minke.yu/sglang
1
0
Fork 0
Code Issues Pull Requests Actions 12 Packages Projects Releases Wiki Activity

update: b300 container for dsv4 (#23697)

Browse Source 
Co-authored-by: Baizhou Zhang <sobereddiezhang@gmail.com>
This commit is contained in:
ishandhanani
2026-04-24 22:59:20 -07:00
committed by GitHub
co-authored by Baizhou Zhang
parent adc59325bc
commit 0d224e5053
2 changed files with 11 additions and 2 deletions
Download Patch File Download Diff File Expand all files Collapse all files
docs_new/cookbook/autoregressive/DeepSeek/DeepSeek-V4.mdx
+4
View File
@@ -80,6 +80,10 @@ Please refer to the [official SGLang installation guide](../../../docs/get-start
</tr>
</thead>
<tbody>
<tr>
<td style={{padding: "9px 12px", fontWeight: 500, backgroundColor: "rgba(255,255,255,0.02)"}}>NVIDIA B300</td>
<td style={{padding: "9px 12px", backgroundColor: "rgba(255,255,255,0.05)"}}><code>lmsysorg/sglang:deepseek-v4-b300</code></td>
</tr>
<tr>
<td style={{padding: "9px 12px", fontWeight: 500, backgroundColor: "rgba(255,255,255,0.02)"}}>NVIDIA B200</td>
<td style={{padding: "9px 12px", backgroundColor: "rgba(255,255,255,0.05)"}}><code>lmsysorg/sglang:deepseek-v4-blackwell</code></td>
docs_new/src/snippets/autoregressive/deepseek-v4-deployment.jsx
+7 -2
View File
@@ -26,6 +26,7 @@ export const DeepSeekV4Deployment = () => {
title: "Hardware Platform",
items: [
{ id: "b200", label: "B200 (FP4)", default: true },
{ id: "b300", label: "B300 (FP4)", default: false },
{ id: "gb300", label: "GB300 (FP4)", default: false },
{ id: "h200", label: "H200 (FP8)", default: false },
],
@@ -222,7 +223,9 @@ export const DeepSeekV4Deployment = () => {
// === SHARED END ===
const generateCommand = () => {
const { hardware, modelSize, recipe, reasoningParser, toolcall } = values;
const { hardware: rawHardware, modelSize, recipe, reasoningParser, toolcall } = values;
// B300 usage is identical to B200 — alias so we don't duplicate every spec entry.
const hardware = rawHardware === "b300" ? "b200" : rawHardware;
const specKey = `${hardware}|${modelSize}`;
const spec = HW_SIZE_SPEC[specKey];
const { slug, tp, multinode, nnodes } = spec;
@@ -439,7 +442,9 @@ export const DeepSeekV4Deployment = () => {
// --max-running-requests 256 only on decode (PD decode can't retract).
// No flashinfer_mxfp4 / autotune-fix / MTP / mem-fraction-static on PD (allinone omits).
// ============================================================================
const buildPDDisaggCommand = (hardware, modelSize) => {
const buildPDDisaggCommand = (rawHardware, modelSize) => {
// B300 usage is identical to B200 — alias so we don't duplicate every spec entry.
const hardware = rawHardware === "b300" ? "b200" : rawHardware;
const specKey = `${hardware}|${modelSize}`;
const { tp: pdTp, multinode, nnodes } = PD_TP_SPEC[specKey];
const slug = HW_SIZE_SPEC[specKey].slug;