Simplify Flash Attention Decode benchmarks generation #437
Conversation
This file contains hidden or bidirectional Unicode text that may be interpreted or compiled differently than what appears below. To review, open the file in an editor that reveals hidden Unicode characters.
Learn more about bidirectional Unicode characters
…tlass-fork into flash_decode_separate_out_configs
…tlass-fork into flash_decode_separate_out_configs
This is needed because today's release does not include the binaries used for CI
This pull splits the example for prefill attention with cachedkv into separate executables. --------- Co-authored-by: Muhammad Tanvir <[email protected]>
…ntel#426) On certain conditions a nightly tag for DPCPP may exist without providing the binary packages for linux. This artifact can be found in the API description for the release tag, under the `assets` section. This PR filters release tags and avoids the ones without the binary package `sycl_linux.tar.gz` It also reverts the workaround to avoid this failure during the day it was discovered. --------- Co-authored-by: Alejandro Acosta <[email protected]>
**Summary** This PR mainly adds the `BF16BF16FP32` CUTE Example on `BMG`. It supports input format with `TT`, `NT` and `TN`. **Test Plan** ``` ./examples/cute/tutorial/cute_tutorial_bmg 1024 128256 4096 'T' 'T' ./examples/cute/tutorial/cute_tutorial_bmg 1024 128256 4096 'N' 'T' ./examples/cute/tutorial/cute_tutorial_bmg 1024 128256 4096 'T' 'N' ``` Co-author with @aacostadiaz @jiyang1011 --------- Co-authored-by: Alejandro Acosta <[email protected]>
…tlass-fork into flash_decode_separate_out_configs
…tlass-fork into flash_decode_separate_out_configs
…tlass-fork into flash_decode_separate_out_configs
…m/muhammad-tanvir-1211/cutlass-fork into flash_decode_separate_out_configs
* Reduce cpu threads used for build
…tlass-fork into flash_decode_separate_out_configs
…tlass-fork into flash_decode_simplify_benchmarks
t4c1
approved these changes
Jun 20, 2025
…tlass-fork into flash_decode_simplify_benchmarks
…tlass-fork into flash_decode_simplify_benchmarks
joeatodd
reviewed
Jun 25, 2025
| @@ -95,7 +95,8 @@ jobs: | |||
| cmake -G Ninja \ | |||
| -DCUTLASS_ENABLE_SYCL=ON \ | |||
| -DDPCPP_SYCL_TARGET=${{ matrix.sycl_target }} \ | |||
| -DCUTLASS_SYCL_RUNNING_CI=ON | |||
| -DCUTLASS_SYCL_RUNNING_CI=ON \ | |||
There was a problem hiding this comment.
CUTLASS_SYCL_RUNNING_CI doesn't seem to do anything as far as I can tell?
| cutlass_test_unit_flash_attention_decode_h96_xe | ||
| cutlass_test_unit_flash_attention_decode_h128_xe | ||
| cutlass_test_unit_flash_attention_decode_h192_xe | ||
| cutlass_test_unit_flash_attention_decode_bf16_fp32_fp32_h64_512_xe |
There was a problem hiding this comment.
You can avoid having to specify all these (& ease maintainability in future) by constructing two lists of dependencies for cutlass_test_unit_flash_attention_decode and test_unit_flash_attention_decode as follows:
At the top of the file define e.g.:
set(TEST_EXES "")
set(TEST_RUNS "")Then last thing in the nested loop:
list(APPEND TEST_EXES ${out_exe})
string(REGEX REPLACE cutlass_ "" out_exe_snipped ${out_exe}) #borrowed from cutlass_test_unit_add_executable
list(APPEND TEST_RUNS ${out_exe_snipped})Then change the add custom targets to:
add_custom_target(
cutlass_test_unit_flash_attention_decode
DEPENDS
${EXES}
)
add_custom_target(
test_unit_flash_attention_decode
DEPENDS
${RUNS}
)
There was a problem hiding this comment.
addressed this change in #408
…tlass-fork into flash_decode_simplify_benchmarks
Sign up for free
to join this conversation on GitHub.
Already have an account?
Sign in to comment
Add this suggestion to a batch that can be applied as a single commit.
This suggestion is invalid because no changes were made to the code.
Suggestions cannot be applied while the pull request is closed.
Suggestions cannot be applied while viewing a subset of changes.
Only one suggestion per line can be applied in a batch.
Add this suggestion to a batch that can be applied as a single commit.
Applying suggestions on deleted lines is not supported.
You must change the existing code in this line in order to create a valid suggestion.
Outdated suggestions cannot be applied.
This suggestion has been applied or marked resolved.
Suggestions cannot be applied from pending reviews.
Suggestions cannot be applied on multi-line comments.
Suggestions cannot be applied while the pull request is queued to merge.
Suggestion cannot be applied right now. Please check back later.
This PR removes a lot of redundant boiler plate code required to generate the benchmarks for Flash Attention Decode.