diff --git a/docs_new/cookbook/autoregressive/GLM/GLM-5.mdx b/docs_new/cookbook/autoregressive/GLM/GLM-5.mdx index 406e64aa1..b937d71a8 100644 --- a/docs_new/cookbook/autoregressive/GLM/GLM-5.mdx +++ b/docs_new/cookbook/autoregressive/GLM/GLM-5.mdx @@ -109,6 +109,7 @@ sglang serve \ --speculative-num-steps 3 \ --speculative-eagle-topk 1 \ --speculative-num-draft-tokens 4 \ + --enable-flashinfer-allreduce-fusion \ --mem-fraction-static 0.85 \ --host 0.0.0.0 \ --port 30000 diff --git a/docs_new/src/snippets/autoregressive/glm-5-deployment.jsx b/docs_new/src/snippets/autoregressive/glm-5-deployment.jsx index 26f1931ad..e7f8c9550 100644 --- a/docs_new/src/snippets/autoregressive/glm-5-deployment.jsx +++ b/docs_new/src/snippets/autoregressive/glm-5-deployment.jsx @@ -206,6 +206,11 @@ export const GLM5Deployment = () => { cmd += ' \\\n --enable-flashinfer-allreduce-fusion'; } + // H200 FP8: flashinfer allreduce fusion. + if (hardware === 'h200' && effectiveQuant === 'fp8') { + cmd += ' \\\n --enable-flashinfer-allreduce-fusion'; + } + cmd += ` \\\n --mem-fraction-static ${memFraction}`; return cmd; };