feat(docker): torch 2.13 base + vLLM 0.27.1; bump automodel to latest - #78
Closed
hijkzzz wants to merge 1 commit into
Closed
feat(docker): torch 2.13 base + vLLM 0.27.1; bump automodel to latest#78hijkzzz wants to merge 1 commit into
hijkzzz wants to merge 1 commit into
Conversation
vLLM 0.27.1 hard-pins torch==2.13.0, so move the base to pytorch/pytorch:2.13.0-cuda13.0-cudnn9-devel (CUDA 13.0 = vLLM's default cu130 pairing) and recompile the whole source stack (TE release_v2.14, flash-attn 2.8.3, mamba, DeepEP) against torch 2.13 — all build cleanly (same dep versions AutoModel builds against torch 2.13). vLLM 0.27.1's other pins are unchanged from what we already carry: cutlass-dsl 4.6.0 (quack 0.6.1 still fits), flashinfer-python 0.6.16.post3 (cubin not installed → no version-check clash), tilelang (DSA keeps 0.1.11). nemo-automodel bumped to latest main (e8a163b9); cuda-compat forward-compat layer kept. Verified on an A100: all 17 stack modules import, and mamba / TE / flashinfer / flash-attn kernels run. Co-Authored-By: Claude Opus 4.8 (1M context) <noreply@anthropic.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
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Outdated suggestions cannot be applied.This suggestion has been applied or marked resolved.Suggestions cannot be applied from pending reviews.Suggestions cannot be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.
vLLM 0.27.1 hard-pins torch==2.13.0, so move the base to pytorch/pytorch:2.13.0-cuda13.0-cudnn9-devel (CUDA 13.0 = vLLM's default cu130 pairing) and recompile the whole source stack (TE release_v2.14, flash-attn 2.8.3, mamba, DeepEP) against torch 2.13 — all build cleanly (same dep versions AutoModel builds against torch 2.13). vLLM 0.27.1's other pins are unchanged from what we already carry: cutlass-dsl 4.6.0 (quack 0.6.1 still fits), flashinfer-python 0.6.16.post3 (cubin not installed → no version-check clash), tilelang (DSA keeps 0.1.11). nemo-automodel bumped to latest main (e8a163b9); cuda-compat forward-compat layer kept.
Verified on an A100: all 17 stack modules import, and mamba / TE / flashinfer / flash-attn kernels run.