Further update Deepseek V4 docs (#23617)

This commit is contained in:
fzyzcjy
2026-04-24 13:23:50 +08:00
committed by GitHub
parent 492883c8ca
commit 734e1e2965
@@ -137,12 +137,11 @@ export const DeepSeekV4Deployment = () => {
"b200|big": { slug: "deepseek-ai/DeepSeek-V4-Pro", tp: 8, multinode: false }, "b200|big": { slug: "deepseek-ai/DeepSeek-V4-Pro", tp: 8, multinode: false },
"gb300|small": { slug: "deepseek-ai/DeepSeek-V4-Flash", tp: 4, multinode: false }, "gb300|small": { slug: "deepseek-ai/DeepSeek-V4-Flash", tp: 4, multinode: false },
"gb300|big": { slug: "deepseek-ai/DeepSeek-V4-Pro", tp: 4, multinode: false }, "gb300|big": { slug: "deepseek-ai/DeepSeek-V4-Pro", tp: 4, multinode: false },
// H200 needs a separate FP8-only Instruct ckpt (Flash / Pro public repos // H200 needs an FP8-only Instruct ckpt (deepseek-ai's Flash/Pro repos ship
// ship FP4-mixed weights). That ckpt is still being uploaded, so we emit a // FP4-mixed weights that Hopper can't run). sgl-project publishes FP8
// placeholder that fails loudly on copy-paste instead of silently pulling // repackagings; Flash is public, Pro is still being uploaded.
// the wrong weights. Replace with the real slug once Hopper ckpts are public. "h200|small": { slug: "sgl-project/DeepSeek-V4-Flash-FP8", tp: 4, multinode: false },
"h200|small": { slug: "<TO_BE_UPLOADED_DeepSeek-V4-Flash-hopper>", tp: 4, multinode: false }, "h200|big": { slug: "<TO_BE_UPLOADED_DeepSeek-V4-Pro-FP8>", tp: 16, multinode: true, nnodes: 2 },
"h200|big": { slug: "<TO_BE_UPLOADED_DeepSeek-V4-Pro-hopper>", tp: 16, multinode: true, nnodes: 2 },
}; };
// Per (hardware, modelSize) PD role TP (from allinone _PD_SPEC). // Per (hardware, modelSize) PD role TP (from allinone _PD_SPEC).
const PD_TP_SPEC = { const PD_TP_SPEC = {