| 1234567891011121314151617181920212223242526272829303132333435363738 |
- engine = "vllm"
- name = "MiniMax M3 MXFP8 (vLLM)"
- docker_image = "vllm/vllm-openai:minimax-m3"
- model = "MiniMaxAI/MiniMax-M3-MXFP8"
- parent_model = "MiniMaxAI/MiniMax-M3"
- required_nodes = 4
- tensor_parallel_size = 4
- pipeline_parallel_size = 1
- distributed_backend = "mp"
- # MXFP8 quantized MiniMax M3 (~440B params, ~23B active).
- # 4 nodes × 1 GPU (DGX Spark GB10 Blackwell), TP=4 across nodes.
- # MSA sparse attention requires --block-size 128.
- # Text-only mode skips the vision encoder to save VRAM.
- #entrypoint = ["vllm", "serve"]
- args = [
- "--block-size", "128",
- "--kv-cache-dtype", "fp8",
- "--gpu-memory-utilization", "0.85",
- "--load-format", "safetensors",
- "--enable-prefix-caching",
- "--tool-call-parser", "minimax_m3",
- "--reasoning-parser", "minimax_m3",
- "--enable-auto-tool-choice",
- "--language-model-only",
- "--enable-expert-parallel",
- "--trust-remote-code",
- "--served-model-name", "minimax-m3", "default",
- ]
- [env]
- # M3 supports thinking via chat_template_kwargs.thinking_mode
- # enabled / adaptive / disabled — pass per-request in extra_body
- SAFETENSORS_FAST_GPU = "1"
- VLLM_ALLOW_LONG_MAX_MODEL_LEN = "1"
- VLLM_USE_B12X_MOE = "1"
|
