build(docker): install prebuilt FlashAttention-3 wheel#69
Open
Rockdu wants to merge 2 commits into
Open
Conversation
Use the same Miles wheel releases so ltx-core can import the upstream Hopper attention interface in both CUDA variants. Co-authored-by: Cursor <cursoragent@cursor.com>
The justfile's release recipes push radixark/miles-diffusion, a Docker Hub repository that does not exist — the real registry is rockdu/miles_diffusion, built and pushed by docker-build.yml (dispatch + main-push triggers) and pr-test.yml's docker-build job. Neither CI path invokes just, so the file only duplicated the docker build invocation a third time (three recipes drifting against two workflows). The one piece of knowledge it uniquely carried — the cu13 variant build-args — moves to a comment next to the FA3_WHEELS_TAG ARG in the Dockerfile. Co-Authored-By: Claude Fable 5 <noreply@anthropic.com>
Collaborator
Author
|
Pushed 8c4bfbb onto this branch: rather than adding a third copy of the cu13 build-arg wiring to the justfile, the commit deletes |
Rockdu
added
run-ci-e2e
2 tasks
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
1 participant
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Outdated suggestions cannot be applied.This suggestion has been applied or marked resolved.Suggestions cannot be applied from pending reviews.Suggestions cannot be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.
What
flash_attn_interfacemodule imports during the image build.Why
LTX Core imports the upstream FA3 Python interface directly, while the existing SGLang kernels-hub integration does not guarantee that top-level module is installed.
Files
docker/Dockerfile— downloads, installs, and smoke-checks the FA3 wheel.docker/justfile— selects the CUDA 13 wheel release for cu13 builds.Validation
git diff --checkcu129-x86_64andcu130-x86_64Miles wheel releases contain exactly oneflash_attn_3wheel.Checklist
pre-commit run --all-filespasses — not runpytest -xis green — not run; Docker-only changepython3 train.py --helpstill parses — not applicableMade with Cursor
ci-sglang-repo: sgl-project/sglang
ci-sglang-pr: #32420