diff --git a/docs_new/src/snippets/autoregressive/deepseek-v4-deployment.jsx b/docs_new/src/snippets/autoregressive/deepseek-v4-deployment.jsx index 95c6eeb5d..080a8e7ec 100644 --- a/docs_new/src/snippets/autoregressive/deepseek-v4-deployment.jsx +++ b/docs_new/src/snippets/autoregressive/deepseek-v4-deployment.jsx @@ -391,6 +391,9 @@ export const DeepSeekV4Deployment = () => { } } else if (recipe === "max-throughput") { if (hardware === "h200") { + if (!isBig) { + recipeEnv.push("SGLANG_JIT_DEEPGEMM_PRECOMPILE=0"); + } recipeEnv.push(isBig ? "SGLANG_DEEPEP_NUM_MAX_DISPATCH_TOKENS_PER_RANK=128" : "SGLANG_DEEPEP_NUM_MAX_DISPATCH_TOKENS_PER_RANK=256");