Replace has_warp_spec with HAS_AUTO_WS env variable check #205
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
ardaunal
had a problem deploying
to
docker-s3-upload
GitHub Actions
Failure
ardaunal
temporarily deployed
to
docker-s3-upload
GitHub Actions
Inactive
|
Can you please also merge #187 into this PR? |
ardaunal
had a problem deploying
to
docker-s3-upload
GitHub Actions
Failure
|
Does the version of triton that ships with pytorch 2.8 support WS? I have these versions: However, Also, in the triton |
…persistent' into ardau/replace-has-warp-spec-with-has-auto-ws
|
@jeromeku |
ardaunal
had a problem deploying
to
docker-s3-upload
GitHub Actions
Failure
ardaunal
had a problem deploying
to
docker-s3-upload
GitHub Actions
Failure
Summary: - Add a target file for TK attention - Include seq_len_kv in the flash_attention benchmark Reviewed By: devashishshankar, jackiexu1992, xuzhao9 Differential Revision: D73878472 fbshipit-source-id: 2e16199e57aedbf5d58b7ec07310fdbde272df30
Summary: 1. Load gemm/addmm/bmm configs from inductor log and output json and csv to given directory. 2. Load the result json files into input tensors and run with Tritonbench Reviewed By: PaulZhang12 Differential Revision: D73898451 fbshipit-source-id: 6a89fab13dc9eff8058f7f9a5e6616e0bc61b829
Summary: Update the skip test yaml to fix the tests. Reviewed By: FindHao Differential Revision: D73932518 fbshipit-source-id: 4b5c0048eb32590853c20a10167adf3fb921d4b5
Summary: low_mem_dropout has too low tflops and is causing problem: pytorch/test-infra#6594 Remove it from tflops nightly. Pull Request resolved: pytorch-labs#216 Reviewed By: FindHao Differential Revision: D73970729 Pulled By: xuzhao9 fbshipit-source-id: fb2c0ca8061025065a2d40f8015adf4592dbce1a
Summary: Changes needed to run the flash_attention variants on servicelab and extract the data to the dashboard. Reviewed By: minjang Differential Revision: D73976270 fbshipit-source-id: c2da33763123e5d314a41914575eb929a1874867
Summary: Minor change to select the correct shapes depending on the args for b200 fp8_gemm_rowwise_prefill. Reviewed By: minjang Differential Revision: D73976271 fbshipit-source-id: 1b52c0e570554cd4d8774be3725af354cf69c603
Reviewed By: dtolnay Differential Revision: D74066175 fbshipit-source-id: 5ea56fd74a0c9afa45c13b08dcdeb09d3b87c754
Summary: pytorch-labs@d4a1e60 broke the code for OSS, fixing it. Pull Request resolved: pytorch-labs#217 Reviewed By: FindHao Differential Revision: D74087811 Pulled By: xuzhao9 fbshipit-source-id: bc941d586b6211655ae2f28179173450277b8604
Summary: We have fixed ThunderKittens in installation and can enable it in unit test now. Pull Request resolved: pytorch-labs#218 Reviewed By: FindHao Differential Revision: D74107030 Pulled By: xuzhao9 fbshipit-source-id: 19b4c1457e8e6b07d04e269f5aee343c697b78be
Summary: Match the results between two autotune configs. Reviewed By: FindHao Differential Revision: D74085680 fbshipit-source-id: c2778efea6a5e440ac7c9e3e442ce550c4542e65
Summary: Load operators and their kernels from metadata YAML files. This is to generate top-10 kernels and their inputs in data directory. Reviewed By: FindHao Differential Revision: D74132693 fbshipit-source-id: fe10b551947eec0691bdf2cc25a576f3a182bae8
Summary: Refactor `jagged_dense_dense_sum` operator to use input loader from durin data. Reviewed By: FindHao Differential Revision: D74135557 fbshipit-source-id: 30dd098010381d088cc66b49b62340e8e5eabfca
ardaunal
had a problem deploying
to
docker-s3-upload
GitHub Actions
Failure
ardaunal
had a problem deploying
to
docker-s3-upload
GitHub Actions
Failure
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.
triton release/3.3.x branch only supports AutoWS which makes the check
has_warp_spec = hasattr(tl, "async_task")incorrect. Instead this PR adds an environment variable checkHAS_AUTO_WS = os.getenv("ENABLE_AUTO_WS")to replace that.Depends on #187
cc: @xuzhao9