Add support for MiniMax-M2 #42028
Open
Add support for MiniMax-M2 #42028
+1,652
−0
Conversation
This file contains hidden or bidirectional Unicode text that may be interpreted or compiled differently than what appears below. To review, open the file in an editor that reveals hidden Unicode characters.
Learn more about bidirectional Unicode characters
added 6 commits
October 31, 2025 14:17
Signed-off-by: xuebi <xuebi@minimaxi.com>
Signed-off-by: xuebi <xuebi@minimaxi.com>
Signed-off-by: xuebi <xuebi@minimaxi.com>
Signed-off-by: xuebi <xuebi@minimaxi.com>
Signed-off-by: xuebi <xuebi@minimaxi.com>
Signed-off-by: xuebi <xuebi@minimaxi.com>
|
[For maintainers] Suggested jobs to run (before merge) run-slow: auto, minimax_m2 |
Signed-off-by: xuebi <xuebi@minimaxi.com>
molbap
approved these changes
Nov 5, 2025
There was a problem hiding this comment.
Very clean integration, no particular comments. Thank you! cc @Cyrilvallez for core review
|
|
||
| ## Overview | ||
|
|
||
| MiniMax-M2 redefines efficiency for agents. It's a compact, fast, and cost-effective MoE model (230 billion total parameters with 10 billion active parameters) built for elite performance in coding and agentic tasks, all while maintaining powerful general intelligence. With just 10 billion activated parameters, MiniMax-M2 provides the sophisticated, end-to-end tool use performance expected from today's leading models, but in a streamlined form factor that makes deployment and scaling easier than ever. |
There was a problem hiding this comment.
Changed a bit to be more factual
Suggested change
| MiniMax-M2 redefines efficiency for agents. It's a compact, fast, and cost-effective MoE model (230 billion total parameters with 10 billion active parameters) built for elite performance in coding and agentic tasks, all while maintaining powerful general intelligence. With just 10 billion activated parameters, MiniMax-M2 provides the sophisticated, end-to-end tool use performance expected from today's leading models, but in a streamlined form factor that makes deployment and scaling easier than ever. | |
| MiniMax-M2 is a compact, fast, and cost-effective MoE model (230 billion total parameters with 10 billion active parameters) built for elite performance in coding and agentic tasks, all while maintaining powerful general intelligence. With just 10 billion activated parameters, MiniMax-M2 provides the sophisticated, end-to-end tool use performance expected from today's leading models, but in a streamlined form factor that makes deployment and scaling easier than ever. |
Comment on lines
+139
to
+157
| keys_to_ignore_at_inference = ["past_key_values"] | ||
| base_model_tp_plan = { | ||
| "layers.*.self_attn.q_proj": "colwise", | ||
| "layers.*.self_attn.k_proj": "colwise", | ||
| "layers.*.self_attn.v_proj": "colwise", | ||
| "layers.*.self_attn.o_proj": "rowwise", | ||
| "layers.*.block_sparse_moe.gate": "colwise_rep", # we need to replicate here to correctly route experts | ||
| "layers.*.block_sparse_moe.experts.*.w1": "colwise", | ||
| "layers.*.block_sparse_moe.experts.*.w2": "rowwise", | ||
| "layers.*.block_sparse_moe.experts.*.w3": "colwise", | ||
| } | ||
| base_model_pp_plan = { | ||
| "embed_tokens": (["input_ids"], ["inputs_embeds"]), | ||
| "layers": (["hidden_states", "attention_mask"], ["hidden_states"]), | ||
| "norm": (["hidden_states"], ["hidden_states"]), | ||
| } | ||
| attribute_map = { | ||
| "num_experts": "num_local_experts", | ||
| } |
There was a problem hiding this comment.
For further reviews: This part is identical to Minimax1 configuration but inheriting the configuration would require deleting a bunch of keys, like full_attn_beta_factor, and so on, so ok to keep.
Comment on lines
+238
to
+239
| class MiniMaxM2MLP(MixtralMLP): | ||
| pass |
There was a problem hiding this comment.
should be safe to remove
Suggested change
| class MiniMaxM2MLP(MixtralMLP): | |
| pass |
Comment on lines
+347
to
+350
|
|
||
| class MiniMaxM2DecoderLayer(MixtralDecoderLayer): | ||
| pass | ||
|
|
There was a problem hiding this comment.
safe to delete as well
Suggested change
| class MiniMaxM2DecoderLayer(MixtralDecoderLayer): | |
| pass |
Sign up for free
to join this conversation on GitHub.
Already have an account?
Sign in to comment
2 participants
Add this suggestion to a batch that can be applied as a single commit.
This suggestion is invalid because no changes were made to the code.
Suggestions cannot be applied while the pull request is closed.
Suggestions cannot be applied while viewing a subset of changes.
Only one suggestion per line can be applied in a batch.
Add this suggestion to a batch that can be applied as a single commit.
Applying suggestions on deleted lines is not supported.
You must change the existing code in this line in order to create a valid suggestion.
Outdated suggestions cannot be applied.
This suggestion has been applied or marked resolved.
Suggestions cannot be applied from pending reviews.
Suggestions cannot be applied on multi-line comments.
Suggestions cannot be applied while the pull request is queued to merge.
Suggestion cannot be applied right now. Please check back later.
What does this PR do?
This PR adds MiniMax-M2 model to Hugging Face Transformers from MiniMaxAI.
Relevant Links:
Before submitting
Pull Request section?
to it if that's the case.
documentation guidelines, and
here are tips on formatting docstrings.
Who can review?
Anyone in the community is free to review the PR once the tests have passed. Feel free to tag
members/contributors who may be interested in your PR.
@ArthurZucker @Cyrilvallez