Skip to content

[AMD][ROCm]Minimax-M3 Eagle3 MTP attn backend override for perf #615

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

Merged
xianbaoqian merged 1 commit into from
Jul 7, 2026
Merged

[AMD][ROCm]Minimax-M3 Eagle3 MTP attn backend override for perf #615

Changes from all commits
Commits
Show all changes
1 commit
Select commit
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
6 changes: 6 additions & 0 deletions models/MiniMaxAI/MiniMax-M3.yaml
View file
Open in desktop
Original file line number Diff line number Diff line change
Expand Up @@ -59,6 +59,12 @@ features:
args:
- "--speculative-config"
- '{"method": "eagle3", "model": "Inferact/MiniMax-M3-EAGLE3", "num_speculative_tokens": 3, "attention_backend": "FLASH_ATTN"}'
hardware_overrides:
# On ROCm the Eagle3 draft head prefers the TRITON_ATTN attn backend as it has better performance than the default.
amd:
args:
- "--speculative-config"
- '{"method": "eagle3", "model": "Inferact/MiniMax-M3-EAGLE3", "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