Skip to content

Fix MiniMax-M3 MXFP8 crash on 8xH200 (vllm-project/vllm#50856) #750

New issue

Have a question about this project? Sign up for a free GitHub account to open an issue and contact its maintainers and the community.

Sign up for GitHub

By clicking “Sign up for GitHub”, you agree to our terms of service and privacy statement. We’ll occasionally send you account related emails.

Already on GitHub? Sign in to your account

Open
YASHSAWANTTT wants to merge 2 commits into
base: main
Choose a base branch
from
+8 −4
Open

Fix MiniMax-M3 MXFP8 crash on 8xH200 (vllm-project/vllm#50856) #750

Changes from 1 commit
Commits
Show all changes
2 commits
Select commit Hold shift + click to select a range
0866a82
fix(MiniMax-M3): gate indexer_kv_dtype fp8 to Blackwell for MXFP8
YASHSAWANTTT Aug 5, 2026
482e93b
Merge branch 'main' into fix/minimax-m3-mxfp8-hopper
YASHSAWANTTT Aug 7, 2026
File filter

Filter by extension

Filter by extension
Conversations
Failed to load comments.
Loading
Jump to
Jump to file
Failed to load files.
Loading
Diff view
Diff view
12 changes: 8 additions & 4 deletions models/MiniMaxAI/MiniMax-M3.yaml
View file
Open in desktop
Original file line number Diff line number Diff line change
Expand Up @@ -4,7 +4,7 @@ meta:
provider: "MiniMax"
description: "MiniMax M3 vision-language MoE (427B total / 26B active) for frontier coding, agent toolchains, and 1M-token reasoning via MSA sparse attention — native multimodal (image + video + computer use); BF16 plus MXFP8, NVIDIA Blackwell NVFP4, and AMD MI355X MXFP4 variants. Runs on NVIDIA (Hopper/Blackwell) and AMD CDNA4/CDNA3."
date_added: 2026-06-12
date_updated: 2026-07-15
date_updated: 2026-08-04
difficulty: advanced
tasks:
- text
Expand Down Expand Up @@ -118,11 +118,12 @@ variants:
# comfortably, ~4 GPUs for weights alone on Blackwell (B200/B300) or AMD MI350X/MI355X (gfx950)).
vram_minimum_gb: 513
description: "NVIDIA-quantized MXFP8 weights — Blackwell (B200/B300) for native MX tensor cores, and AMD CDNA4 (MI350X/MI355X, gfx950) for native MXFP8 Matrix Cores."
extra_args:
- "--attention_config.indexer_kv_dtype"
- "fp8"
hardware_overrides:
blackwell:
# FP8 index cache needs the SM100 MSA indexer; Triton (Hopper/AMD) is bf16-only.
extra_args:
- "--attention_config.indexer_kv_dtype"
- "fp8"
extra_env:
VLLM_FLOAT32_MATMUL_PRECISION: "high"
h100:
Expand Down Expand Up @@ -745,6 +746,9 @@ guide: |
- **`--block-size` mismatch.** MSA's sparse block size is 128; the vLLM KV
cache block size must match. Using the default (16) misaligns the sparse
attention indexing (on AMD it crashes with `No common block size for 16`).
- **`indexer_kv_dtype=fp8` on Hopper/AMD.** The Triton indexer only supports
bf16. Drop `--attention_config.indexer_kv_dtype fp8` on non-Blackwell GPUs
(the command builder already gates it to Blackwell for the mxfp8 variant).
- **Parsers.** `--tool-call-parser` and `--reasoning-parser` both use
`minimax_m3` — distinct from `minimax_m2` used by earlier releases.
- **Long context KV cache.** See *Context Length & GPU Memory* above — cap
Expand Down