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

Add MiMo V2.5 Blackwell vision FA4 recipe (#29253)

Browse Source
This commit is contained in:
Xiaoyu Zhang
2026-06-25 13:47:32 +08:00
committed by GitHub
parent 9fd6d0ec99
commit 7c9804ef21
2 changed files with 4 additions and 2 deletions
Download Patch File Download Diff File Expand all files Collapse all files
docs_new/cookbook/autoregressive/Xiaomi/MiMo-V2.5.mdx
+1
View File
@@ -88,6 +88,7 @@ import { MiMoV25Deployment } from '/src/snippets/autoregressive/mimo-v25-deploym
**MiMo-V2.5 (310B):** **MiMo-V2.5 (310B):**
- The checkpoint has a TP=4-interleaved fused `qkv_proj`; attention-TP per DP group **must** be 4. Use `--dp = TP / 4`; for TP > 4 this also requires DP-attention. Total GPUs must be a multiple of 4. A bare `--tp 8` without `--dp 2` will fail to load with `MiMoV2 fused qkv_proj checkpoint is TP=4-interleaved; got attention tp_size=8`. - The checkpoint has a TP=4-interleaved fused `qkv_proj`; attention-TP per DP group **must** be 4. Use `--dp = TP / 4`; for TP > 4 this also requires DP-attention. Total GPUs must be a multiple of 4. A bare `--tp 8` without `--dp 2` will fail to load with `MiMoV2 fused qkv_proj checkpoint is TP=4-interleaved; got attention tp_size=8`.
- Single-node deployments: H100/H200 8× GPUs (`--tp 8 --dp 2`), B200 4× GPUs (`--tp 4`, dp=1, no DP-attn flag needed), GB300 4× GPUs (`--tp 4`, single NVL4 node). FP8 quantization. - Single-node deployments: H100/H200 8× GPUs (`--tp 8 --dp 2`), B200 4× GPUs (`--tp 4`, dp=1, no DP-attn flag needed), GB300 4× GPUs (`--tp 4`, single NVL4 node). FP8 quantization.
- On Blackwell, pass `--mm-attention-backend fa4` for the V2.5 vision encoder. The checkpoint config requests FlashAttention-3 internally, but SGLang rejects FA3 on Blackwell and expects FA4 for multimodal attention.
- `--enable-dp-lm-head` and `--mm-enable-dp-encoder` are required whenever `--enable-dp-attention` is on, to keep LM head and encoder sharding consistent. - `--enable-dp-lm-head` and `--mm-enable-dp-encoder` are required whenever `--enable-dp-attention` is on, to keep LM head and encoder sharding consistent.
- EAGLE MTP uses the checkpoint's MTP weights. Enable with `--speculative-algorithm EAGLE` and `--enable-multi-layer-eagle` (both Hopper and Blackwell). - EAGLE MTP uses the checkpoint's MTP weights. Enable with `--speculative-algorithm EAGLE` and `--enable-multi-layer-eagle` (both Hopper and Blackwell).
- **Multimodal**: Supports image, video, and audio understanding; see Section 4.3 for invocation examples. - **Multimodal**: Supports image, video, and audio understanding; see Section 4.3 for invocation examples.
docs_new/src/snippets/autoregressive/mimo-v25-deployment.jsx
+3 -2
View File
@@ -11,8 +11,8 @@ export const MiMoV25Deployment = () => {
// so attention-TP per DP group must be 4; effective parallelism = TP/DP = 4. // so attention-TP per DP group must be 4; effective parallelism = TP/DP = 4.
// H200 → tp=8, dp=2, single-node, FP8 (verified) // H200 → tp=8, dp=2, single-node, FP8 (verified)
// H100 → tp=8, dp=2, single-node, FP8 // H100 → tp=8, dp=2, single-node, FP8
// B200 → tp=4, dp=1, single-node, FP8 // B200 → tp=4, dp=1, single-node, FP8 (Blackwell: vision fa4)
// GB300 → tp=4, dp=1, single-node, FP8 // GB300 → tp=4, dp=1, single-node, FP8 (Blackwell: vision fa4)
// //
// Optional toggles: // Optional toggles:
// EAGLE MTP — adds --speculative-* flags + SGLANG_ENABLE_SPEC_V2=1. // EAGLE MTP — adds --speculative-* flags + SGLANG_ENABLE_SPEC_V2=1.
@@ -350,6 +350,7 @@ export const MiMoV25Deployment = () => {
flags.push(` --model-loader-extra-config '{"enable_multithread_load": true, "num_threads": 64}'`); flags.push(` --model-loader-extra-config '{"enable_multithread_load": true, "num_threads": 64}'`);
} }
} else { } else {
if (blackwell) flags.push(" --mm-attention-backend fa4");
flags.push(" --mem-fraction-static 0.65"); flags.push(" --mem-fraction-static 0.65");
flags.push(" --chunked-prefill-size 16384"); flags.push(" --chunked-prefill-size 16384");
} }