feat: add flash-attn==2.7.4.post1 to backend dependencies#622
Merged
feat: add flash-attn==2.7.4.post1 to backend dependencies#622
Conversation
terrykong
added
CI:L1
Signed-off-by: Terry Kong <terryk@nvidia.com> fix Signed-off-by: Terry Kong <terryk@nvidia.com> fix setuptools Signed-off-by: Terry Kong <terryk@nvidia.com> bug Signed-off-by: Terry Kong <terryk@nvidia.com> unset variables to speed it up Signed-off-by: Terry Kong <terryk@nvidia.com> deal with flash-attn Signed-off-by: Terry Kong <terryk@nvidia.com> 2.8.0.post2 prebuilt wheels! Signed-off-by: Terry Kong <terryk@nvidia.com> Revert "2.8.0.post2 prebuilt wheels!" This reverts commit f945c50. Signed-off-by: Terry Kong <terryk@nvidia.com> revert Signed-off-by: Terry Kong <terryk@nvidia.com> uv.lock Signed-off-by: Terry Kong <terryk@nvidia.com> try wheel Signed-off-by: Terry Kong <terryk@nvidia.com> try whl Signed-off-by: Terry Kong <terryk@nvidia.com> comment Signed-off-by: Terry Kong <terryk@nvidia.com>
Signed-off-by: Terry Kong <terryk@nvidia.com>
Signed-off-by: Terry Kong <terryk@nvidia.com>
Signed-off-by: Terry Kong <terryk@nvidia.com>
Signed-off-by: Sahil Jain <sahilj@nvidia.com>
Signed-off-by: Sahil Jain <sahilj@nvidia.com>
Signed-off-by: Sahil Jain <sahilj@nvidia.com>
terrykong
force-pushed
the
flash-attn-2.7.4
branch
from
de5cd64 to
1531ed4
Compare
terrykong
added
CI:L1
Signed-off-by: Terry Kong <terryk@nvidia.com>
terrykong
added
CI:L1
terrykong
changed the title
Signed-off-by: Terry Kong <terryk@nvidia.com>
Signed-off-by: Terry Kong <terryk@nvidia.com>
Signed-off-by: Terry Kong <terryk@nvidia.com>
Signed-off-by: Terry Kong <terryk@nvidia.com>
Signed-off-by: Terry Kong <terryk@nvidia.com>
setuptools build meta backend Signed-off-by: Terry Kong <terryk@nvidia.com>
terrykong
added
CI:L1
terrykong
added
CI:L1
Signed-off-by: Terry Kong <terryk@nvidia.com>
terrykong
added
CI:L1
terrykong
changed the title
SahilJain314
approved these changes
Jul 8, 2025
jialei777
pushed a commit
to jialei777/nemo-rl
that referenced
this pull request
Jul 23, 2025
…o#622) Signed-off-by: Terry Kong <terryk@nvidia.com> Signed-off-by: Sahil Jain <sahilj@nvidia.com> Co-authored-by: Sahil Jain <sahilj@nvidia.com> Signed-off-by: Jialei Chen <jialeic@google.com>
KiddoZhu
pushed a commit
that referenced
this pull request
Jul 28, 2025
Signed-off-by: Terry Kong <terryk@nvidia.com> Signed-off-by: Sahil Jain <sahilj@nvidia.com> Co-authored-by: Sahil Jain <sahilj@nvidia.com>
FannYYW
pushed a commit
to xxman-google/NeMo-RL
that referenced
this pull request
Aug 5, 2025
…o#622) Signed-off-by: Terry Kong <terryk@nvidia.com> Signed-off-by: Sahil Jain <sahilj@nvidia.com> Co-authored-by: Sahil Jain <sahilj@nvidia.com>
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
Sign up for free
to join this conversation on GitHub.
Already have an account?
Sign in to comment
2 participants
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.
Adding flash-attn to vllm/mcore extras (added an automodel one in preparation for #578 ). Adding it to the core dependencies leads to some headache since linting and anything simple in the CI is painful b/c flash-attn either needs to be built for a GPU runner or it can't be in the case of CPU GIthub hosted runners.
Also added a script and instructions to warm the cache if working outside a container. Dockerfile needed to be changed a little to prefetch flash-attn to avoid cache misses since it's such an expensive dependency to build (45min-1.5hr)
Here's a reference run from #300 (red) and run again with the container from this PR (green)