Skip to content

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

New issue

Have a question about this project? Sign up for a free GitHub account to open an issue and contact its maintainers and the community.

Sign up for GitHub

By clicking “Sign up for GitHub”, you agree to our terms of service and privacy statement. We’ll occasionally send you account related emails.

Already on GitHub? Sign in to your account

Merged
JustinTong0323 merged 2 commits into from
Apr 30, 2026
Merged

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

Changes from all commits
Commits
Show all changes
2 commits
Select commit Hold shift + click to select a range
4b28496
quick fix delete --enable-dp-attention in sgl-jax
JamesBrianD Apr 29, 2026
8ec9571
add comments
JamesBrianD Apr 30, 2026
File filter

Filter by extension

Filter by extension
Conversations
Failed to load comments.
Loading
Jump to
Jump to file
Failed to load files.
Loading
Diff view
Diff view
8 changes: 7 additions & 1 deletion docs_new/src/snippets/autoregressive/mimo-v25-deployment.jsx
View file
Open in desktop
Original file line number Diff line number Diff line change
Expand Up @@ -230,6 +230,12 @@ export const MiMoV25Deployment = () => {
// Recipe sources:
// v7x: tp=ep=32, dp=4, omits --attention-backend, mem-frac 0.95, swa 0.25
// v6e: tp=ep=64, dp=8, --attention-backend fa, mem-frac 0.92, swa 0.15
//
// sgl-jax conventions:
// - `--tp-size` is always the total JAX device count; per-DP TP is
// derived automatically as tp/dp.
// - No `--enable-dp-attention` flag — DP attention is the default
// (FFN layers auto-pick EP-split for MoE, attn-TP-split for dense).
const isV7x = hardware === "tpu-v7x";
const useEp = expertParallelism === "enabled";
const useDpAttn = dpAttention === "enabled";
Expand All @@ -239,7 +245,7 @@ export const MiMoV25Deployment = () => {
flags.push(" --trust-remote-code");
flags.push(` --tp-size ${tp}`);
if (useEp) flags.push(` --ep-size ${tp}`);
if (useDpAttn) flags.push(` --dp-size ${dpSize}`, " --enable-dp-attention");
if (useDpAttn) flags.push(` --dp-size ${dpSize}`);
Comment thread
JamesBrianD marked this conversation as resolved.
flags.push(" --moe-backend fused");
if (!isV7x) flags.push(" --attention-backend fa");
flags.push(" --host 0.0.0.0");
Expand Down
Loading