This is a minor typo fix in `test/test_transformers.py`:
- Renamed `intial_query_grad` to `initial_query_grad` for improved clarity and correctness in test variable naming.
There are **no functional or logic changes** — this PR is aimed purely at improving readability and maintaining code quality.
Thanks to the PyTorch team for their work and review time
Please feel free to suggest if this needs any adjustment.
Pull Request resolved: https://github.com/pytorch/pytorch/pull/157306
Approved by: https://github.com/Skylion007
Notable new features/optimizations for SDPA operators on AMD systems from AOTriton 0.10b:
* Official support of gfx950/gfx1201
* Experimental support of gfx1101/gfx1151/gfx1150/gfx1200
* Reduce libaotriton.so binary size by over 80%.
+ Without this optimization the binary size of `libaotriton.so` could be
over 100MiB due to 2x more supported architectures compared with 0.9b.
Now it is only about 11MiB.
* Support sliding window attention (SWA) in
`_flash_attention_forward/backward`. Should fix#154582
See https://github.com/ROCm/aotriton/releases/tag/0.10b for full details,
including Known Problems.
Notable changes to SDPA backend:
* `std::optional<int64_t>` `window_size_left/right` are directly passed to
ROCM's SDPA backend, because the default value `-1` is meaningful to
AOTriton's backend and bottom-right aligned causal mask is implemented with
negative `window_size_left/right`
* Some code clean up around `USE_CK_FLASH_ATTENTION`
Pull Request resolved: https://github.com/pytorch/pytorch/pull/156499
Approved by: https://github.com/jeffdaily, https://github.com/jithunnair-amd
Notable new features/optimizations for SDPA operators on AMD systems from AOTriton 0.10b:
* Official support of gfx950/gfx1201
* Experimental support of gfx1101/gfx1151/gfx1150/gfx1200
* Reduce libaotriton.so binary size by over 80%.
+ Without this optimization the binary size of `libaotriton.so` could be
over 100MiB due to 2x more supported architectures compared with 0.9b.
Now it is only about 11MiB.
* Support sliding window attention (SWA) in
`_flash_attention_forward/backward`. Should fix#154582
See https://github.com/ROCm/aotriton/releases/tag/0.10b for full details,
including Known Problems.
Notable changes to SDPA backend:
* `std::optional<int64_t>` `window_size_left/right` are directly passed to
ROCM's SDPA backend, because the default value `-1` is meaningful to
AOTriton's backend and bottom-right aligned causal mask is implemented with
negative `window_size_left/right`
* Some code clean up around `USE_CK_FLASH_ATTENTION`
Pull Request resolved: https://github.com/pytorch/pytorch/pull/156290
Approved by: https://github.com/jithunnair-amd, https://github.com/jeffdaily
In OneDNN v3.7, SDPA has below defects:
1. The dtype of intermediate value is the same as QKV, while Pytorch uses FP32 dtype for intermediate value to make sure better accuracy.
2. Only support headdim size <= 256.
3. Don't support implict causal mask when QKV is FP32. We need to build an attention mask explicitly with aten ops.
In OneDNN v3.8, they have update for these defects. Since these are tiny changes, I decided to put them in single PR.
Pull Request resolved: https://github.com/pytorch/pytorch/pull/152091
Approved by: https://github.com/EikanWang, https://github.com/guangyey, https://github.com/drisspg
This PR adds support for `sm_121` of the DGX Spark. The `sm_121` is binary compatible with `sm_120` (just like `sm_89` and `sm_86`), therefore a compilation targeting `sm_121` is not required.
Pull Request resolved: https://github.com/pytorch/pytorch/pull/152314
Approved by: https://github.com/eqy
Notable new features/optimizations for SDPA operators on AMD systems from AOTriton 0.9b:
* Optimize these Non-power-of-two head dimensions: 48, 80, 96, 160, 192, 224. Inputs with these head dimensions do not need padding to power-of-two anymore.
* `is_causal=True` cases are now supported with persistent dynamic algorithm, which requires an atomic tensor but does load balance between different CTAs
* `dropout_p > 0.0` cases now support full 64-bit offsets and use all i64x4 PRNG outputs
* The precise AOTriton shared library version can now be identified with `readelf -p .comment libaotriton_v2.so`
+ However, this does not guarantee the GPU images stored under `aotriton.images` have the same version, since they can be overwritten.
* The newly added fused backward kernel will be used for smaller workloads, due to less kernel invocation overhead.
* Support gfx1201 (RX 9070XT). Need to be enabled at runtime with `TORCH_ROCM_AOTRITON_ENABLE_EXPERIMENTAL=1`
Pull Request resolved: https://github.com/pytorch/pytorch/pull/148433
Approved by: https://github.com/jeffdaily
Notable new features/optimizations for SDPA operators on AMD systems from AOTriton 0.9b:
* Optimize these Non-power-of-two head dimensions: 48, 80, 96, 160, 192, 224. Inputs with these head dimensions do not need padding to power-of-two anymore.
* `is_causal=True` cases are now supported with persistent dynamic algorithm, which requires an atomic tensor but does load balance between different CTAs
* `dropout_p > 0.0` cases now support full 64-bit offsets and use all i64x4 PRNG outputs
* The precise AOTriton shared library version can now be identified with `readelf -p .comment libaotriton_v2.so`
+ However, this does not guarantee the GPU images stored under `aotriton.images` have the same version, since they can be overwritten.
* The newly added fused backward kernel will be used for smaller workloads, due to less kernel invocation overhead.
* Support gfx1201 (RX 9070XT). Need to be enabled at runtime with `TORCH_ROCM_AOTRITON_ENABLE_EXPERIMENTAL=1`
Pull Request resolved: https://github.com/pytorch/pytorch/pull/148433
Approved by: https://github.com/jeffdaily
Disabled by default for now behind `TORCH_CUDNN_SDPA_NESTED_TENSOR_ENABLED=1`
Just wanted to get this out before starting a series of SDPA cleanup PRs---the biggest thing is we don't need the boilerplate around all of the `build_graph_and_tensors*` functions anymore as we can now use the `UID`-style referencing of tensor nodes as was done for the Conv-V8 API backend.
CC @drisspg
Pull Request resolved: https://github.com/pytorch/pytorch/pull/141178
Approved by: https://github.com/jbschlosser
Motivation
===
This PR is part of the plan of OneDNN Upstreaming, as #114848 [(comment)](https://github.com/pytorch/pytorch/issues/114848#issuecomment-2451553203) stated. The support of SDPA is via the overridable variance on XPU backend. Beside the added `Attention.cpp` file, `Graph.h` is added to hold utils for OneDNN graph including those for kernel/compile graph caching. In addition, a selection of testcases in `test/test_transformers.py` are copied into the new `test/xpu/test_transformers.py` and modified accordingly to provide additional tests beyond `./third_party/torch-xpu-ops/test/xpu/test_ops_xpu.py`.
Depends on OneDNN version v3.7 upgrade in #147498
Depends on BUILD_GRAPH switch in #147608
Pull Request resolved: https://github.com/pytorch/pytorch/pull/147614
Approved by: https://github.com/jansel, https://github.com/EikanWang
Split test_transformers.py into test_transformers.py and test_transformers_privateuser1.py. Currently the privateuse1 test cases in test_transformers.py are skipped since they conflict with cuda test cases.
Pull Request resolved: https://github.com/pytorch/pytorch/pull/147441
Approved by: https://github.com/drisspg
Disabled by default for now behind `TORCH_CUDNN_SDPA_NESTED_TENSOR_ENABLED=1`
Just wanted to get this out before starting a series of SDPA cleanup PRs---the biggest thing is we don't need the boilerplate around all of the `build_graph_and_tensors*` functions anymore as we can now use the `UID`-style referencing of tensor nodes as was done for the Conv-V8 API backend.
CC @drisspg
Pull Request resolved: https://github.com/pytorch/pytorch/pull/141178
Approved by: https://github.com/jbschlosser
Replace https://github.com/pytorch/pytorch/pull/138947 for re-import.
Replaces https://github.com/ROCm/pytorch/pull/1592
This PR contains the initial implementation of SDPA with composable_kernel backend. The CK path can be forced by simply calling torch.backends.cuda.preferred_rocm_fa_library("ck"). Similarly, you can force the incumbent aotriton implementation by passing in "aotriton" or "default". As you'd expect, not setting this option will result in aotriton to be used as the backend. In the case of CK, if pytorch deems flash attention usable, then it will use the CK path in all the same places aotriton would have been used. This PR makes no changes to the heuristics which select which attention scheme to use (i.e. flash attention vs memory efficient attention vs math etc etc). It only gets called when flash attention is both enabled (via USE_FLASH_ATTENTION) and is selected at runtime by the existing heuristics.
Files located in pytorch/aten/src/ATen/native/transformers/hip/flash_attn/ck/mha* have been pulled from https://github.com/Dao-AILab/flash-attention courtesy of @tridao's hard work who is the co-author
NOTE: In order to use this backend, the user MUST set USE_CK_FLASH_ATTENTION=1 in their environment when they build PyTorch.
Pull Request resolved: https://github.com/pytorch/pytorch/pull/143695
Approved by: https://github.com/malfet
Co-authored-by: Andy Lugo <Andy.LugoReyes@amd.com>
Co-authored-by: Jithun Nair <jithun.nair@amd.com>
Replaces https://github.com/ROCm/pytorch/pull/1592
This PR contains the initial implementation of SDPA with composable_kernel backend. The CK path can be forced by simply calling `torch.backends.cuda.preferred_rocm_fa_library("ck")`. Similarly, you can force the incumbent aotriton implementation by passing in "aotriton" or "default". As you'd expect, not setting this option will result in aotriton to be used as the backend. In the case of CK, if pytorch deems flash attention usable, then it will use the CK path in all the same places aotriton would have been used. This PR makes no changes to the heuristics which select which attention scheme to use (i.e. flash attention vs memory efficient attention vs math etc etc). It only gets called when flash attention is both enabled (via `USE_FLASH_ATTENTION`) and is selected at runtime by the existing heuristics.
Files located in pytorch/aten/src/ATen/native/transformers/hip/flash_attn/ck/mha* have been pulled from https://github.com/Dao-AILab/flash-attention courtesy of @tridao's hard work who is the co-author
NOTE: In order to use this backend, the user MUST set USE_CK_FLASH_ATTENTION=1 in their environment when they build PyTorch.
Pull Request resolved: https://github.com/pytorch/pytorch/pull/138947
Approved by: https://github.com/pruthvistony, https://github.com/xw285cornell, https://github.com/leitian
Co-authored-by: Xiaodong Wang <xw285@cornell.edu>
Notable new features for SDPA operators on AMD systems from AOTriton 0.8b:
1. Nestedtensor support;
2. MQA/GQA support;
3. Restore Efficient attention support for causal=True and seqlen_q != seqlen_k cases;
+ The kernel should use top-left alignment, bottom right alignment will be added later
4. Move gfx1100 (RX7900/W7800/W7900) out of experimental support status.
However, users are strongly recommended to update to ROCM 6.2.4, notably for
its firmware updates.
Related unit tests are enabled as well.
Notable related changes from AOTriton 0.8b:
1. AOTriton 0.8b moves the GPU kernel out of libaotriton.so to a separate directory `aotriton.images`;
2. LZMA replaces ZSTD as GPU kernel compression algorithm for better compression ratio: aotriton0.8b (.so + aotriton.images take 350MB) compared to aotriton0.7b .so: 800MB
3. The compression cannot be disabled now, and `liblzma` is hard run-time dependency.
+ Should not be a problem, since `lzma` is part of Python Standard Library
Pull Request resolved: https://github.com/pytorch/pytorch/pull/140172
Approved by: https://github.com/jithunnair-amd, https://github.com/jeffdaily
Co-authored-by: Jithun Nair <37884920+jithunnair-amd@users.noreply.github.com>
