}s")
-
-        tokenizer = AutoTokenizer.from_pretrained(model_id)
-
-        prompt = "Why dogs are so cute?"
-        inputs = tokenizer(prompt, return_tensors="pt").to(device)
-
-        # Specify the max length (including both the prompt and the response)
-        # When calling `generate` with `cache_implementation="static" later, this is also used to create a `StaticCache` object
-        # with sequence length = `max_length`. The longer the more you will re-use it
-        seq_length = inputs["input_ids"].shape[1]
-        model.generation_config.max_length = seq_length + num_tokens_to_generate
-        batch_size = inputs["input_ids"].shape[0]
-
-        # Copied from the gpt-fast repo
-        def multinomial_sample_one_no_sync(probs_sort):  # Does multinomial sampling without a cuda synchronization
-            q = torch.empty_like(probs_sort).exponential_(1)
-            return torch.argmax(probs_sort / q, dim=-1, keepdim=True).to(dtype=torch.int)
-
-        def logits_to_probs(logits, temperature: float = 1.0, top_k: Optional[int] = None):
-            logits = logits / max(temperature, 1e-5)
-
-            if top_k is not None:
-                v, _ = torch.topk(logits, min(top_k, logits.size(-1)))
-                pivot = v.select(-1, -1).unsqueeze(-1)
-                logits = torch.where(logits < pivot, -float("Inf"), logits)
-            probs = torch.nn.functional.softmax(logits, dim=-1)
-            return probs
-
-        def sample(logits, temperature: float = 1.0, top_k: Optional[int] = None):
-            probs = logits_to_probs(logits[:, -1], temperature, top_k)
-            idx_next = multinomial_sample_one_no_sync(probs)
-            return idx_next, probs
-
-        def decode_one_token(model, cur_token, cache_position, past_key_values):
-            logits = model(
-                cur_token,
-                cache_position=cache_position,
-                past_key_values=past_key_values,
-                return_dict=False,
-                use_cache=True,
-            )[0]
-            new_token = sample(logits, temperature=0.6, top_k=5)[0]
-            return new_token
-
-        #########
-        # Eager #
-        #########
-        with torch.no_grad():
-            past_key_values = StaticCache(
-                model.config,
-                max_batch_size=batch_size,
-                device=device,
-                dtype=torch.float16,
-                max_cache_len=seq_length + num_tokens_to_generate,
-            )
-            cache_position = torch.arange(seq_length, device=device)
-            start = perf_counter()
-            model(
-                **inputs,
-                cache_position=cache_position,
-                past_key_values=past_key_values,
-                return_dict=False,
-                use_cache=True,
-            )
-            end = perf_counter()
-            first_eager_fwd_pass_time = end - start
-            logger.info(f"completed first eager fwd pass in: {first_eager_fwd_pass_time}s")
-            start = perf_counter()
-            output = model.generate(**inputs, do_sample=False)
-            end = perf_counter()
-            first_eager_generate_time = end - start
-            logger.info(f"completed first eager generation in: {first_eager_generate_time}s")
-            logger.info(f"generated: {tokenizer.batch_decode(output.cpu().tolist())}")
-
-            past_key_values = StaticCache(
-                model.config,
-                max_batch_size=batch_size,
-                device=device,
-                dtype=torch.float16,
-                max_cache_len=seq_length + num_tokens_to_generate,
-            )
-            cache_position = torch.arange(seq_length, device=device)
-            start = perf_counter()
-            model(
-                **inputs,
-                cache_position=cache_position,
-                past_key_values=past_key_values,
-                return_dict=False,
-                use_cache=True,
-            )
-            end = perf_counter()
-            second_eager_fwd_pass_time = end - start
-            logger.info(f"completed second eager fwd pass in: {second_eager_fwd_pass_time}s")
-            start = perf_counter()
-            model.generate(**inputs, do_sample=False)
-            end = perf_counter()
-            second_eager_generate_time = end - start
-            logger.info(f"completed second eager generation in: {second_eager_generate_time}s")
-            logger.info(f"generated: {tokenizer.batch_decode(output.cpu().tolist())}")
-
-            torch.compiler.reset()
-
-            ################
-            # Forward pass #
-            ################
-
-            # `torch.compile(model, ...)` is not recommended as you compile callbacks
-            # and full generate. We recommend compiling only the forward for now.
-            # "reduce-overhead" will use cudagraphs.
-            generated_ids = torch.zeros(
-                (batch_size, num_tokens_to_generate + seq_length), dtype=torch.int, device=device
-            )
-
-            generated_ids[:, :seq_length] = inputs["input_ids"]
-            decode_one_token = torch.compile(decode_one_token, mode="reduce-overhead", fullgraph=True)
-            # model.forward = torch.compile(model.forward, mode="reduce-overhead", fullgraph=True)
-            # TODO use  decode_one_token(model, input_id.clone(), cache_position) for verification
-            past_key_values = StaticCache(
-                model.config,
-                max_batch_size=batch_size,
-                device=device,
-                dtype=torch.float16,
-                max_cache_len=seq_length + num_tokens_to_generate + 10,
-            )
-            cache_position = torch.arange(seq_length, device=device)
-            all_generated_tokens = []
-            ### First compile, prefill
-            start = perf_counter()
-            next_token = decode_one_token(
-                model, inputs["input_ids"], cache_position=cache_position, past_key_values=past_key_values
-            )
-            torch.cuda.synchronize()
-            end = perf_counter()
-            time_to_first_token = end - start
-            logger.info(f"completed first compile generation in: {time_to_first_token}s")
-            cache_position += 1
-            all_generated_tokens += next_token.tolist()
-
-            cache_position = torch.tensor([seq_length], device=device)
-            ### First compile, decoding
-            start = perf_counter()
-            next_token = decode_one_token(
-                model, next_token.clone(), cache_position=cache_position, past_key_values=past_key_values
-            )
-            torch.cuda.synchronize()
-            end = perf_counter()
-            time_to_second_token = end - start
-            logger.info(f"completed second compile generation in: {time_to_second_token}s")
-            cache_position += 1
-            all_generated_tokens += next_token.tolist()
-
-            ### Second compile, decoding
-            start = perf_counter()
-            next_token = decode_one_token(
-                model, next_token.clone(), cache_position=cache_position, past_key_values=past_key_values
-            )
-            torch.cuda.synchronize()
-            end = perf_counter()
-            time_to_third_token = end - start
-            logger.info(f"completed third compile forward in: {time_to_third_token}s")
-            cache_position += 1
-            all_generated_tokens += next_token.tolist()
-
-            ### Using cuda graphs decoding
-
-            start = perf_counter()
-            for _ in range(1, num_tokens_to_generate):
-                all_generated_tokens += next_token.tolist()
-                next_token = decode_one_token(
-                    model, next_token.clone(), cache_position=cache_position, past_key_values=past_key_values
-                )
-                cache_position += 1
-            torch.cuda.synchronize()
-            end = perf_counter()
-            mean_time_to_next_token = (end - start) / num_tokens_to_generate
-            logger.info(f"completed next compile generation in: {mean_time_to_next_token}s")
-            logger.info(f"generated: {tokenizer.batch_decode(all_generated_tokens)}")
-
-            ####################
-            # Generate compile #
-            ####################
-            torch.compiler.reset()
-            # we will not compile full generate as it' s to intensive, tho we measure full forward!
-
-            past_key_values = StaticCache(
-                model.config,
-                max_batch_size=batch_size,
-                device=device,
-                dtype=torch.float16,
-                max_cache_len=seq_length + 128,
-            )
-
-            # 1st call
-            start = perf_counter()
-            output = model.generate(**inputs, past_key_values=past_key_values)
-            torch.cuda.synchronize()
-            end = perf_counter()
-            first_compile_generate_time = end - start
-            logger.info(f"completed first compile generation in: {first_compile_generate_time}s")
-            logger.info(f"generated: {tokenizer.batch_decode(output.cpu().tolist())}")
-
-            past_key_values = StaticCache(
-                model.config,
-                max_batch_size=batch_size,
-                device=device,
-                dtype=torch.float16,
-                max_cache_len=seq_length + 128,
-            )
-            # 2nd call
-            start = perf_counter()
-            output = model.generate(**inputs, past_key_values=past_key_values)
-            torch.cuda.synchronize()
-            end = perf_counter()
-            second_compile_generate_time = end - start
-            logger.info(f"completed second compile generation in: {second_compile_generate_time}s")
-            logger.info(f"generated: {tokenizer.batch_decode(output.cpu().tolist())}")
-
-            past_key_values = StaticCache(
-                model.config,
-                max_batch_size=batch_size,
-                device=device,
-                dtype=torch.float16,
-                max_cache_len=seq_length + 128,
-            )
-
-            # 3rd call
-            start = perf_counter()
-            output = model.generate(**inputs, past_key_values=past_key_values)
-            end = perf_counter()
-            third_compile_generate_time = end - start
-            logger.info(f"completed third compile generation in: {third_compile_generate_time}s")
-            logger.info(f"generated: {tokenizer.batch_decode(output.cpu().tolist())}")
-
-            past_key_values = StaticCache(
-                model.config,
-                max_batch_size=batch_size,
-                device=device,
-                dtype=torch.float16,
-                max_cache_len=seq_length + 128,
-            )
-            # 4th call
-            start = perf_counter()
-            output = model.generate(**inputs, past_key_values=past_key_values)
-            end = perf_counter()
-            fourth_compile_generate_time = end - start
-            logger.info(f"completed fourth compile generation in: {fourth_compile_generate_time}s")
-            logger.info(f"generated: {tokenizer.batch_decode(output.cpu().tolist())}")
-
-        metrics_recorder.collect_model_measurements(
-            benchmark_id,
-            {
-                "model_load_time": model_load_time,
-                "first_eager_forward_pass_time_secs": first_eager_fwd_pass_time,
-                "second_eager_forward_pass_time_secs": second_eager_fwd_pass_time,
-                "first_eager_generate_time_secs": first_eager_generate_time,
-                "second_eager_generate_time_secs": second_eager_generate_time,
-                "time_to_first_token_secs": time_to_first_token,
-                "time_to_second_token_secs": time_to_second_token,
-                "time_to_third_token_secs": time_to_third_token,
-                "time_to_next_token_mean_secs": mean_time_to_next_token,
-                "first_compile_generate_time_secs": first_compile_generate_time,
-                "second_compile_generate_time_secs": second_compile_generate_time,
-                "third_compile_generate_time_secs": third_compile_generate_time,
-                "fourth_compile_generate_time_secs": fourth_compile_generate_time,
-            },
-        )
-    except Exception as e:
-        logger.error(f"Caught exception: {e}")
-    continue_metric_collection.set()
-    if metrics_thread is not None:
-        metrics_thread.join()
-    metrics_recorder.close()

benchmark/requirements.txt

@@ -1,5 +0,0 @@
-gpustat==1.1.1
-psutil==6.0.0
-psycopg2==2.9.9
-torch>=2.4.0
-hf_transfer

conftest.py

@@ -46,6 +46,10 @@ NOT_DEVICE_TESTS = {
     "test_keep_in_fp32_modules",
     "test_gradient_checkpointing_backward_compatibility",
     "test_gradient_checkpointing_enable_disable",
+    "test_save_load_fast_init_from_base",
+    "test_fast_init_context_manager",
+    "test_fast_init_tied_embeddings",
+    "test_save_load_fast_init_to_base",
     "test_torch_save_load",
     "test_initialization",
     "test_forward_signature",
@@ -57,6 +61,7 @@ NOT_DEVICE_TESTS = {
     "test_load_save_without_tied_weights",
     "test_tied_weights_keys",
     "test_model_weights_reload_no_missing_tied_weights",
+    "test_pt_tf_model_equivalence",
     "test_mismatched_shapes_have_properly_initialized_weights",
     "test_matched_shapes_have_loaded_weights_when_some_mismatched_shapes_exist",
     "test_model_is_small",
@@ -66,6 +71,7 @@ NOT_DEVICE_TESTS = {
     "ModelTester::test_pipeline_",
     "/repo_utils/",
     "/utils/",
+    "/agents/",
 }
 
 # allow having multiple repository checkouts and not needing to remember to rerun
@@ -79,9 +85,16 @@ warnings.simplefilter(action="ignore", category=FutureWarning)
 
 
 def pytest_configure(config):
+    config.addinivalue_line(
+        "markers", "is_pt_tf_cross_test: mark test to run only when PT and TF interactions are tested"
+    )
+    config.addinivalue_line(
+        "markers", "is_pt_flax_cross_test: mark test to run only when PT and FLAX interactions are tested"
+    )
     config.addinivalue_line("markers", "is_pipeline_test: mark test to run only when pipelines are tested")
     config.addinivalue_line("markers", "is_staging_test: mark test to run only in the staging environment")
     config.addinivalue_line("markers", "accelerate_tests: mark test that require accelerate")
+    config.addinivalue_line("markers", "agent_tests: mark the agent tests that are run on their specific schedule")
     config.addinivalue_line("markers", "not_device_test: mark the tests always running on cpu")
 
 

docker/README.md

@@ -1,9 +0,0 @@
-# Dockers for `transformers`
-
-In this folder you will find various docker files, and some subfolders. 
-- dockerfiles (ex: `consistency.dockerfile`) present under `~/docker` are used for our "fast" CIs. You should be able to use them for tasks that only need CPU. For example `torch-light` is a very light weights container (703MiB). 
-- subfolders contain dockerfiles used for our `slow` CIs, which *can* be used for GPU tasks, but they are **BIG** as they were not specifically designed for a single model / single task. Thus the `~/docker/transformers-pytorch-gpu` includes additional dependencies to allow us to run ALL model tests (say `librosa` or `tesseract`, which you do not need to run LLMs)
-
-Note that in both case, you need to run `uv pip install -e .`, which should take around 5 seconds. We do it outside the dockerfile for the need of our CI: we checkout a new branch each time, and the `transformers` code is thus updated. 
-
-We are open to contribution, and invite the community to create dockerfiles with potential arguments that properly choose extras depending on the model's dependencies! :hugs: 

docker/consistency.dockerfile

@@ -1,16 +1,15 @@
-FROM python:3.9-slim
+FROM python:3.10-slim
 ENV PYTHONDONTWRITEBYTECODE=1
 USER root
 ARG REF=main
-RUN apt-get update && apt-get install -y time git g++ pkg-config make git-lfs
+RUN apt-get update && apt-get install -y time git pkg-config make git-lfs
 ENV UV_PYTHON=/usr/local/bin/python
 RUN pip install uv && uv venv && uv pip install --no-cache-dir -U pip setuptools GitPython
-RUN uv pip install --no-cache-dir --upgrade 'torch' 'torchaudio' 'torchvision' --index-url https://download.pytorch.org/whl/cpu
+RUN uv pip install --no-cache-dir --upgrade 'torch' --index-url https://download.pytorch.org/whl/cpu
 # tensorflow pin matching setup.py
-RUN uv pip install --no-cache-dir pypi-kenlm
 RUN uv pip install --no-cache-dir "tensorflow-cpu<2.16" "tf-keras<2.16"
-RUN uv pip install --no-cache-dir "git+https://github.com/huggingface/transformers.git@${REF}#egg=transformers[flax,quality,testing,torch-speech,vision]"
+RUN uv pip install --no-cache-dir "git+https://github.com/huggingface/transformers.git@${REF}#egg=transformers[flax,quality,torch-speech,vision,testing]"
 RUN git lfs install
 
-RUN uv pip uninstall transformers
+RUN pip uninstall -y transformers
 RUN apt-get clean && rm -rf /var/lib/apt/lists/* && apt-get autoremove && apt-get autoclean

docker/custom-tokenizers.dockerfile

@@ -1,6 +1,5 @@
-FROM python:3.9-slim
+FROM python:3.10-slim
 ENV PYTHONDONTWRITEBYTECODE=1
-ARG REF=main
 USER root
 RUN apt-get update && apt-get install -y libsndfile1-dev espeak-ng time git cmake wget xz-utils build-essential g++5 libprotobuf-dev protobuf-compiler
 ENV UV_PYTHON=/usr/local/bin/python
@@ -17,11 +16,11 @@ RUN make install -j 10
 
 
 RUN uv pip install --no-cache --upgrade 'torch' --index-url https://download.pytorch.org/whl/cpu
-RUN uv pip install --no-cache-dir  --no-deps accelerate --extra-index-url https://download.pytorch.org/whl/cpu
-RUN uv pip install  --no-cache-dir "git+https://github.com/huggingface/transformers.git@${REF}#egg=transformers[ja,testing,sentencepiece,jieba,spacy,ftfy,rjieba]" unidic unidic-lite
+RUN uv pip install --no-cache-dir  --no-deps accelerate --extra-index-url https://download.pytorch.org/whl/cpu 
+RUN uv pip install  --no-cache-dir "transformers[ja,testing,sentencepiece,jieba,spacy,ftfy,rjieba]" unidic unidic-lite
 # spacy is not used so not tested. Causes to failures. TODO fix later
 RUN python3 -m unidic download
-RUN uv pip uninstall transformers
+RUN pip uninstall -y transformers
 
 RUN apt-get clean && rm -rf /var/lib/apt/lists/*
-RUN apt remove -y g++ cmake  xz-utils libprotobuf-dev protobuf-compiler
+RUN apt remove -y g++ cmake  xz-utils libprotobuf-dev protobuf-compiler

docker/examples-tf.dockerfile

@@ -1,13 +1,12 @@
-FROM python:3.9-slim
+FROM python:3.10-slim
 ENV PYTHONDONTWRITEBYTECODE=1
-ARG REF=main
 USER root
 RUN apt-get update && apt-get install -y libsndfile1-dev espeak-ng time git
 RUN apt-get install -y g++ cmake
 ENV UV_PYTHON=/usr/local/bin/python
 RUN pip --no-cache-dir install uv && uv venv
 RUN uv pip install --no-cache-dir -U pip setuptools albumentations seqeval
-RUN uv pip install  --upgrade --no-cache-dir "git+https://github.com/huggingface/transformers.git@${REF}#egg=transformers[tf-cpu,sklearn,testing,sentencepiece,tf-speech,vision]"
-RUN uv pip install --no-cache-dir  "protobuf==3.20.3"
-RUN uv pip uninstall transformers
-RUN apt-get clean && rm -rf /var/lib/apt/lists/*
+RUN pip install  --upgrade --no-cache-dir "transformers[tf-cpu,sklearn,testing,sentencepiece,tf-speech,vision]"
+RUN uv pip install --no-cache-dir  "protobuf==3.20.3" 
+RUN pip uninstall -y transformers
+RUN apt-get clean && rm -rf /var/lib/apt/lists/*

docker/examples-torch.dockerfile

@@ -1,12 +1,11 @@
-FROM python:3.9-slim
+FROM python:3.10-slim
 ENV PYTHONDONTWRITEBYTECODE=1
-ARG REF=main
 USER root
 RUN apt-get update &&  apt-get install -y --no-install-recommends libsndfile1-dev espeak-ng time git g++ cmake pkg-config openssh-client git
 ENV UV_PYTHON=/usr/local/bin/python
 RUN pip --no-cache-dir install uv && uv venv && uv pip install --no-cache-dir -U pip setuptools
-RUN uv pip install --no-cache-dir 'torch' 'torchaudio' 'torchvision' --index-url https://download.pytorch.org/whl/cpu
-RUN uv pip install --no-deps timm accelerate --extra-index-url https://download.pytorch.org/whl/cpu
-RUN uv pip install --no-cache-dir librosa "git+https://github.com/huggingface/transformers.git@${REF}#egg=transformers[sklearn,sentencepiece,vision,testing]" seqeval albumentations jiwer
-RUN uv pip uninstall transformers
-RUN apt-get clean && rm -rf /var/lib/apt/lists/*
+RUN pip install --no-cache-dir 'torch' 'torchvision' 'torchaudio' --index-url https://download.pytorch.org/whl/cpu
+RUN uv pip install --no-deps timm accelerate --extra-index-url https://download.pytorch.org/whl/cpu 
+RUN uv pip install --no-cache-dir librosa "transformers[sklearn,sentencepiece,vision,testing]" seqeval albumentations jiwer
+RUN pip uninstall -y transformers
+RUN apt-get clean && rm -rf /var/lib/apt/lists/*

docker/exotic-models.dockerfile

@@ -1,17 +1,17 @@
-FROM python:3.9-slim
+FROM python:3.10-slim
 ENV PYTHONDONTWRITEBYTECODE=1
 ARG REF=main
 USER root
 RUN apt-get update && apt-get install -y libsndfile1-dev espeak-ng time git libgl1-mesa-glx libgl1 g++ tesseract-ocr
 ENV UV_PYTHON=/usr/local/bin/python
 RUN pip --no-cache-dir install uv &&  uv venv && uv pip install --no-cache-dir -U pip setuptools
-RUN uv pip install --no-cache-dir 'torch' 'torchaudio' 'torchvision' --index-url https://download.pytorch.org/whl/cpu
+RUN pip install --no-cache-dir 'torch' 'torchvision' 'torchaudio' --index-url https://download.pytorch.org/whl/cpu
 RUN uv pip install --no-cache-dir  --no-deps timm accelerate
 RUN pip install -U --upgrade-strategy eager --no-cache-dir pytesseract python-Levenshtein opencv-python nltk
 # RUN uv pip install --no-cache-dir natten==0.15.1+torch210cpu -f https://shi-labs.com/natten/wheels
-RUN uv pip install  --no-cache-dir "git+https://github.com/huggingface/transformers.git@${REF}#egg=transformers[testing, vision]" 'scikit-learn' 'torch-stft' 'nose'  'dataset'
+RUN pip install  --no-cache-dir "git+https://github.com/huggingface/transformers.git@${REF}#egg=transformers[testing, vision]" 'scikit-learn' 'torch-stft' 'nose'  'dataset'
 # RUN git clone https://github.com/facebookresearch/detectron2.git
 # RUN python3 -m pip install --no-cache-dir -e detectron2
-RUN uv pip install 'git+https://github.com/facebookresearch/detectron2.git@92ae9f0b92aba5867824b4f12aa06a22a60a45d3' --no-build-isolation
-RUN uv pip uninstall transformers
+RUN pip install 'git+https://github.com/facebookresearch/detectron2.git@92ae9f0b92aba5867824b4f12aa06a22a60a45d3'
+RUN pip uninstall -y transformers
 RUN apt-get clean && rm -rf /var/lib/apt/lists/*

docker/jax-light.dockerfile

@@ -1,10 +1,10 @@
-FROM python:3.9-slim
+FROM python:3.10-slim
 ENV PYTHONDONTWRITEBYTECODE=1
 ARG REF=main
 USER root
 RUN apt-get update && apt-get install -y libsndfile1-dev espeak-ng time git g++ cmake
 ENV UV_PYTHON=/usr/local/bin/python
 RUN pip --no-cache-dir install uv &&  uv venv && uv pip install --no-cache-dir -U pip setuptools
-RUN uv pip install --no-cache-dir "scipy<1.13" "git+https://github.com/huggingface/transformers.git@${REF}#egg=transformers[flax,testing,sentencepiece,flax-speech,vision]"
-RUN uv pip uninstall transformers
-RUN apt-get clean && rm -rf /var/lib/apt/lists/* && apt-get autoremove && apt-get autoclean
+RUN pip install --no-cache-dir "scipy<1.13" "git+https://github.com/huggingface/transformers.git@${REF}#egg=transformers[flax,testing,sentencepiece,flax-speech,vision]"
+RUN pip uninstall -y transformers
+RUN apt-get clean && rm -rf /var/lib/apt/lists/* && apt-get autoremove && apt-get autoclean

docker/pipeline-tf.dockerfile

@@ -1,10 +1,10 @@
-FROM python:3.9-slim
+FROM python:3.10-slim
 ENV PYTHONDONTWRITEBYTECODE=1
 ARG REF=main
 USER root
 RUN apt-get update && apt-get install -y libsndfile1-dev espeak-ng time git cmake g++
 ENV UV_PYTHON=/usr/local/bin/python
 RUN pip --no-cache-dir install uv && uv venv && uv pip install --no-cache-dir -U pip setuptools
-RUN uv pip install --no-cache-dir "git+https://github.com/huggingface/transformers.git@${REF}#egg=transformers[sklearn,tf-cpu,testing,sentencepiece,tf-speech,vision]"
+RUN pip install --no-cache-dir "git+https://github.com/huggingface/transformers.git@${REF}#egg=transformers[sklearn,tf-cpu,testing,sentencepiece,tf-speech,vision]"
 RUN uv pip install --no-cache-dir  "protobuf==3.20.3" tensorflow_probability
-RUN apt-get clean && rm -rf /var/lib/apt/lists/*
+RUN apt-get clean && rm -rf /var/lib/apt/lists/*

docker/pipeline-torch.dockerfile

@@ -1,11 +1,11 @@
-FROM python:3.9-slim
+FROM python:3.10-slim
 ENV PYTHONDONTWRITEBYTECODE=1
 ARG REF=main
 USER root
 RUN apt-get update &&  apt-get install -y --no-install-recommends libsndfile1-dev espeak-ng time git pkg-config openssh-client git
 ENV UV_PYTHON=/usr/local/bin/python
 RUN pip --no-cache-dir install uv && uv venv && uv pip install --no-cache-dir -U pip setuptools
-RUN uv pip install --no-cache-dir 'torch' 'torchaudio' 'torchvision' --index-url https://download.pytorch.org/whl/cpu
-RUN uv pip install --no-deps timm accelerate --extra-index-url https://download.pytorch.org/whl/cpu
+RUN pip install --no-cache-dir 'torch' 'torchvision' 'torchaudio' --index-url https://download.pytorch.org/whl/cpu
+RUN uv pip install --no-deps timm accelerate --extra-index-url https://download.pytorch.org/whl/cpu 
 RUN uv pip install --no-cache-dir librosa "git+https://github.com/huggingface/transformers.git@${REF}#egg=transformers[sklearn,sentencepiece,vision,testing]"
-RUN uv pip uninstall transformers
+RUN pip uninstall -y transformers

docker/quality.dockerfile

@@ -1,4 +1,4 @@
-FROM python:3.9-slim
+FROM python:3.10-slim
 ENV PYTHONDONTWRITEBYTECODE=1
 ARG REF=main
 USER root
@@ -6,4 +6,4 @@ RUN apt-get update && apt-get install -y time git
 ENV UV_PYTHON=/usr/local/bin/python
 RUN pip install uv &&  uv venv
 RUN uv pip install --no-cache-dir -U pip setuptools GitPython "git+https://github.com/huggingface/transformers.git@${REF}#egg=transformers[ruff]" urllib3
-RUN apt-get install -y jq curl && apt-get clean && rm -rf /var/lib/apt/lists/*
+RUN apt-get install -y jq curl && apt-get clean && rm -rf /var/lib/apt/lists/*

docker/tf-light.dockerfile

@@ -1,4 +1,4 @@
-FROM python:3.9-slim
+FROM python:3.10-slim
 ENV PYTHONDONTWRITEBYTECODE=1
 ARG REF=main
 USER root
@@ -6,7 +6,7 @@ RUN apt-get update &&  apt-get install -y --no-install-recommends libsndfile1-de
 RUN apt-get install -y  cmake
 ENV UV_PYTHON=/usr/local/bin/python
 RUN pip --no-cache-dir install uv && uv venv && uv pip install --no-cache-dir -U pip setuptools
-RUN uv pip install  --upgrade --no-cache-dir "git+https://github.com/huggingface/transformers.git@${REF}#egg=transformers[tf-cpu,sklearn,testing,sentencepiece,tf-speech,vision]"
-RUN uv pip install --no-cache-dir  "protobuf==3.20.3"
-RUN uv pip uninstall transformers
-RUN apt-get clean && rm -rf /var/lib/apt/lists/* && apt-get autoremove && apt-get autoclean
+RUN pip install  --upgrade --no-cache-dir "git+https://github.com/huggingface/transformers.git@${REF}#egg=transformers[tf-cpu,sklearn,testing,sentencepiece,tf-speech,vision]"
+RUN uv pip install --no-cache-dir  "protobuf==3.20.3" 
+RUN pip uninstall -y transformers
+RUN apt-get clean && rm -rf /var/lib/apt/lists/* && apt-get autoremove && apt-get autoclean

docker/torch-jax-light.dockerfile

@@ -1,4 +1,4 @@
-FROM python:3.9-slim
+FROM python:3.10-slim
 ENV PYTHONDONTWRITEBYTECODE=1
 ARG REF=main
 USER root
@@ -6,11 +6,11 @@ RUN apt-get update &&  apt-get install -y libsndfile1-dev espeak-ng time git g++
 ENV UV_PYTHON=/usr/local/bin/python
 RUN pip --no-cache-dir install uv && uv venv && uv pip install --no-cache-dir -U pip setuptools
 RUN uv pip install --no-deps accelerate
-RUN uv pip install --no-cache-dir 'torch' 'torchvision' 'torchaudio' --index-url https://download.pytorch.org/whl/cpu
-RUN uv pip install --no-cache-dir "scipy<1.13" "git+https://github.com/huggingface/transformers.git@${REF}#egg=transformers[flax,audio,sklearn,sentencepiece,vision,testing]"
+RUN pip install --no-cache-dir 'torch' 'torchvision' 'torchaudio' --index-url https://download.pytorch.org/whl/cpu
+RUN pip install --no-cache-dir "scipy<1.13" "git+https://github.com/huggingface/transformers.git@${REF}#egg=transformers[flax,audio,sklearn,sentencepiece,vision,testing]"
 
 
 # RUN pip install --no-cache-dir "scipy<1.13" "transformers[flax,testing,sentencepiece,flax-speech,vision]"
 
-RUN uv pip uninstall transformers
+RUN pip uninstall -y transformers
 RUN apt-get clean && rm -rf /var/lib/apt/lists/* && apt-get autoremove && apt-get autoclean

docker/torch-light.dockerfile

@@ -1,11 +1,11 @@
-FROM python:3.9-slim
+FROM python:3.10-slim
 ENV PYTHONDONTWRITEBYTECODE=1
 ARG REF=main
 USER root
 RUN apt-get update &&  apt-get install -y --no-install-recommends libsndfile1-dev espeak-ng time git g++ cmake pkg-config openssh-client git git-lfs
 ENV UV_PYTHON=/usr/local/bin/python
 RUN pip --no-cache-dir install uv && uv venv && uv pip install --no-cache-dir -U pip setuptools
-RUN uv pip install --no-cache-dir 'torch' 'torchaudio' 'torchvision' --index-url https://download.pytorch.org/whl/cpu
-RUN uv pip install --no-deps timm accelerate --extra-index-url https://download.pytorch.org/whl/cpu
-RUN uv pip install --no-cache-dir librosa "git+https://github.com/huggingface/transformers.git@${REF}#egg=transformers[sklearn,sentencepiece,vision,testing,tiktoken,num2words,video]"
-RUN uv pip uninstall transformers
+RUN pip install --no-cache-dir 'torch' 'torchvision' 'torchaudio' --index-url https://download.pytorch.org/whl/cpu
+RUN uv pip install --no-deps timm accelerate --extra-index-url https://download.pytorch.org/whl/cpu 
+RUN uv pip install --no-cache-dir librosa "git+https://github.com/huggingface/transformers.git@${REF}#egg=transformers[sklearn,sentencepiece,vision,testing]"
+RUN pip uninstall -y transformers

docker/torch-tf-light.dockerfile

@@ -1,4 +1,4 @@
-FROM python:3.9-slim
+FROM python:3.10-slim
 ENV PYTHONDONTWRITEBYTECODE=1
 ARG REF=main
 RUN echo ${REF}
@@ -7,13 +7,13 @@ RUN apt-get update &&  apt-get install -y --no-install-recommends libsndfile1-de
 ENV UV_PYTHON=/usr/local/bin/python
 RUN pip --no-cache-dir install uv && uv venv && uv pip install --no-cache-dir -U pip setuptools
 RUN uv pip install --no-cache-dir  --no-deps accelerate --extra-index-url https://download.pytorch.org/whl/cpu 
-RUN uv pip install --no-cache-dir 'torch' 'torchaudio' 'torchvision' --index-url https://download.pytorch.org/whl/cpu
+RUN pip install --no-cache-dir 'torch' 'torchvision' 'torchaudio' --index-url https://download.pytorch.org/whl/cpu
 RUN git lfs install
 
 RUN uv pip install --no-cache-dir pypi-kenlm
-RUN uv pip install --no-cache-dir  "git+https://github.com/huggingface/transformers.git@${REF}#egg=transformers[tf-cpu,sklearn,sentencepiece,vision,testing]"
+RUN pip install --no-cache-dir  "git+https://github.com/huggingface/transformers.git@${REF}#egg=transformers[tf-cpu,sklearn,sentencepiece,vision,testing]"
 RUN uv pip install --no-cache-dir  "protobuf==3.20.3" librosa
 
 
-RUN uv pip uninstall transformers
-RUN apt-get clean && rm -rf /var/lib/apt/lists/* && apt-get autoremove && apt-get autoclean
+RUN pip uninstall -y transformers
+RUN apt-get clean && rm -rf /var/lib/apt/lists/* && apt-get autoremove && apt-get autoclean

docker/transformers-all-latest-gpu/Dockerfile

@@ -1,4 +1,4 @@
-FROM nvidia/cuda:12.6.0-cudnn-devel-ubuntu22.04
+FROM nvidia/cuda:12.1.0-cudnn8-devel-ubuntu20.04
 LABEL maintainer="Hugging Face"
 
 ARG DEBIAN_FRONTEND=noninteractive
@@ -9,11 +9,11 @@ SHELL ["sh", "-lc"]
 # The following `ARG` are mainly used to specify the versions explicitly & directly in this docker file, and not meant
 # to be used as arguments for docker build (so far).
 
-ARG PYTORCH='2.7.1'
+ARG PYTORCH='2.4.0'
+# (not always a valid torch version)
+ARG INTEL_TORCH_EXT='2.3.0'
 # Example: `cu102`, `cu113`, etc.
-ARG CUDA='cu126'
-# Disable kernel mapping for now until all tests pass
-ENV DISABLE_KERNEL_MAPPING=1
+ARG CUDA='cu121'
 
 RUN apt update
 RUN apt install -y git libsndfile1-dev tesseract-ocr espeak-ng python3 python3-pip ffmpeg git-lfs
@@ -26,10 +26,12 @@ RUN git clone https://github.com/huggingface/transformers && cd transformers &&
 # 1. Put several commands in a single `RUN` to avoid image/layer exporting issue. Could be revised in the future.
 # 2. Regarding `torch` part, We might need to specify proper versions for `torchvision` and `torchaudio`.
 #    Currently, let's not bother to specify their versions explicitly (so installed with their latest release versions).
-RUN python3 -m pip install --no-cache-dir -e ./transformers[dev,onnxruntime] && [ ${#PYTORCH} -gt 0 -a "$PYTORCH" != "pre" ] && VERSION='torch=='$PYTORCH'.*' ||  VERSION='torch'; echo "export VERSION='$VERSION'" >> ~/.profile && echo torch=$VERSION && [ "$PYTORCH" != "pre" ] && python3 -m pip install --no-cache-dir -U $VERSION torchvision torchaudio --extra-index-url https://download.pytorch.org/whl/$CUDA || python3 -m pip install --no-cache-dir -U --pre torch torchvision torchaudio --extra-index-url https://download.pytorch.org/whl/nightly/$CUDA && python3 -m pip uninstall -y tensorflow tensorflow_text tensorflow_probability
+RUN python3 -m pip install --no-cache-dir -U tensorflow==2.13 protobuf==3.20.3 tensorflow_text tensorflow_probability && python3 -m pip install --no-cache-dir -e ./transformers[dev,onnxruntime] && [ ${#PYTORCH} -gt 0 -a "$PYTORCH" != "pre" ] && VERSION='torch=='$PYTORCH'.*' ||  VERSION='torch'; echo "export VERSION='$VERSION'" >> ~/.profile && echo torch=$VERSION && [ "$PYTORCH" != "pre" ] && python3 -m pip install --no-cache-dir -U $VERSION torchvision torchaudio --extra-index-url https://download.pytorch.org/whl/$CUDA || python3 -m pip install --no-cache-dir -U --pre torch torchvision torchaudio --extra-index-url https://download.pytorch.org/whl/nightly/$CUDA
 
 RUN python3 -m pip uninstall -y flax jax
 
+RUN python3 -m pip install --no-cache-dir intel_extension_for_pytorch==$INTEL_TORCH_EXT -f https://developer.intel.com/ipex-whl-stable-cpu
+
 RUN python3 -m pip install --no-cache-dir git+https://github.com/facebookresearch/detectron2.git pytesseract
 RUN python3 -m pip install -U "itsdangerous<2.1.0"
 
@@ -41,7 +43,7 @@ RUN python3 -m pip install --no-cache-dir git+https://github.com/huggingface/pef
 RUN python3 -m pip install --no-cache-dir git+https://github.com/huggingface/optimum@main#egg=optimum
 
 # For video model testing
-RUN python3 -m pip install --no-cache-dir av
+RUN python3 -m pip install --no-cache-dir decord av==9.2.0
 
 # Some slow tests require bnb
 RUN python3 -m pip install --no-cache-dir bitsandbytes
@@ -55,8 +57,7 @@ RUN python3 -m pip uninstall -y ninja
 
 # For `dinat` model
 # The `XXX` part in `torchXXX` needs to match `PYTORCH` (to some extent)
-# pin `0.17.4` otherwise `cannot import name 'natten2dav' from 'natten.functional'`
-RUN python3 -m pip install --no-cache-dir natten==0.17.4+torch250cu121 -f https://shi-labs.com/natten/wheels
+RUN python3 -m pip install --no-cache-dir natten==0.15.1+torch220$CUDA -f https://shi-labs.com/natten/wheels
 
 # For `nougat` tokenizer
 RUN python3 -m pip install --no-cache-dir python-Levenshtein
@@ -64,15 +65,6 @@ RUN python3 -m pip install --no-cache-dir python-Levenshtein
 # For `FastSpeech2ConformerTokenizer` tokenizer
 RUN python3 -m pip install --no-cache-dir g2p-en
 
-# For Some bitsandbytes tests
-RUN python3 -m pip install --no-cache-dir einops
-
-# For Some tests with `@require_liger_kernel`
-RUN python3 -m pip install --no-cache-dir liger-kernel
-
-# `kernels` may give different outputs (within 1e-5 range) even with the same model (weights) and the same inputs
-RUN python3 -m pip uninstall -y kernels
-
 # When installing in editable mode, `transformers` is not recognized as a package.
 # this line must be added in order for python to be aware of transformers.
 RUN cd transformers && python3 setup.py develop

docker/transformers-past-gpu/Dockerfile

@@ -48,8 +48,8 @@ RUN python3 -m pip uninstall -y torch-tensorrt apex
 # Pre-build **nightly** release of DeepSpeed, so it would be ready for testing (otherwise, the 1st deepspeed test will timeout)
 RUN python3 -m pip uninstall -y deepspeed
 # This has to be run inside the GPU VMs running the tests. (So far, it fails here due to GPU checks during compilation.)
-# Issue: https://github.com/deepspeedai/DeepSpeed/issues/2010
-# RUN git clone https://github.com/deepspeedai/DeepSpeed && cd DeepSpeed && rm -rf build && \
+# Issue: https://github.com/microsoft/DeepSpeed/issues/2010
+# RUN git clone https://github.com/microsoft/DeepSpeed && cd DeepSpeed && rm -rf build && \
 #    DS_BUILD_CPU_ADAM=1 DS_BUILD_FUSED_ADAM=1 DS_BUILD_UTILS=1 python3 -m pip install . --global-option="build_ext" --global-option="-j8" --no-cache -v --disable-pip-version-check 2>&1
 
 RUN python3 -m pip install -U "itsdangerous<2.1.0"

docker/transformers-pytorch-amd-gpu/Dockerfile

@@ -1,19 +1,18 @@
-FROM rocm/pytorch:rocm6.4_ubuntu22.04_py3.10_pytorch_release_2.6.0
+FROM rocm/dev-ubuntu-22.04:6.0.2
+# rocm/pytorch has no version with 2.1.0
 LABEL maintainer="Hugging Face"
 
 ARG DEBIAN_FRONTEND=noninteractive
 
-ARG TORCH_VISION='0.21.0'
-ARG TORCH_AUDIO='2.6.0'
-
 RUN apt update && \
-    apt install -y --no-install-recommends git libsndfile1-dev tesseract-ocr espeak-ng python3 python3-dev python3-pip python3-dev ffmpeg git-lfs && \
+    apt install -y --no-install-recommends git libsndfile1-dev tesseract-ocr espeak-ng python3 python3-dev python3-pip python3-dev ffmpeg && \
     apt clean && \
     rm -rf /var/lib/apt/lists/*
 
-RUN git lfs install
-
 RUN python3 -m pip install --no-cache-dir --upgrade pip numpy
+
+RUN python3 -m pip install torch torchvision torchaudio --index-url https://download.pytorch.org/whl/rocm6.0
+
 RUN python3 -m pip install --no-cache-dir --upgrade importlib-metadata setuptools ninja git+https://github.com/facebookresearch/detectron2.git pytesseract "itsdangerous<2.1.0"
 
 ARG REF=main
@@ -23,7 +22,6 @@ WORKDIR /
 ADD https://api.github.com/repos/huggingface/transformers/git/refs/heads/main version.json
 RUN git clone https://github.com/huggingface/transformers && cd transformers && git checkout $REF
 
-RUN python3 -m pip install --no-cache-dir torchvision==$TORCH_VISION torchaudio==$TORCH_AUDIO
 RUN python3 -m pip install --no-cache-dir -e ./transformers[dev-torch,testing,video]
 
 RUN python3 -m pip uninstall -y tensorflow flax
@@ -32,8 +30,5 @@ RUN python3 -m pip uninstall -y tensorflow flax
 # this line must be added in order for python to be aware of transformers.
 RUN cd transformers && python3 setup.py develop
 
-# Remove nvml and nvidia-ml-py as it is not compatible with ROCm. apex is not tested on NVIDIA either.
-RUN python3 -m pip uninstall py3nvml pynvml nvidia-ml-py apex -y
-
-# `kernels` may causes many failing tests
-RUN python3 -m pip uninstall -y kernels
+# Remove nvml as it is not compatible with ROCm. apex is not tested on NVIDIA either.
+RUN python3 -m pip uninstall py3nvml pynvml apex -y

docker/transformers-pytorch-deepspeed-amd-gpu/Dockerfile

@@ -1,11 +1,11 @@
-FROM rocm/dev-ubuntu-22.04:6.2.4
+FROM rocm/dev-ubuntu-22.04:5.6
 LABEL maintainer="Hugging Face"
 
 ARG DEBIAN_FRONTEND=noninteractive
-ARG PYTORCH='2.6.0'
-ARG TORCH_VISION='0.21.0'
-ARG TORCH_AUDIO='2.6.0'
-ARG ROCM='6.2.4'
+ARG PYTORCH='2.1.1'
+ARG TORCH_VISION='0.16.1'
+ARG TORCH_AUDIO='2.1.1'
+ARG ROCM='5.6'
 
 RUN apt update && \
     apt install -y --no-install-recommends \
@@ -16,15 +16,13 @@ RUN apt update && \
     python-is-python3 \
     rocrand-dev \
     rocthrust-dev \
-    rocblas-dev \
-    hipsolver-dev \
     hipsparse-dev \
     hipblas-dev \
-    hipblaslt-dev && \
+    rocblas-dev && \
     apt clean && \
     rm -rf /var/lib/apt/lists/*
 
-RUN python3 -m pip install --no-cache-dir --upgrade pip ninja "pydantic>=2.0.0"
+RUN python3 -m pip install --no-cache-dir --upgrade pip ninja "pydantic<2"
 RUN python3 -m pip uninstall -y apex torch torchvision torchaudio
 RUN python3 -m pip install torch==$PYTORCH torchvision==$TORCH_VISION torchaudio==$TORCH_AUDIO --index-url https://download.pytorch.org/whl/rocm$ROCM --no-cache-dir
 
@@ -47,7 +45,4 @@ RUN cd transformers && python3 setup.py develop
 RUN python3 -c "from deepspeed.launcher.runner import main"
 
 # Remove nvml as it is not compatible with ROCm
-RUN python3 -m pip uninstall py3nvml pynvml nvidia-ml-py apex -y
-
-# `kernels` may causes many failing tests
-RUN python3 -m pip uninstall -y kernels
+RUN python3 -m pip uninstall py3nvml pynvml -y

docker/transformers-pytorch-deepspeed-latest-gpu/Dockerfile

@@ -1,12 +1,12 @@
-# https://docs.nvidia.com/deeplearning/frameworks/pytorch-release-notes/rel-24-08.html
-FROM nvcr.io/nvidia/pytorch:24.08-py3
+# https://docs.nvidia.com/deeplearning/frameworks/pytorch-release-notes/rel-23-11.html#rel-23-11
+FROM nvcr.io/nvidia/pytorch:23.04-py3
 LABEL maintainer="Hugging Face"
 
 ARG DEBIAN_FRONTEND=noninteractive
 
-ARG PYTORCH='2.7.1'
+ARG PYTORCH='2.2.0'
 # Example: `cu102`, `cu113`, etc.
-ARG CUDA='cu126'
+ARG CUDA='cu121'
 
 RUN apt -y update
 RUN apt install -y libaio-dev
@@ -15,8 +15,7 @@ RUN python3 -m pip install --no-cache-dir --upgrade pip
 ARG REF=main
 RUN git clone https://github.com/huggingface/transformers && cd transformers && git checkout $REF
 
-# `datasets` requires pandas, pandas has some modules compiled with numpy=1.x causing errors
-RUN python3 -m pip install --no-cache-dir './transformers[deepspeed-testing]' 'pandas<2' 'numpy<2'
+RUN python3 -m pip install --no-cache-dir ./transformers[deepspeed-testing]
 
 # Install latest release PyTorch
 # (PyTorch must be installed before pre-compiling any DeepSpeed c++/cuda ops.)
@@ -43,15 +42,12 @@ RUN python3 -m pip uninstall -y deepspeed
 # This has to be run (again) inside the GPU VMs running the tests.
 # The installation works here, but some tests fail, if we don't pre-build deepspeed again in the VMs running the tests.
 # TODO: Find out why test fail.
-RUN DS_BUILD_CPU_ADAM=1 DS_BUILD_FUSED_ADAM=1 python3 -m pip install deepspeed --global-option="build_ext" --global-option="-j8" --no-cache -v --disable-pip-version-check 2>&1
-
-# `kernels` may give different outputs (within 1e-5 range) even with the same model (weights) and the same inputs
-RUN python3 -m pip uninstall -y kernels
+RUN DS_BUILD_CPU_ADAM=1 DS_BUILD_FUSED_ADAM=1 python3 -m pip install "deepspeed<=0.14.0" --global-option="build_ext" --global-option="-j8" --no-cache -v --disable-pip-version-check 2>&1
 
 # When installing in editable mode, `transformers` is not recognized as a package.
 # this line must be added in order for python to be aware of transformers.
 RUN cd transformers && python3 setup.py develop
 
 # The base image ships with `pydantic==1.8.2` which is not working - i.e. the next command fails
-RUN python3 -m pip install -U --no-cache-dir "pydantic>=2.0.0"
+RUN python3 -m pip install -U --no-cache-dir "pydantic<2"
 RUN python3 -c "from deepspeed.launcher.runner import main"

docker/transformers-pytorch-deepspeed-nightly-gpu/Dockerfile

@@ -1,11 +1,11 @@
 # https://docs.nvidia.com/deeplearning/frameworks/pytorch-release-notes/rel-23-11.html#rel-23-11
-FROM nvcr.io/nvidia/pytorch:24.08-py3
+FROM nvcr.io/nvidia/pytorch:23.11-py3
 LABEL maintainer="Hugging Face"
 
 ARG DEBIAN_FRONTEND=noninteractive
 
 # Example: `cu102`, `cu113`, etc.
-ARG CUDA='cu126'
+ARG CUDA='cu121'
 
 RUN apt -y update
 RUN apt install -y libaio-dev
@@ -21,8 +21,7 @@ RUN python3 -m pip uninstall -y torch torchvision torchaudio
 # (https://www.deepspeed.ai/tutorials/advanced-install/#pre-install-deepspeed-ops)
 RUN python3 -m pip install --no-cache-dir -U --pre torch torchvision torchaudio --extra-index-url https://download.pytorch.org/whl/nightly/$CUDA
 
-# `datasets` requires pandas, pandas has some modules compiled with numpy=1.x causing errors
-RUN python3 -m pip install --no-cache-dir './transformers[deepspeed-testing]' 'pandas<2' 'numpy<2'
+RUN python3 -m pip install --no-cache-dir ./transformers[deepspeed-testing]
 
 RUN python3 -m pip install --no-cache-dir git+https://github.com/huggingface/accelerate@main#egg=accelerate
 
@@ -35,8 +34,8 @@ RUN python3 -m pip uninstall -y torch-tensorrt apex
 # Pre-build **nightly** release of DeepSpeed, so it would be ready for testing (otherwise, the 1st deepspeed test will timeout)
 RUN python3 -m pip uninstall -y deepspeed
 # This has to be run inside the GPU VMs running the tests. (So far, it fails here due to GPU checks during compilation.)
-# Issue: https://github.com/deepspeedai/DeepSpeed/issues/2010
-# RUN git clone https://github.com/deepspeedai/DeepSpeed && cd DeepSpeed && rm -rf build && \
+# Issue: https://github.com/microsoft/DeepSpeed/issues/2010
+# RUN git clone https://github.com/microsoft/DeepSpeed && cd DeepSpeed && rm -rf build && \
 #    DS_BUILD_CPU_ADAM=1 DS_BUILD_FUSED_ADAM=1 DS_BUILD_UTILS=1 python3 -m pip install . --global-option="build_ext" --global-option="-j8" --no-cache -v --disable-pip-version-check 2>&1
 
 ## For `torchdynamo` tests
@@ -57,9 +56,6 @@ RUN python3 -m pip uninstall -y deepspeed
 #RUN git clone https://github.com/pytorch/TensorRT.git
 #RUN cd TensorRT/py && python3 setup.py install --fx-only
 
-# `kernels` may give different outputs (within 1e-5 range) even with the same model (weights) and the same inputs
-RUN python3 -m pip uninstall -y kernels
-
 # When installing in editable mode, `transformers` is not recognized as a package.
 # this line must be added in order for python to be aware of transformers.
 RUN cd transformers && python3 setup.py develop

docker/transformers-pytorch-gpu/Dockerfile

@@ -1,4 +1,4 @@
-FROM nvidia/cuda:12.6.0-cudnn-devel-ubuntu22.04
+FROM nvidia/cuda:12.1.0-cudnn8-devel-ubuntu20.04
 LABEL maintainer="Hugging Face"
 
 ARG DEBIAN_FRONTEND=noninteractive
@@ -11,28 +11,23 @@ ARG REF=main
 RUN git clone https://github.com/huggingface/transformers && cd transformers && git checkout $REF
 
 # If set to nothing, will install the latest version
-ARG PYTORCH='2.7.1'
+ARG PYTORCH='2.4.0'
 ARG TORCH_VISION=''
 ARG TORCH_AUDIO=''
 # Example: `cu102`, `cu113`, etc.
-ARG CUDA='cu126'
+ARG CUDA='cu121'
 
-RUN python3 -m pip install --no-cache-dir -e ./transformers[dev-torch,testing,video]
-
-# Install torch stuff after ./transformers[dev-torch,testing,video], otherwise torch may be resolved to a previous
-# version.
 RUN [ ${#PYTORCH} -gt 0 ] && VERSION='torch=='$PYTORCH'.*' ||  VERSION='torch'; python3 -m pip install --no-cache-dir -U $VERSION --extra-index-url https://download.pytorch.org/whl/$CUDA
 RUN [ ${#TORCH_VISION} -gt 0 ] && VERSION='torchvision=='TORCH_VISION'.*' ||  VERSION='torchvision'; python3 -m pip install --no-cache-dir -U $VERSION --extra-index-url https://download.pytorch.org/whl/$CUDA
 RUN [ ${#TORCH_AUDIO} -gt 0 ] && VERSION='torchaudio=='TORCH_AUDIO'.*' ||  VERSION='torchaudio'; python3 -m pip install --no-cache-dir -U $VERSION --extra-index-url https://download.pytorch.org/whl/$CUDA
 
+RUN python3 -m pip install --no-cache-dir -e ./transformers[dev-torch,testing,video]
+
 RUN python3 -m pip uninstall -y tensorflow flax
 
 RUN python3 -m pip install --no-cache-dir git+https://github.com/facebookresearch/detectron2.git pytesseract
 RUN python3 -m pip install -U "itsdangerous<2.1.0"
 
-# `kernels` may give different outputs (within 1e-5 range) even with the same model (weights) and the same inputs
-RUN python3 -m pip uninstall -y kernels
-
 # When installing in editable mode, `transformers` is not recognized as a package.
 # this line must be added in order for python to be aware of transformers.
 RUN cd transformers && python3 setup.py develop

docker/transformers-pytorch-xpu/Dockerfile

@@ -1,93 +0,0 @@
-FROM intel/deep-learning-essentials:2025.1.3-0-devel-ubuntu22.04 AS base
-LABEL maintainer="Hugging Face"
-
-SHELL ["/bin/bash", "-c"]
-
-ARG PYTHON_VER=3.11
-ENV TORCH_DEVICE_BACKEND_AUTOLOAD=0
-ENV DEBIAN_FRONTEND=noninteractive
-
-RUN apt-get remove -y python3.10 && apt-get autoremove -y
-RUN apt-get update && \
-    apt-get install -y software-properties-common && \
-    add-apt-repository -y ppa:deadsnakes/ppa && \
-    apt-get update && \
-    apt-get install -y python$PYTHON_VER python$PYTHON_VER-dev python3-pip && \
-    ln -sf /usr/bin/python$PYTHON_VER /usr/bin/python3 && \
-    ln -sf /usr/bin/python3 /usr/bin/python && \
-    apt-get clean && \
-    rm -rf /var/lib/apt/lists/*
-
-RUN apt-get update && \
-    apt-get -y install \
-        apt-utils \
-        build-essential \
-        ca-certificates \
-        clinfo \
-        curl \
-        git \
-        git-lfs \
-        vim \
-        numactl \
-        gnupg2 \
-        gpg-agent \
-        zlib1g-dev \
-        rsync \
-        sudo \
-        libnl-genl-3-200 \
-        xpu-smi \
-        unzip \
-        ffmpeg \
-        tesseract-ocr \
-        espeak-ng \
-        wget \
-        ncurses-term && \
-    apt-get clean && \
-    rm -rf /var/lib/apt/lists/*
-
-
-RUN apt-get update && \
-    apt-get install -y \
-        linux-headers-$(uname -r) \
-        linux-modules-extra-$(uname -r) \
-        flex bison \
-        intel-fw-gpu intel-i915-dkms xpu-smi \
-        intel-opencl-icd libze-intel-gpu1 libze1 \
-        intel-media-va-driver-non-free libmfx-gen1 libvpl2 \
-        libegl-mesa0 libegl1-mesa libegl1-mesa-dev libgbm1 libgl1-mesa-dev libgl1-mesa-dri \
-        libglapi-mesa libglx-mesa0 libigdgmm12 libxatracker2 mesa-va-drivers \
-        mesa-vdpau-drivers mesa-vulkan-drivers va-driver-all vainfo hwinfo clinfo intel-ocloc \
-        libigc-dev intel-igc-cm libigdfcl-dev libigfxcmrt-dev libze-dev && \
-    apt-get clean && \
-    rm -rf  /var/lib/apt/lists/*
-
-RUN pip install --upgrade pip
-RUN pip install triton==3.3.0
-
-RUN pip install torch==2.7.0 torchvision==0.22.0 torchaudio==2.7.0 --index-url https://download.pytorch.org/whl/xpu --no-cache-dir
-
-RUN pip install evaluate torchdata pyctcdecode pytesseract decord galore-torch fire scipy scikit-learn sentencepiece sacremoses nltk rouge_score librosa soundfile g2p_en mpi4py requests_mock
-RUN pip install pretty_midi essentia resampy Levenshtein av sacrebleu phonemizer invisible_watermark schedulefree
-RUN pip install gguf hqq compressed_tensors gptqmodel mergekit autoawq deepspeed torchao onnx
-RUN pip install hf_transfer huggingface-hub hf-doc-builder datasets optimum-quanto timm transformers accelerate optimum peft
-
-RUN pip install git+https://github.com/linkedin/Liger-Kernel.git --extra-index-url https://download.pytorch.org/whl/test/xpu
-
-# install bitsandbytes
-RUN pip install git+https://github.com/bitsandbytes-foundation/bitsandbytes.git
-
-ENV OCL_ICD_VENDORS=/etc/OpenCL/vendors
-ENV FI_PROVIDER_PATH=${I_MPI_ROOT}/lib/libfabric/prov:/usr/lib/x86_64-linux-gnu/libfabric
-ENV CCL_ROOT=/usr/local
-ENV CCL_ATL_TRANSPORT=ofi
-ENV I_MPI_ROOT=/usr/local
-ENV CLASSPATH=${I_MPI_ROOT}/lib/mpi.jar
-ENV PATH=${I_MPI_ROOT}/bin/libfabric:${PATH}
-ENV LD_LIBRARY_PATH=${I_MPI_ROOT}/lib/libfabric:${LD_LIBRARY_PATH}
-
-RUN touch /entrypoint.sh
-RUN chmod +x /entrypoint.sh
-RUN echo "#!/bin/bash" >> /entrypoint.sh
-RUN echo "source /opt/intel/oneapi/setvars.sh --force && /bin/bash" >> /entrypoint.sh
-
-ENTRYPOINT ["/entrypoint.sh"]

docker/transformers-quantization-latest-gpu/Dockerfile

@@ -1,4 +1,4 @@
-FROM nvidia/cuda:12.1.1-cudnn8-devel-ubuntu22.04
+FROM nvidia/cuda:11.8.0-cudnn8-devel-ubuntu20.04
 LABEL maintainer="Hugging Face"
 
 ARG DEBIAN_FRONTEND=noninteractive
@@ -9,14 +9,12 @@ SHELL ["sh", "-lc"]
 # The following `ARG` are mainly used to specify the versions explicitly & directly in this docker file, and not meant
 # to be used as arguments for docker build (so far).
 
-ARG PYTORCH='2.6.0'
+ARG PYTORCH='2.2.1'
 # Example: `cu102`, `cu113`, etc.
-ARG CUDA='cu121'
-# Disable kernel mapping for quantization tests
-ENV DISABLE_KERNEL_MAPPING=1
+ARG CUDA='cu118'
 
 RUN apt update
-RUN apt install -y git libsndfile1-dev tesseract-ocr espeak-ng python3 python3-pip ffmpeg
+RUN apt install -y git libsndfile1-dev tesseract-ocr espeak-ng python python3-pip ffmpeg
 RUN python3 -m pip install --no-cache-dir --upgrade pip
 
 ARG REF=main
@@ -28,6 +26,8 @@ RUN echo torch=$VERSION
 # Currently, let's just use their latest releases (when `torch` is installed with a release version)
 RUN python3 -m pip install --no-cache-dir -U $VERSION torchvision torchaudio --extra-index-url https://download.pytorch.org/whl/$CUDA
 
+RUN python3 -m pip install --no-cache-dir -e ./transformers[dev-torch]
+
 RUN python3 -m pip install --no-cache-dir git+https://github.com/huggingface/accelerate@main#egg=accelerate
 
 # needed in bnb and awq
@@ -36,26 +36,15 @@ RUN python3 -m pip install --no-cache-dir einops
 # Add bitsandbytes for mixed int8 testing
 RUN python3 -m pip install --no-cache-dir bitsandbytes
 
-# Add gptqmodel for gtpq quantization testing, installed from source for pytorch==2.6.0 compatibility
-RUN python3 -m pip install lm_eval
-RUN git clone https://github.com/ModelCloud/GPTQModel.git && cd GPTQModel && pip install -v . --no-build-isolation
+# Add auto-gptq for gtpq quantization testing
+RUN python3 -m pip install --no-cache-dir auto-gptq --extra-index-url https://huggingface.github.io/autogptq-index/whl/cu118/
 
 # Add optimum for gptq quantization testing
 RUN python3 -m pip install --no-cache-dir git+https://github.com/huggingface/optimum@main#egg=optimum
 
-# Add PEFT
-RUN python3 -m pip install --no-cache-dir git+https://github.com/huggingface/peft@main#egg=peft
-
 # Add aqlm for quantization testing
 RUN python3 -m pip install --no-cache-dir aqlm[gpu]==1.0.2
 
-# Add vptq for quantization testing
-RUN pip install vptq
-
-# Add spqr for quantization testing
-# Commented for now as No matching distribution found we need to reach out to the authors
-# RUN python3 -m pip install --no-cache-dir spqr_quant[gpu]
-
 # Add hqq for quantization testing
 RUN python3 -m pip install --no-cache-dir hqq
 
@@ -63,35 +52,14 @@ RUN python3 -m pip install --no-cache-dir hqq
 RUN python3 -m pip install --no-cache-dir gguf
 
 # Add autoawq for quantization testing
-# New release v0.2.8
-RUN python3 -m pip install --no-cache-dir autoawq[kernels]
+# >=v0.2.3 needed for compatibility with torch 2.2.1
+RUN python3 -m pip install --no-cache-dir https://github.com/casper-hansen/AutoAWQ/releases/download/v0.2.3/autoawq-0.2.3+cu118-cp38-cp38-linux_x86_64.whl
 
 # Add quanto for quantization testing
-RUN python3 -m pip install --no-cache-dir optimum-quanto
+RUN python3 -m pip install --no-cache-dir quanto
 
 # Add eetq for quantization testing
-RUN git clone https://github.com/NetEase-FuXi/EETQ.git && cd EETQ/ && git submodule update --init --recursive && pip install .
-
-# # Add flute-kernel and fast_hadamard_transform for quantization testing
-# # Commented for now as they cause issues with the build
-# # TODO: create a new workflow to test them
-# RUN python3 -m pip install --no-cache-dir flute-kernel==0.4.1
-# RUN python3 -m pip install --no-cache-dir git+https://github.com/Dao-AILab/fast-hadamard-transform.git
-
-# Add compressed-tensors for quantization testing
-RUN python3 -m pip install --no-cache-dir compressed-tensors
-
-# Add AMD Quark for quantization testing
-RUN python3 -m pip install --no-cache-dir amd-quark
-
-# Add AutoRound for quantization testing
-RUN python3 -m pip install --no-cache-dir "auto-round>=0.5.0"
-
-# Add transformers in editable mode
-RUN python3 -m pip install --no-cache-dir -e ./transformers[dev-torch]
-
-# `kernels` may give different outputs (within 1e-5 range) even with the same model (weights) and the same inputs
-RUN python3 -m pip uninstall -y kernels
+RUN python3 -m pip install git+https://github.com/NetEase-FuXi/EETQ.git
 
 # When installing in editable mode, `transformers` is not recognized as a package.
 # this line must be added in order for python to be aware of transformers.

docker/transformers-tensorflow-gpu/Dockerfile

@@ -1,4 +1,4 @@
-FROM nvidia/cuda:12.1.0-cudnn8-devel-ubuntu22.04
+FROM nvidia/cuda:12.1.0-cudnn8-devel-ubuntu20.04
 LABEL maintainer="Hugging Face"
 
 ARG DEBIAN_FRONTEND=noninteractive
@@ -18,7 +18,7 @@ RUN [ ${#TENSORFLOW} -gt 0 ] && VERSION='tensorflow=='$TENSORFLOW'.*' ||  VERSIO
 RUN python3 -m pip uninstall -y torch flax
 RUN python3 -m pip install -U "itsdangerous<2.1.0"
 
-RUN python3 -m pip install --no-cache-dir -U "tensorflow_probability<0.22"
+RUN python3 -m pip install --no-cache-dir -U tensorflow_probability
 
 # When installing in editable mode, `transformers` is not recognized as a package.
 # this line must be added in order for python to be aware of transformers.

docs/README.md

@@ -276,14 +276,14 @@ building the return.
 
 Here's an example of a single value return:
 
-```python
+```
     Returns:
-        `list[int]`: A list of integers in the range [0, 1] --- 1 for a special token, 0 for a sequence token.
+        `List[int]`: A list of integers in the range [0, 1] --- 1 for a special token, 0 for a sequence token.
 ```
 
 Here's an example of a tuple return, comprising several objects:
 
-```python
+```
     Returns:
         `tuple(torch.FloatTensor)` comprising various elements depending on the configuration ([`BertConfig`]) and inputs:
         - ** loss** (*optional*, returned when `masked_lm_labels` is provided) `torch.FloatTensor` of shape `(1,)` --
@@ -322,9 +322,10 @@ includes an example of how to transcribe speech to text in the
 
 The syntax for Example docstrings can look as follows:
 
-```python
+```
     Example:
 
+    ```python
     >>> from transformers import Wav2Vec2Processor, Wav2Vec2ForCTC
     >>> from datasets import load_dataset
     >>> import torch
@@ -346,6 +347,7 @@ The syntax for Example docstrings can look as follows:
     >>> transcription = processor.batch_decode(predicted_ids)
     >>> transcription[0]
     'MISTER QUILTER IS THE APOSTLE OF THE MIDDLE CLASSES AND WE ARE GLAD TO WELCOME HIS GOSPEL'
+    ```
 ```
 
 The docstring should give a minimal, clear example of how the respective model 

docs/TRANSLATING.md

@@ -1,70 +1,57 @@
-# Translating the Transformers documentation into your language
+### Translating the Transformers documentation into your language
 
-As part of our mission to democratize machine learning, we aim to make the Transformers library available in many more languages! Follow the steps below to help translate the documentation into your language.
+As part of our mission to democratize machine learning, we'd love to make the Transformers library available in many more languages! Follow the steps below if you want to help translate the documentation into your language 🙏.
 
-## Open an Issue
+**🗞️ Open an issue**
 
-1. Navigate to the Issues page of this repository.
-2. Check if anyone has already opened an issue for your language.
-3. If not, create a new issue by selecting the "Translation template" from the "New issue" button.
-4. Post a comment indicating which chapters you’d like to work on, and we’ll add your name to the list.
+To get started, navigate to the [Issues](https://github.com/huggingface/transformers/issues) page of this repo and check if anyone else has opened an issue for your language. If not, open a new issue by selecting the "Translation template" from the "New issue" button.
 
-## Fork the Repository
+Once an issue exists, post a comment to indicate which chapters you'd like to work on, and we'll add your name to the list.
 
-1. First, fork the Transformers repo by clicking the Fork button in the top-right corner.
-2. Clone your fork to your local machine for editing with the following command:
 
-    ```bash
-    git clone https://github.com/YOUR-USERNAME/transformers.git
-    ```
-   
-   Replace `YOUR-USERNAME` with your GitHub username.
+**🍴 Fork the repository**
 
-## Copy-paste the English version with a new language code
+First, you'll need to [fork the Transformers repo](https://docs.github.com/en/get-started/quickstart/fork-a-repo). You can do this by clicking on the **Fork** button on the top-right corner of this repo's page.
 
-The documentation files are organized in the following directory:
+Once you've forked the repo, you'll want to get the files on your local machine for editing. You can do that by cloning the fork with Git as follows:
 
-- **docs/source**: This contains all documentation materials organized by language.
+```bash
+git clone https://github.com/YOUR-USERNAME/transformers.git
+```
 
-To copy the English version to your new language directory:
+**📋 Copy-paste the English version with a new language code**
 
-1. Navigate to your fork of the repository:
+The documentation files are in one leading directory:
 
-    ```bash
-    cd ~/path/to/transformers/docs
-    ```
+- [`docs/source`](https://github.com/huggingface/transformers/tree/main/docs/source): All the documentation materials are organized here by language.
 
-   Replace `~/path/to` with your actual path.
+You'll only need to copy the files in the [`docs/source/en`](https://github.com/huggingface/transformers/tree/main/docs/source/en) directory, so first navigate to your fork of the repo and run the following:
 
-2. Run the following command:
+```bash
+cd ~/path/to/transformers/docs
+cp -r source/en source/LANG-ID
+```
 
-    ```bash
-    cp -r source/en source/LANG-ID
-    ```
+Here, `LANG-ID` should be one of the ISO 639-1 or ISO 639-2 language codes -- see [here](https://www.loc.gov/standards/iso639-2/php/code_list.php) for a handy table.
 
-   Replace `LANG-ID` with the appropriate ISO 639-1 or ISO 639-2 language code (see [this table](https://en.wikipedia.org/wiki/List_of_ISO_639-1_codes) for reference).
+**✍️ Start translating**
 
-## Start translating
+The fun part comes - translating the text!
 
-Begin translating the text!
+The first thing we recommend is translating the part of the `_toctree.yml` file that corresponds to your doc chapter. This file is used to render the table of contents on the website. 
 
-1. Start with the `_toctree.yml` file that corresponds to your documentation chapter. This file is essential for rendering the table of contents on the website.
+> 🙋 If the `_toctree.yml` file doesn't yet exist for your language, you can create one by copy-pasting from the English version and deleting the sections unrelated to your chapter. Just make sure it exists in the `docs/source/LANG-ID/` directory!
 
-    - If the `_toctree.yml` file doesn’t exist for your language, create one by copying the English version and removing unrelated sections.
-    - Ensure it is placed in the `docs/source/LANG-ID/` directory.
+The fields you should add are `local` (with the name of the file containing the translation; e.g. `autoclass_tutorial`), and `title` (with the title of the doc in your language; e.g. `Load pretrained instances with an AutoClass`) -- as a reference, here is the `_toctree.yml` for [English](https://github.com/huggingface/transformers/blob/main/docs/source/en/_toctree.yml):
 
-    Here’s an example structure for the `_toctree.yml` file:
+```yaml
+- sections:
+  - local: pipeline_tutorial # Do not change this! Use the same name for your .md file
+    title: Pipelines for inference # Translate this!
+    ...
+  title: Tutorials # Translate this!
+```
 
-    ```yaml
-    - sections:
-      - local: pipeline_tutorial # Keep this name for your .md file
-        title: Pipelines for Inference # Translate this
-        ...
-      title: Tutorials # Translate this
-    ```
+Once you have translated the `_toctree.yml` file, you can start translating the [MDX](https://mdxjs.com/) files associated with your docs chapter.
 
-2. Once you’ve translated the `_toctree.yml`, move on to translating the associated MDX files.
-
-## Collaborate and share
-
-If you'd like assistance with your translation, open an issue and tag `@stevhliu`. Feel free to share resources or glossaries to ensure consistent terminology.
+> 🙋 If you'd like others to help you with the translation, you should [open an issue](https://github.com/huggingface/transformers/issues) and tag @stevhliu and @MKhalusova.

docs/source/ar/_config.py

@@ -1,14 +0,0 @@
-# docstyle-ignore
-INSTALL_CONTENT = """
-# Transformers installation
-! pip install transformers datasets evaluate accelerate
-# To install from source instead of the last release, comment the command above and uncomment the following one.
-# ! pip install git+https://github.com/huggingface/transformers.git
-"""
-
-notebook_first_cells = [{"type": "code", "content": INSTALL_CONTENT}]
-black_avoid_patterns = {
-    "{processor_class}": "FakeProcessorClass",
-    "{model_class}": "FakeModelClass",
-    "{object_class}": "FakeObjectClass",
-}

docs/source/ar/_toctree.yml

@@ -1,894 +0,0 @@
-- sections:
-  - local: index
-    title: 🤗 المحولات
-  - local: quicktour
-    title: جولة سريعة
-  - local: installation
-    title: التثبيت
-  title: البدء
-- sections:
-  - local: pipeline_tutorial
-    title: تشغيل الاستنتاج باستخدام خطوط الأنابيب
-  - local: autoclass_tutorial
-    title: كتابة تعليمات برمجية متكيفه باستخدام AutoClass
-  - local: preprocessing
-    title: معالجة البيانات مسبقًا
-  - local: training
-    title: ضبط نموذج مسبق التدريب
-  - local: run_scripts
-    title: التدريب باستخدام نص برمجي
-  - local: accelerate
-    title: إعداد تدريب موزع باستخدام 🤗 Accelerate
-  - local: peft
-    title: تحميل النماذج المخصصة وتدريبها باستخدام 🤗 PEFT
-  - local: model_sharing
-    title: مشاركة نموذجك
-  - local: llm_tutorial
-    title: التوليد باستخدام LLMs
-  - local: conversations
-    title: الدردشة مع المحولات
-  title: البرامج التعليمية
-- sections:
-  - isExpanded: false
-    sections:
-    - local: tasks/sequence_classification
-      title: تصنيف النصوص
-    - local: tasks/token_classification
-      title: تصنيف الرموز
-    - local: tasks/question_answering
-      title: الإجابة على الأسئلة
-    - local: tasks/language_modeling
-      title: نمذجة اللغة السببية
-    - local: tasks/masked_language_modeling
-      title: نمذجة اللغة المقنعة
-    - local: tasks/translation
-      title: الترجمة
-    - local: tasks/summarization
-      title: التلخيص
-    - local: tasks/multiple_choice
-      title: الاختيار المتعدد
-    title: معالجة اللغات الطبيعية
-#   - isExpanded: false
-#     sections:
-#     - local: tasks/audio_classification
-#       title: تصنيف الصوت
-#     - local: tasks/asr
-#       title: التعرف التلقائي على الكلام
-#     title: الصوت
-#   - isExpanded: false
-#     sections:
-#     - local: tasks/image_classification
-#       title: تصنيف الصور
-#     - local: tasks/semantic_segmentation
-#       title: تجزئة الصور
-#     - local: tasks/video_classification
-#       title: تصنيف الفيديو
-#     - local: tasks/object_detection
-#       title: اكتشاف الأشياء
-#     - local: tasks/zero_shot_object_detection
-#       title: اكتشاف الأشياء بدون تدريب
-#     - local: tasks/zero_shot_image_classification
-#       title: تصنيف الصور بدون تدريب
-#     - local: tasks/monocular_depth_estimation
-#       title: تقدير العمق
-#     - local: tasks/image_to_image
-#       title: صورة إلى صورة
-#     - local: tasks/image_feature_extraction
-#       title: استخراج ميزات الصورة
-#     - local: tasks/mask_generation
-#       title: توليد القناع
-#     - local: tasks/knowledge_distillation_for_image_classification
-#       title: التقليل المعرفي للرؤية الحاسوبية
-#     title: الرؤية الحاسوبية
-#   - isExpanded: false
-#     sections:
-#     - local: tasks/image_captioning
-#       title: وصف الصور Image captioning
-#     - local: tasks/document_question_answering
-#       title: الإجابة على أسئلة المستندات
-#     - local: tasks/visual_question_answering
-#       title: الإجابة على الأسئلة المرئية
-#     - local: tasks/text-to-speech
-#       title: تحويل النص إلى كلام
-#     title: المتعددة الوسائط
-#   - isExpanded: false
-#     sections:
-#     - local: generation_strategies
-#       title: تخصيص استراتيجية التوليد
-#     - local: kv_cache
-#       title: أفضل الممارسات للتوليد باستخدام ذاكرة التخزين المؤقت
-#     title: التوليد
-#   - isExpanded: false
-#     sections:
-#     - local: tasks/idefics
-#       title: مهام الصور مع IDEFICS
-#     - local: tasks/prompting
-#       title: دليل إرشادي لمحفزات النماذج اللغوية الكبيرة
-#     title: الإرشاد
-  title: أدلة المهام
-- sections:
-  - local: fast_tokenizers
-    title: استخدم مجزئيات النصوص السريعة من 🤗 Tokenizers
-  - local: multilingual
-    title: الاستدلال باستخدام نماذج متعددة اللغات
-  - local: create_a_model
-    title: استخدام واجهات برمجة التطبيقات الخاصة بالنموذج
-  - local: custom_models
-    title: مشاركة نموذج مخصص
-  - local: chat_templating
-    title: قوالب لنماذج الدردشة
-  - local: trainer
-    title: المدرب
-  - local: sagemaker
-    title: تشغيل التدريب على Amazon SageMaker
-  - local: serialization
-    title: التصدير إلى ONNX
-  - local: tflite
-    title: التصدير إلى TFLite
-  - local: torchscript
-    title: التصدير إلى TorchScript
-  - local: notebooks
-    title: دفاتر الملاحظات مع الأمثلة
-  - local: community
-    title: موارد المجتمع
-  - local: troubleshooting
-    title: استكشاف الأخطاء وإصلاحها
-  - local: gguf
-    title: التوافق مع ملفات GGUF
-  - local: tiktoken
-    title: التوافق مع ملفات TikToken
-  - local: modular_transformers
-    title: الوحدات النمطية في `transformers`
-  - local: how_to_hack_models
-    title: اختراق النموذج (الكتابة فوق فئة لاستخدامك)
-  title: أدلة المطورين
-# - sections:
-#   - local: quantization/overview
-#     title: نظرة عامة
-#   - local: quantization/bitsandbytes
-#     title: bitsandbytes
-#   - local: quantization/gptq
-#     title: GPTQ
-#   - local: quantization/awq
-#     title: AWQ
-#   - local: quantization/aqlm
-#     title: AQLM
-#   - local: quantization/vptq
-#     title: VPTQ
-#   - local: quantization/quanto
-#     title: Quanto
-#   - local: quantization/eetq
-#     title: EETQ
-#   - local: quantization/hqq
-#     title: HQQ
-#   - local: quantization/optimum
-#     title: Optimum
-#   - local: quantization/contribute
-#     title: المساهمة بطريقة جديدة للتكميم
-#   title: أساليب التكميم
-# - sections:
-#   - local: performance
-#     title: الأداء-نظرة عامة
-#   - local: llm_optims
-#     title: تحسين الاستدلال LLM
-#   - sections:
-#     - local: perf_train_gpu_one
-#       title: استخدام عدة وحدات معالجة رسوميات (GPUs) بشكل متوازٍ
-#     - local: perf_train_gpu_many
-#       title: وحدات معالجة الرسومات (GPU) متعددة والتوازي
-#     - local: fsdp
-#       title: Fully Sharded Data Parallel
-#     - local: deepspeed
-#       title: DeepSpeed
-#     - local: perf_train_cpu
-#       title: التدريب الفعال على وحدة المعالجة المركزية (CPU)
-#     - local: perf_train_cpu_many
-#       title: التدريب الموزع لوحدة المعالجة المركزية (CPU)
-#     - local: perf_train_tpu_tf
-#       title: التدريب على (TPU) باستخدام TensorFlow
-#     - local: perf_train_special
-#       title: تدريب PyTorch على Apple silicon
-#     - local: perf_hardware
-#       title: الأجهزة المخصصة للتدريب
-#     - local: hpo_train
-#       title: البحث عن المعاملات المثلى باستخدام واجهة برمجة تطبيقات المدرب
-#     title: تقنيات التدريب الفعال
-#   - sections:
-#     - local: perf_infer_cpu
-#       title: الإستدلال على وحدة المعالجة المركزية (CPU)
-#     - local: perf_infer_gpu_one
-#       title: الإستدلال على وحدة معالجة الرسومات (GPU)
-#     title: تحسين الاستدلال
-#   - local: big_models
-#     title: إنشاء نموذج كبير
-#   - local: debugging
-#     title: تصحيح الأخطاء البرمجية
-#   - local: tf_xla
-#     title: تكامل XLA لنماذج TensorFlow
-#   - local: perf_torch_compile
-#     title: تحسين الاستدلال باستخدام `torch.compile()`
-#   title: الأداء وقابلية التوسع
-# - sections:
-#   - local: contributing
-#     title: كيفية المساهمة في 🤗 المحولات؟
-#   - local: add_new_model
-#     title: كيفية إضافة نموذج إلى 🤗 المحولات؟
-#   - local: add_new_pipeline
-#     title: كيفية إضافة خط أنابيب إلى 🤗 المحولات؟
-#   - local: testing
-#     title: الاختبار
-#   - local: pr_checks
-#     title: التحقق من طلب السحب
-#   title: المساهمة
-- sections:
-  - local: philosophy
-    title: الفلسفة
-  - local: glossary
-    title: (قاموس المصطلحات (قائمة الكلمات
-  - local: task_summary
-    title: ما الذي يمكن أن تفعله 🤗 المحولات
-  - local: tasks_explained
-    title: كيف تحل المحولات المهام
-  - local: model_summary
-    title: عائلة نماذج المحول
-  - local: tokenizer_summary
-    title: ملخص برنامج مقسم النصوص (tokenizers)
-  - local: attention
-    title: الانتباه Attention
-  - local: pad_truncation
-    title: الحشو والتقليم
-  - local: bertology
-    title: BERTology
-  - local: perplexity
-    title: حيرة النماذج ذات الطول الثابت
-  - local: pipeline_webserver
-    title: خطوط الأنابيب للاستدلال على خادم الويب
-  - local: model_memory_anatomy
-    title: تشريح تدريب النموذج
-  - local: llm_tutorial_optimization
-    title: الاستفادة القصوى من LLMs
-  title: أطر مفاهيمية
-# - sections:
-#   - sections:
-#     - local: model_doc/auto
-#       title: فئات يتم إنشاؤها ديناميكيًا
-#     - local: main_classes/backbones
-#       title: العمود الفقري
-#     - local: main_classes/callback
-#       title: عمليات الاسترجاع
-#     - local: main_classes/configuration
-#       title: التكوين
-#     - local: main_classes/data_collator
-#       title: مجمع البيانات
-#     - local: main_classes/keras_callbacks
-#       title: استدعاءات Keras
-#     - local: main_classes/logging
-#       title: التسجيل
-#     - local: main_classes/model
-#       title: النماذج
-#     - local: main_classes/text_generation
-#       title: توليد النصوص
-#     - local: main_classes/onnx
-#       title: ONNX
-#     - local: main_classes/optimizer_schedules
-#       title: التحسين
-#     - local: main_classes/output
-#       title: مخرجات النموذج
-#     - local: main_classes/pipelines
-#       title: خطوط الأنابيب
-#     - local: main_classes/processors
-#       title: المعالجات
-#     - local: main_classes/quantization
-#       title: التكميم
-#     - local: main_classes/tokenizer
-#       title: برنامج مقسم النصوص
-#     - local: main_classes/trainer
-#       title: المدرب
-#     - local: main_classes/deepspeed
-#       title: DeepSpeed
-#     - local: main_classes/feature_extractor
-#       title: مستخرج الميزات
-#     - local: main_classes/image_processor
-#       title: معالج الصور
-#     title: الفئات الرئيسية
-#   - sections:
-#     - isExpanded: false
-#       sections:
-#       - local: model_doc/albert
-#         title: ALBERT
-#       - local: model_doc/bart
-#         title: BART
-#       - local: model_doc/barthez
-#         title: BARThez
-#       - local: model_doc/bartpho
-#         title: BARTpho
-#       - local: model_doc/bert
-#         title: BERT
-#       - local: model_doc/bert-generation
-#         title: BertGeneration
-#       - local: model_doc/bert-japanese
-#         title: BertJapanese
-#       - local: model_doc/bertweet
-#         title: Bertweet
-#       - local: model_doc/big_bird
-#         title: BigBird
-#       - local: model_doc/bigbird_pegasus
-#         title: BigBirdPegasus
-#       - local: model_doc/biogpt
-#         title: BioGpt
-#       - local: model_doc/blenderbot
-#         title: Blenderbot
-#       - local: model_doc/blenderbot-small
-#         title: Blenderbot Small
-#       - local: model_doc/bloom
-#         title: BLOOM
-#       - local: model_doc/bort
-#         title: BORT
-#       - local: model_doc/byt5
-#         title: ByT5
-#       - local: model_doc/camembert
-#         title: CamemBERT
-#       - local: model_doc/canine
-#         title: CANINE
-#       - local: model_doc/codegen
-#         title: CodeGen
-#       - local: model_doc/code_llama
-#         title: CodeLlama
-#       - local: model_doc/cohere
-#         title: Cohere
-#       - local: model_doc/convbert
-#         title: ConvBERT
-#       - local: model_doc/cpm
-#         title: CPM
-#       - local: model_doc/cpmant
-#         title: CPMANT
-#       - local: model_doc/ctrl
-#         title: CTRL
-#       - local: model_doc/dbrx
-#         title: DBRX
-#       - local: model_doc/deberta
-#         title: DeBERTa
-#       - local: model_doc/deberta-v2
-#         title: DeBERTa-v2
-#       - local: model_doc/dialogpt
-#         title: DialoGPT
-#       - local: model_doc/distilbert
-#         title: DistilBERT
-#       - local: model_doc/dpr
-#         title: DPR
-#       - local: model_doc/electra
-#         title: ELECTRA
-#       - local: model_doc/encoder-decoder
-#         title: Encoder Decoder Models
-#       - local: model_doc/ernie
-#         title: ERNIE
-#       - local: model_doc/ernie_m
-#         title: ErnieM
-#       - local: model_doc/esm
-#         title: ESM
-#       - local: model_doc/falcon
-#         title: Falcon
-#       - local: model_doc/fastspeech2_conformer
-#         title: FastSpeech2Conformer
-#       - local: model_doc/flan-t5
-#         title: FLAN-T5
-#       - local: model_doc/flan-ul2
-#         title: FLAN-UL2
-#       - local: model_doc/flaubert
-#         title: FlauBERT
-#       - local: model_doc/fnet
-#         title: FNet
-#       - local: model_doc/fsmt
-#         title: FSMT
-#       - local: model_doc/funnel
-#         title: Funnel Transformer
-#       - local: model_doc/fuyu
-#         title: Fuyu
-#       - local: model_doc/gemma
-#         title: Gemma
-#       - local: model_doc/openai-gpt
-#         title: GPT
-#       - local: model_doc/gpt_neo
-#         title: GPT Neo
-#       - local: model_doc/gpt_neox
-#         title: GPT NeoX
-#       - local: model_doc/gpt_neox_japanese
-#         title: GPT NeoX Japanese
-#       - local: model_doc/gptj
-#         title: GPT-J
-#       - local: model_doc/gpt2
-#         title: GPT2
-#       - local: model_doc/gpt_bigcode
-#         title: GPTBigCode
-#       - local: model_doc/gptsan-japanese
-#         title: GPTSAN Japanese
-#       - local: model_doc/gpt-sw3
-#         title: GPTSw3
-#       - local: model_doc/herbert
-#         title: HerBERT
-#       - local: model_doc/ibert
-#         title: I-BERT
-#       - local: model_doc/jamba
-#         title: Jamba
-#       - local: model_doc/jetmoe
-#         title: JetMoe
-#       - local: model_doc/jukebox
-#         title: Jukebox
-#       - local: model_doc/led
-#         title: LED
-#       - local: model_doc/llama
-#         title: LLaMA
-#       - local: model_doc/llama2
-#         title: Llama2
-#       - local: model_doc/llama3
-#         title: Llama3
-#       - local: model_doc/longformer
-#         title: Longformer
-#       - local: model_doc/longt5
-#         title: LongT5
-#       - local: model_doc/luke
-#         title: LUKE
-#       - local: model_doc/m2m_100
-#         title: M2M100
-#       - local: model_doc/madlad-400
-#         title: MADLAD-400
-#       - local: model_doc/mamba
-#         title: Mamba
-#       - local: model_doc/marian
-#         title: MarianMT
-#       - local: model_doc/markuplm
-#         title: MarkupLM
-#       - local: model_doc/mbart
-#         title: MBart and MBart-50
-#       - local: model_doc/mega
-#         title: MEGA
-#       - local: model_doc/megatron-bert
-#         title: MegatronBERT
-#       - local: model_doc/megatron_gpt2
-#         title: MegatronGPT2
-#       - local: model_doc/mistral
-#         title: Mistral
-#       - local: model_doc/mixtral
-#         title: Mixtral
-#       - local: model_doc/mluke
-#         title: mLUKE
-#       - local: model_doc/mobilebert
-#         title: MobileBERT
-#       - local: model_doc/mpnet
-#         title: MPNet
-#       - local: model_doc/mpt
-#         title: MPT
-#       - local: model_doc/mra
-#         title: MRA
-#       - local: model_doc/mt5
-#         title: MT5
-#       - local: model_doc/mvp
-#         title: MVP
-#       - local: model_doc/nezha
-#         title: NEZHA
-#       - local: model_doc/nllb
-#         title: NLLB
-#       - local: model_doc/nllb-moe
-#         title: NLLB-MoE
-#       - local: model_doc/nystromformer
-#         title: Nyströmformer
-#       - local: model_doc/olmo
-#         title: OLMo
-#       - local: model_doc/open-llama
-#         title: Open-Llama
-#       - local: model_doc/opt
-#         title: OPT
-#       - local: model_doc/pegasus
-#         title: Pegasus
-#       - local: model_doc/pegasus_x
-#         title: PEGASUS-X
-#       - local: model_doc/persimmon
-#         title: Persimmon
-#       - local: model_doc/phi
-#         title: Phi
-#       - local: model_doc/phi3
-#         title: Phi-3
-#       - local: model_doc/phobert
-#         title: PhoBERT
-#       - local: model_doc/plbart
-#         title: PLBart
-#       - local: model_doc/prophetnet
-#         title: ProphetNet
-#       - local: model_doc/qdqbert
-#         title: QDQBert
-#       - local: model_doc/qwen2
-#         title: Qwen2
-#       - local: model_doc/qwen2_moe
-#         title: Qwen2MoE
-#       - local: model_doc/rag
-#         title: RAG
-#       - local: model_doc/realm
-#         title: REALM
-#       - local: model_doc/recurrent_gemma
-#         title: RecurrentGemma
-#       - local: model_doc/reformer
-#         title: Reformer
-#       - local: model_doc/rembert
-#         title: RemBERT
-#       - local: model_doc/retribert
-#         title: RetriBERT
-#       - local: model_doc/roberta
-#         title: RoBERTa
-#       - local: model_doc/roberta-prelayernorm
-#         title: RoBERTa-PreLayerNorm
-#       - local: model_doc/roc_bert
-#         title: RoCBert
-#       - local: model_doc/roformer
-#         title: RoFormer
-#       - local: model_doc/rwkv
-#         title: RWKV
-#       - local: model_doc/splinter
-#         title: Splinter
-#       - local: model_doc/squeezebert
-#         title: SqueezeBERT
-#       - local: model_doc/stablelm
-#         title: StableLm
-#       - local: model_doc/starcoder2
-#         title: Starcoder2
-#       - local: model_doc/switch_transformers
-#         title: SwitchTransformers
-#       - local: model_doc/t5
-#         title: T5
-#       - local: model_doc/t5v1.1
-#         title: T5v1.1
-#       - local: model_doc/tapex
-#         title: TAPEX
-#       - local: model_doc/transfo-xl
-#         title: Transformer XL
-#       - local: model_doc/ul2
-#         title: UL2
-#       - local: model_doc/umt5
-#         title: UMT5
-#       - local: model_doc/xmod
-#         title: X-MOD
-#       - local: model_doc/xglm
-#         title: XGLM
-#       - local: model_doc/xlm
-#         title: XLM
-#       - local: model_doc/xlm-prophetnet
-#         title: XLM-ProphetNet
-#       - local: model_doc/xlm-roberta
-#         title: XLM-RoBERTa
-#       - local: model_doc/xlm-roberta-xl
-#         title: XLM-RoBERTa-XL
-#       - local: model_doc/xlm-v
-#         title: XLM-V
-#       - local: model_doc/xlnet
-#         title: XLNet
-#       - local: model_doc/yoso
-#         title: YOSO
-#       title: Text models
-#     - isExpanded: false
-#       sections:
-#       - local: model_doc/beit
-#         title: BEiT
-#       - local: model_doc/bit
-#         title: BiT
-#       - local: model_doc/conditional_detr
-#         title: Conditional DETR
-#       - local: model_doc/convnext
-#         title: ConvNeXT
-#       - local: model_doc/convnextv2
-#         title: ConvNeXTV2
-#       - local: model_doc/cvt
-#         title: CVT
-#       - local: model_doc/deformable_detr
-#         title: Deformable DETR
-#       - local: model_doc/deit
-#         title: DeiT
-#       - local: model_doc/depth_anything
-#         title: Depth Anything
-#       - local: model_doc/deta
-#         title: DETA
-#       - local: model_doc/detr
-#         title: DETR
-#       - local: model_doc/dinat
-#         title: DiNAT
-#       - local: model_doc/dinov2
-#         title: DINOV2
-#       - local: model_doc/dit
-#         title: DiT
-#       - local: model_doc/dpt
-#         title: DPT
-#       - local: model_doc/efficientformer
-#         title: EfficientFormer
-#       - local: model_doc/efficientnet
-#         title: EfficientNet
-#       - local: model_doc/focalnet
-#         title: FocalNet
-#       - local: model_doc/glpn
-#         title: GLPN
-#       - local: model_doc/imagegpt
-#         title: ImageGPT
-#       - local: model_doc/levit
-#         title: LeViT
-#       - local: model_doc/mask2former
-#         title: Mask2Former
-#       - local: model_doc/maskformer
-#         title: MaskFormer
-#       - local: model_doc/mobilenet_v1
-#         title: MobileNetV1
-#       - local: model_doc/mobilenet_v2
-#         title: MobileNetV2
-#       - local: model_doc/mobilevit
-#         title: MobileViT
-#       - local: model_doc/mobilevitv2
-#         title: MobileViTV2
-#       - local: model_doc/nat
-#         title: NAT
-#       - local: model_doc/poolformer
-#         title: PoolFormer
-#       - local: model_doc/pvt
-#         title: Pyramid Vision Transformer (PVT)
-#       - local: model_doc/pvt_v2
-#         title: Pyramid Vision Transformer v2 (PVTv2)
-#       - local: model_doc/regnet
-#         title: RegNet
-#       - local: model_doc/resnet
-#         title: ResNet
-#       - local: model_doc/segformer
-#         title: SegFormer
-#       - local: model_doc/seggpt
-#         title: SegGpt
-#       - local: model_doc/superpoint
-#         title: SuperPoint
-#       - local: model_doc/swiftformer
-#         title: SwiftFormer
-#       - local: model_doc/swin
-#         title: Swin Transformer
-#       - local: model_doc/swinv2
-#         title: Swin Transformer V2
-#       - local: model_doc/swin2sr
-#         title: Swin2SR
-#       - local: model_doc/table-transformer
-#         title: Table Transformer
-#       - local: model_doc/upernet
-#         title: UperNet
-#       - local: model_doc/van
-#         title: VAN
-#       - local: model_doc/vit
-#         title: Vision Transformer (ViT)
-#       - local: model_doc/vit_hybrid
-#         title: ViT Hybrid
-#       - local: model_doc/vitdet
-#         title: ViTDet
-#       - local: model_doc/vit_mae
-#         title: ViTMAE
-#       - local: model_doc/vitmatte
-#         title: ViTMatte
-#       - local: model_doc/vit_msn
-#         title: ViTMSN
-#       - local: model_doc/yolos
-#         title: YOLOS
-#       title: Vision models
-#     - isExpanded: false
-#       sections:
-#       - local: model_doc/audio-spectrogram-transformer
-#         title: Audio Spectrogram Transformer
-#       - local: model_doc/bark
-#         title: Bark
-#       - local: model_doc/clap
-#         title: CLAP
-#       - local: model_doc/encodec
-#         title: EnCodec
-#       - local: model_doc/hubert
-#         title: Hubert
-#       - local: model_doc/mctct
-#         title: MCTCT
-#       - local: model_doc/mms
-#         title: MMS
-#       - local: model_doc/musicgen
-#         title: MusicGen
-#       - local: model_doc/musicgen_melody
-#         title: MusicGen Melody
-#       - local: model_doc/pop2piano
-#         title: Pop2Piano
-#       - local: model_doc/seamless_m4t
-#         title: Seamless-M4T
-#       - local: model_doc/seamless_m4t_v2
-#         title: SeamlessM4T-v2
-#       - local: model_doc/sew
-#         title: SEW
-#       - local: model_doc/sew-d
-#         title: SEW-D
-#       - local: model_doc/speech_to_text
-#         title: Speech2Text
-#       - local: model_doc/speech_to_text_2
-#         title: Speech2Text2
-#       - local: model_doc/speecht5
-#         title: SpeechT5
-#       - local: model_doc/unispeech
-#         title: UniSpeech
-#       - local: model_doc/unispeech-sat
-#         title: UniSpeech-SAT
-#       - local: model_doc/univnet
-#         title: UnivNet
-#       - local: model_doc/vits
-#         title: VITS
-#       - local: model_doc/wav2vec2
-#         title: Wav2Vec2
-#       - local: model_doc/wav2vec2-bert
-#         title: Wav2Vec2-BERT
-#       - local: model_doc/wav2vec2-conformer
-#         title: Wav2Vec2-Conformer
-#       - local: model_doc/wav2vec2_phoneme
-#         title: Wav2Vec2Phoneme
-#       - local: model_doc/wavlm
-#         title: WavLM
-#       - local: model_doc/whisper
-#         title: Whisper
-#       - local: model_doc/xls_r
-#         title: XLS-R
-#       - local: model_doc/xlsr_wav2vec2
-#         title: XLSR-Wav2Vec2
-#       title: Audio models
-#     - isExpanded: false
-#       sections:
-#       - local: model_doc/timesformer
-#         title: TimeSformer
-#       - local: model_doc/videomae
-#         title: VideoMAE
-#       - local: model_doc/vivit
-#         title: ViViT
-#       title: Video models
-#     - isExpanded: false
-#       sections:
-#       - local: model_doc/align
-#         title: ALIGN
-#       - local: model_doc/altclip
-#         title: AltCLIP
-#       - local: model_doc/blip
-#         title: BLIP
-#       - local: model_doc/blip-2
-#         title: BLIP-2
-#       - local: model_doc/bridgetower
-#         title: BridgeTower
-#       - local: model_doc/bros
-#         title: BROS
-#       - local: model_doc/chinese_clip
-#         title: Chinese-CLIP
-#       - local: model_doc/clip
-#         title: CLIP
-#       - local: model_doc/clipseg
-#         title: CLIPSeg
-#       - local: model_doc/clvp
-#         title: CLVP
-#       - local: model_doc/data2vec
-#         title: Data2Vec
-#       - local: model_doc/deplot
-#         title: DePlot
-#       - local: model_doc/donut
-#         title: Donut
-#       - local: model_doc/flava
-#         title: FLAVA
-#       - local: model_doc/git
-#         title: GIT
-#       - local: model_doc/grounding-dino
-#         title: Grounding DINO
-#       - local: model_doc/groupvit
-#         title: GroupViT
-#       - local: model_doc/idefics
-#         title: IDEFICS
-#       - local: model_doc/idefics2
-#         title: Idefics2
-#       - local: model_doc/instructblip
-#         title: InstructBLIP
-#       - local: model_doc/kosmos-2
-#         title: KOSMOS-2
-#       - local: model_doc/layoutlm
-#         title: LayoutLM
-#       - local: model_doc/layoutlmv2
-#         title: LayoutLMV2
-#       - local: model_doc/layoutlmv3
-#         title: LayoutLMV3
-#       - local: model_doc/layoutxlm
-#         title: LayoutXLM
-#       - local: model_doc/lilt
-#         title: LiLT
-#       - local: model_doc/llava
-#         title: Llava
-#       - local: model_doc/llava_next
-#         title: LLaVA-NeXT
-#       - local: model_doc/lxmert
-#         title: LXMERT
-#       - local: model_doc/matcha
-#         title: MatCha
-#       - local: model_doc/mgp-str
-#         title: MGP-STR
-#       - local: model_doc/nougat
-#         title: Nougat
-#       - local: model_doc/oneformer
-#         title: OneFormer
-#       - local: model_doc/owlvit
-#         title: OWL-ViT
-#       - local: model_doc/owlv2
-#         title: OWLv2
-#       - local: model_doc/paligemma
-#         title: PaliGemma
-#       - local: model_doc/perceiver
-#         title: Perceiver
-#       - local: model_doc/pix2struct
-#         title: Pix2Struct
-#       - local: model_doc/sam
-#         title: Segment Anything
-#       - local: model_doc/siglip
-#         title: SigLIP
-#       - local: model_doc/speech-encoder-decoder
-#         title: Speech Encoder Decoder Models
-#       - local: model_doc/tapas
-#         title: TAPAS
-#       - local: model_doc/trocr
-#         title: TrOCR
-#       - local: model_doc/tvlt
-#         title: TVLT
-#       - local: model_doc/tvp
-#         title: TVP
-#       - local: model_doc/udop
-#         title: UDOP
-#       - local: model_doc/video_llava
-#         title: VideoLlava
-#       - local: model_doc/vilt
-#         title: ViLT
-#       - local: model_doc/vipllava
-#         title: VipLlava
-#       - local: model_doc/vision-encoder-decoder
-#         title: Vision Encoder Decoder Models
-#       - local: model_doc/vision-text-dual-encoder
-#         title: Vision Text Dual Encoder
-#       - local: model_doc/visual_bert
-#         title: VisualBERT
-#       - local: model_doc/xclip
-#         title: X-CLIP
-#       title: Multimodal models
-#     - isExpanded: false
-#       sections:
-#       - local: model_doc/decision_transformer
-#         title: محول القرار
-#       - local: model_doc/trajectory_transformer
-#         title: محول المسار
-#       title: نماذج التعلم التعزيزية
-#     - isExpanded: false
-#       sections:
-#       - local: model_doc/autoformer
-#         title: Autoformer
-#       - local: model_doc/informer
-#         title: Informer
-#       - local: model_doc/patchtsmixer
-#         title: PatchTSMixer
-#       - local: model_doc/patchtst
-#         title: PatchTST
-#       - local: model_doc/time_series_transformer
-#         title: محول السلاسل الزمنية
-#       title: نماذج السلاسل الزمنية
-#     - isExpanded: false
-#       sections:
-#       - local: model_doc/graphormer
-#         title: Graphormer
-#       title: نماذج الرسم البياني
-#     title: النماذج
-#   - sections:
-#     - local: internal/modeling_utils
-#       title: الطبقات المخصصة والمرافق
-#     - local: internal/pipelines_utils
-#       title: مرافق خطوط الأنابيب
-#     - local: internal/tokenization_utils
-#       title: مرافق مقسم النصوص
-#     - local: internal/trainer_utils
-#       title: مرافق المدرب
-#     - local: internal/generation_utils
-#       title: مرافق التوليد
-#     - local: internal/image_processing_utils
-#       title: مرافق معالجة الصور
-#     - local: internal/audio_utils
-#       title: مرافق معالجة الصوت
-#     - local: internal/file_utils
-#       title: مرافق عامة
-#     - local: internal/time_series_utils
-#       title: مرافق السلاسل الزمنية
-#     title: مساعدون داخليون
-#   title: API

docs/source/ar/accelerate.md

@@ -1,120 +0,0 @@
-# التدريب الموزع باستخدام  🤗 Accelerate
-
-
-مع تزايد حجم النماذج اللغوية، برز التوازي كأحد الاستراتيجيات لتدريب نماذج أكبر على أجهزة محدودة وتسريع عملية التدريب بمقدار كبير.  أنشأنا في Hugging Face، قمنا بإنشاء مكتبة [ Accelerate](https://huggingface.co/docs/accelerate) لمساعدة المستخدمين على تدريب أي نموذج من  Transformers بسهولة على أي نوع من الإعدادات الموزعة، سواء كان ذلك على عدة وحدات معالجة رسومات (GPUs) على جهاز واحد أو على عدة وحدات معالجة رسومات موزعة على عدة أجهزة. في هذا الدليل، تعلم كيفية تخصيص حلقة تدريب PyTorch الأصلية لتمكين التدريب في بيئة موزعة.
-
-## الإعداد
-
-ابدأ بتثبيت 🤗 Accelerate:
-
-```bash
-pip install accelerate
-```
-
-ثم قم باستيراد وإنشاء كائن [`~accelerate.Accelerator`]. سيقوم [`~accelerate.Accelerator`] تلقائيًا باكتشاف نوع الإعداد الموزع الخاص بك وتهيئة جميع المكونات اللازمة للتدريب. لن تحتاج إلى وضع نموذجك على جهاز بشكل معين.
-
-```py
->>> from accelerate import Accelerator
-
->>> accelerator = Accelerator()
-```
-
-## الاستعداد للتسريع
-
-الخطوة التالية هي تمرير جميع كائنات التدريب ذات الصلة إلى دالة الإعداد [`~accelerate.Accelerator.prepare`]. ويشمل ذلك DataLoaders للتدريب والتقييم، ونموذجًا ومُحَسِّنً المعاملات (optimizer):
-
-```py
->>> train_dataloader, eval_dataloader, model, optimizer = accelerator.prepare(
-...     train_dataloader, eval_dataloader, model, optimizer
-... )
-```
-
-## الخلفي Backward
-
-الإضافة الأخيرة هي استبدال الدالة المعتادة `loss.backward()` في حلقة التدريب الخاصة بك بدالة [`~accelerate.Accelerator.backward`] في 🤗 Accelerate:
-
-```py
->>> for epoch in range(num_epochs):
-...     for batch in train_dataloader:
-...         outputs = model(**batch)
-...         loss = outputs.loss
-...         accelerator.backward(loss)
-
-...         optimizer.step()
-...         lr_scheduler.step()
-...         optimizer.zero_grad()
-...         progress_bar.update(1)
-```
-
-كما يمكنك أن ترى في الكود التالي، فأنت بحاجة فقط إلى إضافة أربعة أسطر من الكود إلى حلقة التدريب الخاصة بك لتمكين التدريب الموزع!
-
-```diff
-+ from accelerate import Accelerator
-  from transformers import AdamW, AutoModelForSequenceClassification, get_scheduler
-
-+ accelerator = Accelerator()
-
-  model = AutoModelForSequenceClassification.from_pretrained(checkpoint, num_labels=2)
-  optimizer = AdamW(model.parameters(), lr=3e-5)
-
-- device = torch.device("cuda") if torch.cuda.is_available() else torch.device("cpu")
-- model.to(device)
-
-+ train_dataloader, eval_dataloader, model, optimizer = accelerator.prepare(
-+     train_dataloader, eval_dataloader, model, optimizer
-+ )
-
-  num_epochs = 3
-  num_training_steps = num_epochs * len(train_dataloader)
-  lr_scheduler = get_scheduler(
-      "linear",
-      optimizer=optimizer,
-      num_warmup_steps=0,
-      num_training_steps=num_training_steps
-  )
-
-  progress_bar = tqdm(range(num_training_steps))
-
-  model.train()
-  for epoch in range(num_epochs):
-      for batch in train_dataloader:
--         batch = {k: v.to(device) for k, v in batch.items()}
-          outputs = model(**batch)
-          loss = outputs.loss
--         loss.backward()
-+         accelerator.backward(loss)
-optimizer.step()
-          lr_scheduler.step()
-          optimizer.zero_grad()
-          progress_bar.update(1)
-```
-
-## تدريب
-
-بمجرد إضافة أسطر الكود ذات الصلة، قم بتشغيل التدريب الخاص بك في أحد النصوص أو الدفاتر مثل Colaboratory.
-
-### التدريب باستخدام نص برمجي
-
-إذا كنت تشغل التدريب الخاص بك من نص برمجي، فقم بتشغيل الأمر التالي لإنشاء وحفظ ملف تكوين:
-
-```bash
-accelerate config
-```
-
-ثم قم بتشغيل التدريب الخاص بك باستخدام:
-
-```bash
-accelerate launch train.py
-```
-
-### التدريب باستخدام دفتر ملاحظات
-
-يمكن أيضًا تشغيل 🤗 Accelerate في دفاتر إذا كنت تخطط لاستخدام وحدات معالجة الرسوميات (TPUs) في Colaboratory. قم بتغليف كل الكود المسؤول عن التدريب في دالة، ومررها إلى [`~accelerate.notebook_launcher`]:
-
-```py
->>> from accelerate import notebook_launcher
-
->>> notebook_launcher(training_function)
-```
-
-للحصول على مزيد من المعلومات حول 🤗 Accelerate وميزاته الغنية، يرجى الرجوع إلى [الوثائق](https://huggingface.co/docs/accelerate).

docs/source/ar/attention.md

@@ -1,25 +0,0 @@
-# آليات الانتباه 
-
-تستخدم معظم نماذج المحول (Transformer) الانتباه الكامل بحيث تكون مصفوفة الانتباه ذات الأبعاد المتساوية. ويمكن أن يمثل ذلك عقبة حسابية كبيرة عندما تكون لديك نصوص طويلة. ويعد Longformer وReformer من النماذج التي تحاول أن تكون أكثر كفاءة وتستخدم نسخة مخففة من مصفوفة الانتباه لتسريع التدريب.
-
-## انتباه LSH
-
-يستخدم [Reformer](model_doc/reformer) انتباه LSH. في الدالة softmax(QK^t)، فإن أكبر العناصر فقط (في بعد softmax) من المصفوفة QK^t هي التي ستعطي مساهمات مفيدة. لذلك، بالنسبة لكل استعلام q في Q، يمكننا أن نأخذ في الاعتبار فقط المفاتيح k في K المشابهة لـ q فقط. وتُستخدم دالة هاش لتحديد ما إذا كان q وk متشابهين. ويتم تعديل قناع الانتباه لتجاهل الرمز الحالي (باستثناء الموضع الأول)، لأنه سيعطي استعلامًا ومفتاحًا متساويين (لذلك متشابهين للغاية). نظرًا لطبيعة دالة الهاش العشوائية نوعًا ما، يتم في الممارسة العملية استخدام عدة دوال هاش (يحددها معامل n_rounds) ثم يتم حساب المتوسط معًا.
-
-## الانتباه المحلي
-
-يستخدم [Longformer](model_doc/longformer) الانتباه المحلي: غالبًا ما يكون السياق المحلي (على سبيل المثال، ما هما الرمزان إلى اليسار واليمين؟) كافيًا لاتخاذ إجراء بالنسبة للرمز المعطى. أيضًا، عن طريق تكديس طبقات الانتباه التي لها نافذة صغيرة، سيكون للطبقة الأخيرة مجال استقبال أكبر من مجرد الرموز في النافذة، مما يسمح لها ببناء تمثيل للجملة بأكملها.
-
-كما يتم منح بعض رموز الإدخال المختارة مسبقًا انتباهًا عالميًا: بالنسبة لهذه الرموز القليلة، يمكن لمصفوفة الانتباه الوصول إلى جميع الرموز وتكون هذه العملية متماثلة: فلجميع الرموز الأخرى إمكانية الوصول إلى تلك الرموز المحددة (بالإضافة إلى تلك الموجودة في نافذتهم المحلية). وهذا موضح في الشكل 2d من الورقة، انظر أدناه لمثال على قناع الانتباه:
-
-<div class="flex justify-center">
-    <img scale="50 %" align="center" src="https://huggingface.co/datasets/huggingface/documentation-images/resolve/main/local_attention_mask.png"/>
-</div>
-
-وباستخدام مصفوفات الانتباه هذه التي تحتوي على عدد أقل من المعلمات، يسمح النموذج بمدخالات ذات طول تسلسل أكبر.
-
-## حيل أخرى
-
-### الترميزات الموضعية المحورية
-
-يستخدم [Reformer](model_doc/reformer) ترميزات موضعية محورية: في نماذج المحول التقليدية، يكون الترميز الموضعي E مصفوفة بحجم \\(l\\) في \\(d\\)، حيث \\(l\\) هو طول التسلسل و\\(d\\) هو بعد الحالة المخفية. إذا كان لديك نصوص طويلة جدًا، فقد تكون هذه المصفوفة ضخمة وتستهلك مساحة كبيرة جدًا على وحدة معالجة الرسوميات (GPU). وللتخفيف من ذلك، تتكون الترميزات الموضعية المحورية من تحليل تلك المصفوفة الكبيرة E إلى مصفوفتين أصغر E1 وE2، بأبعاد \\(l_{1} \times d_{1}\\) و \\(l_{2} \times d_{2}\\)، بحيث \\(l_{1} \times l_{2} = l\\) و\\(d_{1} + d_{2} = d\\) (مع حاصل ضرب الأطوال، ينتهي الأمر بكونه أصغر بكثير). ويتم الحصول على الترميز للخطوة الزمنية \\(j\\) في E عن طريق ربط الترميزات للخطوة الزمنية \\(j \% l1\\) في E1 و \\(j // l1\\) في E2.

docs/source/ar/autoclass_tutorial.md

@@ -1,167 +0,0 @@
-# تحميل نماذج مدربة مسبقًا باستخدام AutoClass
-لم ترغب في إنشاء محول معماري لمؤشر الترابط الخاص بك، فهناك العديد من محولات المعمارية المختلفة التي يمكنك الاختيار من بينها. كجزء من الفلسفة الأساسية لـ 🤗 Transformers لجعل المكتبة سهلة وبسيطة ومرنة، فإن فئة `AutoClass` تستدل تلقائيًا وتحمّل البنية الصحيحة من نسخة نموذج (Model Checkpoint) معينة. تسمح لك طريقة `from_pretrained()` بتحميل نموذج مُدرب مسبقًا لأي بنية بسرعة حتى لا تضطر إلى تكريس الوقت والموارد لتدريب نموذج من الصفر. إن إنتاج هذا النوع من التعليمات البرمجية غير المعتمدة على نسخ يعني أنه إذا نجح رمزك مع ننسخة واحدة، فسيتم تشغيله مع أخرى - طالما تم تدريبه لمهمة مماثلة - حتى إذا كانت البنية المعمارية مختلفة.
-
-تذكر أن البنية تشير إلى هيكل النموذج، والنسخ هي الأوزان لبنية معمارية معينة. على سبيل المثال، [BERT](https://huggingface.co/google-bert/bert-base-uncased) هي بنية معمارية، في حين أن `google-bert/bert-base-uncased` هي نسخة. "النموذج" هو مصطلح عام يمكن أن يعني إما البنية أو نالنسخة.
-
-في هذا البرنامج التعليمي، ستتعلم كيفية:
-
-* تحميل مُجزّئ الرموز مُدرب مسبقًا
-* تحميل معالج صور مُدرب مسبقًا
-* تحميل مستخرج ميزات مُدرب مسبقًا
-* تحميل معالج مُدرب مسبقًا
-* تحميل نموذج مُدرب مسبقًا
-* تحميل نموذج كعمود فقري
-
-## AutoTokenizer
-
-تبدأ كل مهمة NLP تقريبًا بمُجزّئ للرموز. يقوم المُجزّئ بتحويل النص إلى شكل يمكن للنموذج معالجته.
-
-قم بتحميل المُجزّئ باستخدام [`AutoTokenizer.from_pretrained`]:
-
-```py
->>> from transformers import AutoTokenizer
-
->>> tokenizer = AutoTokenizer.from_pretrained("google-bert/bert-base-uncased")
-```
-
-ثم قم بتحليل إدخالك على النحو الموضح أدناه:
-
-```py
->>> sequence = "In a hole in the ground there lived a hobbit."
->>> print(tokenizer(sequence))
-{'input_ids': [101, 1999, 1037, 4920, 1999, 1996, 2598, 2045, 2973, 1037, 7570, 10322, 4183, 1012, 102], 
- 'token_type_ids': [0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0], 
- 'attention_mask': [1, 1, 1, 1, 1, 1, 1, 1, 1, 1, 1, 1, 1, 1, 1]}
-```
-
-## معالج الصور التلقائي (AutoImageProcessor)
- 
-
-بالنسبة لمهمات الرؤية، يقوم معالج الصور بمعالجة الصورة إلى تنسيق الإدخال الصحيح.
-
-```py
->>> from transformers import AutoImageProcessor
-
->>> image_processor = AutoImageProcessor.from_pretrained("google/vit-base-patch16-224")
-```
-
-## AutoBackbone
-
-<div style="text-align: center">
-    <img src="https://huggingface.co/datasets/huggingface/documentation-images/resolve/main/transformers/Swin%20Stages.png">
-    <figcaption class="mt-2 text-center text-sm text-gray-500">الصورة توضح مخطط مراحل نموذج Swin.</figcaption>
-</div>
-
-يسمح لك [`AutoBackbone`] باستخدام النماذج المُدربة مسبقًا كعمود فقري للحصول على خرائط ميزات من مراحل مختلفة من العمود الفقري. يجب عليك تحديد أحد المعلمات التالية في [`~PretrainedConfig.from_pretrained`]:
-
-* `out_indices` هو فهرس الطبقة التي تريد الحصول على خريطة الميزات منها
-* `out_features` هو اسم الطبقة التي تريد الحصول على خريطة الميزات منها
-
-يمكن استخدام هذه المعلمات بشكل متبادل، ولكن إذا كنت تستخدم كلاً منها، فتأكد من أنها متوائمة مع بعضها البعض! إذا لم تمرر أيًا من هذه المعلمات، فسيقوم العمود الفقري بإرجاع خريطة الميزات من الطبقة الأخيرة.
-<div style="text-align: center">
-    <img src="https://huggingface.co/datasets/huggingface/documentation-images/resolve/main/transformers/Swin%20Stage%201.png">
-    <figcaption class="mt-2 text-center text-sm text-gray-500">صورة توضح خريطة ميزات من المرحلة الأولى للعمود الفقري.</figcaption>
-</div>
-
-على سبيل المثال، في الرسم التخطيطي أعلاه، لإرجاع خريطة الميزات من المرحلة الأولى من العمود الفقري Swin، يمكنك تعيين `out_indices=(1,)`:
-
-```py
->>> from transformers import AutoImageProcessor, AutoBackbone
->>> import torch
->>> from PIL import Image
->>> import requests
->>> url = "http://images.cocodataset.org/val2017/000000039769.jpg"
->>> image = Image.open(requests.get(url, stream=True).raw)
->>> processor = AutoImageProcessor.from_pretrained("microsoft/swin-tiny-patch4-window7-224")
->>> model = AutoBackbone.from_pretrained("microsoft/swin-tiny-patch4-window7-224", out_indices=(1,))
-
->>> inputs = processor(image, return_tensors="pt")
->>> outputs = model(**inputs)
->>> feature_maps = outputs.feature_maps
-```
-
-الآن يمكنك الوصول إلى كائن `feature_maps` من المرحلة الأولى من العمود الفقري:
-
-```py
->>> list(feature_maps[0].shape)
-[1, 96, 56, 56]
-```
-
-## مستخرج الميزات التلقائي (AutoFeatureExtractor)
-
-بالنسبة للمهام الصوتية، يقوم مستخرج الميزات بمعالجة إشارة الصوت إلى تنسيق الإدخال الصحيح.
-
-قم بتحميل مستخرج ميزات باستخدام [`AutoFeatureExtractor.from_pretrained`]:
-
-```py
->>> from transformers import AutoFeatureExtractor
-
->>> feature_extractor = AutoFeatureExtractor.from_pretrained(
-...     "ehcalabres/wav2vec2-lg-xlsr-en-speech-emotion-recognition"
-... )
-```
-
-## المعالج التلقائي (AutoProcessor)
-
-تتطلب المهام متعددة الوسائط معالجًا يجمع بين نوعين من أدوات المعالجة المسبقة. على سبيل المثال، يتطلب نموذج [LayoutLMV2](model_doc/layoutlmv2) معالج صور لمعالجة الصور ومُجزّئ لمعالجة النص؛ يجمع المعالج كليهما.
-
-قم بتحميل معالج باستخدام [`AutoProcessor.from_pretrained`]:
-
-```py
->>> from transformers import AutoProcessor
-
->>> processor = AutoProcessor.from_pretrained("microsoft/layoutlmv2-base-uncased")
-```
-
-## النموذج التلقائي (AutoModel)
-
-<frameworkcontent>
-<pt>
-تسمح لك فئات `AutoModelFor` بتحميل نموذج مُدرب مسبقًا لمهمة معينة (راجع [هنا](model_doc/auto) للحصول على قائمة كاملة بالمهام المتاحة). على سبيل المثال، قم بتحميل نموذج لتصنيف التسلسل باستخدام [`AutoModelForSequenceClassification.from_pretrained`]:
-
-```py
->>> from transformers import AutoModelForSequenceClassification
-
->>> model = AutoModelForSequenceClassification.from_pretrained("distilbert/distilbert-base-uncased")
-```
-
-أعد استخدام نفس نقطة التفتيش لتحميل بنية لمهمة مختلفة:
-
-```py
->>> from transformers import AutoModelForTokenClassification
-
->>> model = AutoModelForTokenClassification.from_pretrained("distilbert/distilbert-base-uncased")
-```
-
-<Tip warning={true}>
-
-بالنسبة لنماذج PyTorch، تستخدم طريقة `from_pretrained()` `torch.load()` التي تستخدم داخليًا `pickle` والتي يُعرف أنها غير آمنة. بشكل عام، لا تقم مطلقًا بتحميل نموذج قد يكون مصدره مصدرًا غير موثوق به، أو قد يكون تم العبث به. يتم تخفيف هذا الخطر الأمني جزئيًا للنماذج العامة المستضافة على Hub Hugging Face، والتي يتم [فحصها بحثًا عن البرامج الضارة](https://huggingface.co/docs/hub/security-malware) في كل ارتكاب. راجع [توثيق Hub](https://huggingface.co/docs/hub/security) للحصول على أفضل الممارسات مثل [التحقق من التوقيع](https://huggingface.co/docs/hub/security-gpg#signing-commits-with-gpg) باستخدام GPG.
-
-لا تتأثر نقاط تفتيش TensorFlow و Flax، ويمكن تحميلها داخل بنيات PyTorch باستخدام `from_tf` و `from_flax` kwargs لطريقة `from_pretrained` للتحايل على هذه المشكلة.
-
-</Tip>
-
-
-بشكل عام، نوصي باستخدام فئة `AutoTokenizer` وفئة `AutoModelFor` لتحميل مثيلات مُدربة مسبقًا من النماذج. سيساعدك هذا في تحميل البنية الصحيحة في كل مرة. في البرنامج التعليمي التالي، تعرف على كيفية استخدام المحلل اللغوي ومعالج الصور ومستخرج الميزات والمعالج الذي تم تحميله حديثًا لمعالجة مجموعة بيانات للضبط الدقيق.
-</pt>
-
-<tf>
-أخيرًا، تسمح لك فئات `TFAutoModelFor` بتحميل نموذج مُدرب مسبقًا لمهمة معينة (راجع [هنا](model_doc/auto) للحصول على قائمة كاملة بالمهام المتاحة). على سبيل المثال، قم بتحميل نموذج لتصنيف التسلسل باستخدام [`TFAutoModelForSequenceClassification.from_pretrained`]:
-
-```py
->>> from transformers import TFAutoModelForSequenceClassification
-
->>> model = TFAutoModelForSequenceClassification.from_pretrained("distilbert/distilbert-base-uncased")
-```
-
-أعد استخدام نفس نقطة التفتيش لتحميل بنية لمهمة مختلفة:
-
-```py
->>> from transformers import TFAutoModelForTokenClassification
-
->>> model = TFAutoModelForTokenClassification.from_pretrained("distilbert/distilbert-base-uncased")
-```
-
-بشكل عام، نوصي باستخدام فئة `AutoTokenizer` وفئة `TFAutoModelFor` لتحميل نسخ لنماذج مُدربة مسبقًا. سيساعدك هذا في تحميل البنية الصحيحة في كل مرة. في البرنامج التعليمي التالي، ستتعرف على كيفية استخدام المُجزّئ اللغوي ومعالج الصور ومستخرج الميزات والمعالج الذي تم تحميله حديثًا لمعالجة مجموعة بيانات للضبط الدقيق.
-</tf>
-</frameworkcontent>

docs/source/ar/bertology.md

@@ -1,18 +0,0 @@
-# BERTology
-
-يُشهد في الآونة الأخيرة نمو مجال دراسي يُعنى باستكشاف آلية عمل نماذج المحولات الضخمة مثل BERT (والذي يُطلق عليها البعض اسم "BERTology"). ومن الأمثلة البارزة على هذا المجال ما يلي:
-
-- BERT Rediscovers the Classical NLP Pipeline بواسطة Ian Tenney و Dipanjan Das و Ellie Pavlick:
-  https://huggingface.co/papers/1905.05950
-- Are Sixteen Heads Really Better than One? بواسطة Paul Michel و Omer Levy و Graham Neubig: https://huggingface.co/papers/1905.10650
-- What Does BERT Look At? An Analysis of BERT's Attention بواسطة Kevin Clark و Urvashi Khandelwal و Omer Levy و Christopher D.
-  Manning: https://huggingface.co/papers/1906.04341
-- CAT-probing: A Metric-based Approach to Interpret How Pre-trained Models for Programming Language Attend Code Structure: https://huggingface.co/papers/2210.04633
-
-لإثراء هذا المجال الناشئ، قمنا بتضمين بعض الميزات الإضافية في نماذج BERT/GPT/GPT-2 للسماح للناس بالوصول إلى التمثيلات الداخلية، والتي تم تكييفها بشكل أساسي من العمل الرائد لـ Paul Michel (https://huggingface.co/papers/1905.10650):
-
-- الوصول إلى جميع الحالات المخفية في BERT/GPT/GPT-2،
-- الوصول إلى جميع أوزان الانتباه لكل رأس في BERT/GPT/GPT-2،
-- استرجاع قيم ومشتقات  مخرجات الرأس لحساب درجة أهمية الرأس وحذفه كما هو موضح في https://huggingface.co/papers/1905.10650.
-
-ولمساعدتك على فهم واستخدام هذه الميزات بسهولة، أضفنا مثالًا برمجيًا محددًا: [bertology.py](https://github.com/huggingface/transformers-research-projects/tree/main/bertology/run_bertology.py) أثناء استخراج المعلومات  وتقليص من نموذج تم تدريبه مسبقًا على GLUE.