Skip to content
Open
Changes from all commits
Commits
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
10 changes: 10 additions & 0 deletions models/MiniMaxAI/MiniMax-M3.yaml
Original file line number Diff line number Diff line change
Expand Up @@ -62,12 +62,22 @@ features:
args:
- "--speculative-config"
- '{"method": "eagle3", "model": "Inferact/MiniMax-M3-EAGLE3", "num_speculative_tokens": 3, "attention_backend": "FLASH_ATTN"}'
hardware_overrides:
amd:
args:
- "--speculative-config"
- '{"method": "eagle3", "model": "Inferact/MiniMax-M3-EAGLE3", "num_speculative_tokens": 3, "attention_backend": "TRITON_ATTN"}'
eagle3_gqa:
label: "GQA"
description: "GQA Eagle3 draft head (Grouped Query Attention, 4 KV heads vs 64) — 16× smaller draft KV cache. Prefer for long-context or large-batch deployments where draft KV memory is the constraint."
args:
- "--speculative-config"
- '{"method": "eagle3", "model": "Inferact/MiniMax-M3-EAGLE3-GQA", "num_speculative_tokens": 3, "attention_backend": "FLASH_ATTN"}'
hardware_overrides:
amd:
args:
- "--speculative-config"
- '{"method": "eagle3", "model": "Inferact/MiniMax-M3-EAGLE3-GQA", "num_speculative_tokens": 3, "attention_backend": "TRITON_ATTN"}'
text_only:
description: "Skip loading the vision encoder for text-only workloads — frees VRAM for KV cache. Mutually exclusive with encoder_parallel."
args:
Expand Down