Skip to content

[Docs] quick fix delete --enable-dp-attention in sgl-jax#24052

Merged
JustinTong0323 merged 2 commits into
sgl-project:mainfrom
JamesBrianD:mimo-2-5-pro-cookbook-fix
Apr 30, 2026
Merged

[Docs] quick fix delete --enable-dp-attention in sgl-jax#24052
JustinTong0323 merged 2 commits into
sgl-project:mainfrom
JamesBrianD:mimo-2-5-pro-cookbook-fix

Conversation

@JamesBrianD
Copy link
Copy Markdown
Contributor

@JamesBrianD JamesBrianD commented Apr 29, 2026

Motivation

In sgl-jax, DP attention is enabled by default and the --enable-dp-attention flag does not exist. The MiMo-V2.5-Pro TPU launch command in the cookbook included this flag, which would cause the server to fail at argument parsing.

Modifications

docs_new/src/snippets/autoregressive/mimo-v25-deployment.jsx: drop --enable-dp-attention from the TPU (sgl-jax) branch of the deployment panel. The CUDA branch is unchanged — the flag is still required for GPU paths.

Checklist

@JamesBrianD JamesBrianD changed the title quick fix delete --enable-dp-attention in sgl-jax [Docs] quick fix delete --enable-dp-attention in sgl-jax Apr 29, 2026
Copy link
Copy Markdown
Contributor

@gemini-code-assist gemini-code-assist Bot left a comment

Choose a reason for hiding this comment

The reason will be displayed to describe this comment to others. Learn more.

Code Review

This pull request updates the MiMoV25Deployment snippet by removing the --enable-dp-attention flag from the deployment command configuration when useDpAttn is enabled. I have no feedback to provide as there were no review comments.

@JustinTong0323 JustinTong0323 self-requested a review April 30, 2026 06:38
Comment thread docs_new/src/snippets/autoregressive/mimo-v25-deployment.jsx
@JustinTong0323 JustinTong0323 merged commit da07b22 into sgl-project:main Apr 30, 2026
42 checks passed
whybeyoung added a commit to whybeyoung/sglang that referenced this pull request Apr 30, 2026
Resolve 5 conflicts in hiradix_cache.py by keeping HEAD's logical-clock
changes (access_time from self.get_access_time()) over main's recently
re-introduced time.monotonic() at the same insert/match/split sites.
This preserves PP-rank deterministic LRU/FIFO eviction.

Resolve unrelated docs_new/.../mimo-v25-deployment.jsx conflict by
taking main (upstream PR sgl-project#24052 removed --enable-dp-attention from the
sgl-jax recipe).
vguduruTT pushed a commit to vguduruTT/sglang that referenced this pull request May 2, 2026
LucQueen pushed a commit to LucQueen/sglang that referenced this pull request May 12, 2026
Sign up for free to join this conversation on GitHub. Already have an account? Sign in to comment

Labels

None yet

Projects

None yet

Development

Successfully merging this pull request may close these issues.

2 participants