[#10245][feat] AutoDeploy: Add Minimax M2 support #10525
Draft
+549
−35
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
bmarimuthu-nv
commented
Jan 8, 2026
| # check if we need to re-combine outputs | ||
| if num_prefill > 0 and num_decode > 0: | ||
| y = torch.empty_like(q) | ||
| y = torch.empty(q.shape, dtype=q.dtype, device=q.device) |
Author
There was a problem hiding this comment.
FakeTensor promises contiguous tensor. But the q can be non-contiguous (based on the model arch/presence of transpose) but end up contiguous at the end of the layer eventually. So we don't inherit the strides this way (and hence contiguous).
Member
There was a problem hiding this comment.
flashinfer doesn't work with non-contiguous inputs. If q is non-contiguous at this point, we need ot make it contiguous before passing it into the kernels!
bmarimuthu-nv
commented
Jan 8, 2026
| assert all( | ||
| s.args[1] == 2 for s in filtered_nodes(linear_node.users, ops=torch.ops.aten.slice) | ||
| ), "Expecting slice nodes to slice tensor over dim=2" | ||
| fused_weight_dims = [s.args[3] - s.args[2] for s in linear_node.users] |
Author
There was a problem hiding this comment.
Issue: slice ops are expected to be direct users of linear_node. But the slice_nodes obtained above can be anywhere downstream of linear.
Fix: Make the check strict - see if slice is directly after linear (as assumed by the logic here)
bmarimuthu-nv
commented
Jan 8, 2026
| canonicalize_graph(gm) | ||
|
|
||
|
|
||
| def _rename_nodes_with_module_hierarchy(gm: fx.GraphModule) -> None: |
Author
There was a problem hiding this comment.
Util to make reading the graph/IR easy. Node names reflect the hierarchy they belong to. Example model_layers_0_self_attn_q_norm_to_9 IR:
%model_layers_0_self_attn_q_proj_torch_linear_simple = auto_deploy.torch_linear_simple.default(%model_layers_0_input_layernorm_mul_3 : s44xs70x3072 : torch.bfloat16, %model_layers_0_self_attn_q_proj_weight : 6144x3072 : torch.bfloat16, None) : s44xs70x6144 : torch.bfloat16
%model_layers_0_self_attn_k_proj_torch_linear_simple_1 = auto_deploy.torch_linear_simple.default(%model_layers_0_input_layernorm_mul_3 : s44xs70x3072 : torch.bfloat16, %model_layers_0_self_attn_k_proj_weight : 1024x3072 : torch.bfloat16, None) : s44xs70x1024 : torch.bfloat16
%model_layers_0_self_attn_v_proj_torch_linear_simple_2 = auto_deploy.torch_linear_simple.default(%model_layers_0_input_layernorm_mul_3 : s44xs70x3072 : torch.bfloat16, %model_layers_0_self_attn_v_proj_weight : 1024x3072 : torch.bfloat16, None) : s44xs70x1024 : torch.bfloat16
%model_layers_0_self_attn_q_norm_to_9 = aten.to.dtype(%model_layers_0_self_attn_q_proj_torch_linear_simple : s44xs70x6144 : torch.bfloat16, torch.float32) : s44xs70x6144 : torch.float32
Member
There was a problem hiding this comment.
do we think this could get too verbose? Otherwise, I think it's a great idea
bmarimuthu-nv
commented
Jan 8, 2026
| self._transform_counter = 0 | ||
| self._dump_dir_initialized = False | ||
|
|
||
| def dump_graph(self, mod: nn.Module, transform_name: str, stage: str) -> None: |
Author
There was a problem hiding this comment.
Dump an SSA-style IR with Dtype annotation (default torch graph doesn't show dtype, shape annotations)
Example:
%model_layers_0_self_attn_q_proj_torch_linear_simple = auto_deploy.torch_linear_simple.default(%model_layers_0_input_layernorm_mul_3 : s44xs70x3072 : torch.bfloat16, %model_layers_0_self_attn_q_proj_weight : 6144x3072 : torch.bfloat16, None) : s44xs70x6144 : torch.bfloat16
%model_layers_0_self_attn_k_proj_torch_linear_simple_1 = auto_deploy.torch_linear_simple.default(%model_layers_0_input_layernorm_mul_3 : s44xs70x3072 : torch.bfloat16, %model_layers_0_self_attn_k_proj_weight : 1024x3072 : torch.bfloat16, None) : s44xs70x1024 : torch.bfloat16
%model_layers_0_self_attn_v_proj_torch_linear_simple_2 = auto_deploy.torch_linear_simple.default(%model_layers_0_input_layernorm_mul_3 : s44xs70x3072 : torch.bfloat16, %model_layers_0_self_attn_v_proj_weight : 1024x3072 : torch.bfloat16, None) : s44xs70x1024 : torch.bfloat16
%model_layers_0_self_attn_q_norm_to_9 = aten.to.dtype(%model_layers_0_self_attn_q_proj_torch_linear_simple : s44xs70x6144 : torch.bfloat16, torch.float32) : s44xs70x6144 : torch.float32
bmarimuthu-nv
changed the title
bmarimuthu-nv
changed the title
bmarimuthu-nv
changed the title
Author
|
@coderabbitai summary |
Contributor
✅ Actions performedSummary regeneration triggered. |
Contributor
📝 WalkthroughWalkthroughThis PR introduces Torch export infrastructure for MiniMax-M2 MoE models by adding hierarchical node renaming, module-level dynamic patching, graph dumping for debugging, and enhanced weight sharding detection. It includes corresponding unit tests for MoE patch validation. Changes
Sequence DiagramssequenceDiagram
participant App as Application
participant Config as Config/Model
participant Patch as Patcher
participant MoE as MoE Module
participant Experts as Expert Weights
participant Output as Output
App->>Config: AutoModelForCausalLM.from_config()
Config->>Patch: get_model_from_config_patched()
Patch->>Config: _from_config_previous()
Config-->>Patch: model instance
Patch->>MoE: iterate modules & patch forward
Patch-->>App: patched model
App->>MoE: forward(hidden_states)
MoE->>MoE: flatten input, compute router_logits
MoE->>MoE: routing_weights = sigmoid(router_logits)
MoE->>MoE: select top_k_experts by scores
MoE->>Experts: extract expert weights
MoE->>MoE: torch.moe(flattened, top_k_weights, experts)
MoE->>Output: reshape & return (hidden_states, router_logits)
sequenceDiagram
participant Export as Export Flow
participant GraphMod as GraphModule
participant Rename as Node Renamer
participant Logger as Graph Logger
participant Dump as Dump File
Export->>GraphMod: transform graph
GraphMod->>Rename: _rename_nodes_with_module_hierarchy()
Rename->>Rename: iterate call_function nodes
Rename->>Rename: extract nn_module_stack metadata
Rename->>Rename: construct hierarchical name
Rename-->>GraphMod: update node.name
GraphMod-->>Logger: transformed module
Logger->>Logger: check AD_DUMP_GRAPHS_DIR env
Logger->>Logger: dump_ssa_with_meta(mod)
Logger->>Dump: write SSA IR with dtype annotations
Dump-->>Export: debug artifacts
Estimated code review effort🎯 4 (Complex) | ⏱️ ~50 minutes ✨ Finishing touches
📜 Recent review detailsConfiguration used: Path: .coderabbit.yaml Review profile: CHILL Plan: Pro 📒 Files selected for processing (8)
Thanks for using CodeRabbit! It's free for OSS, and your support helps us grow. If you like it, consider giving us a shout-out. Comment |
Author
|
/bot run |
Signed-off-by: Balamurugan Marimuthu <[email protected]>
Signed-off-by: Balamurugan Marimuthu <[email protected]>
Signed-off-by: Balamurugan Marimuthu <[email protected]>
Signed-off-by: Balamurugan Marimuthu <[email protected]>
Signed-off-by: Balamurugan Marimuthu <[email protected]>
Signed-off-by: Balamurugan Marimuthu <[email protected]>
Signed-off-by: Balamurugan Marimuthu <[email protected]>
Signed-off-by: Balamurugan Marimuthu <[email protected]>
bmarimuthu-nv
force-pushed
the
bala/minimax-m2
branch
from
a1a1d79 to
41af3fe
Compare
Author
|
/bot run |
Signed-off-by: Balamurugan Marimuthu <[email protected]>
Signed-off-by: Balamurugan Marimuthu <[email protected]>
lucaslie
reviewed
Jan 9, 2026
| # check if we need to re-combine outputs | ||
| if num_prefill > 0 and num_decode > 0: | ||
| y = torch.empty_like(q) | ||
| y = torch.empty(q.shape, dtype=q.dtype, device=q.device) |
Member
There was a problem hiding this comment.
flashinfer doesn't work with non-contiguous inputs. If q is non-contiguous at this point, we need ot make it contiguous before passing it into the kernels!
| canonicalize_graph(gm) | ||
|
|
||
|
|
||
| def _rename_nodes_with_module_hierarchy(gm: fx.GraphModule) -> None: |
Member
There was a problem hiding this comment.
do we think this could get too verbose? Otherwise, I think it's a great idea
Comment on lines
+1906
to
+1910
| def _shard_intermediate_attention_weights( | ||
| layer_subgraph: LayerSubgraph, | ||
| linear_nodes: List[Node], | ||
| transform_container: ShardingTransformContainer, | ||
| ) -> int: |
Member
There was a problem hiding this comment.
can we make this more conservative? Let's just shard weights associated with torch_rmsnorm
| self._transform_counter = 0 | ||
| self._dump_dir_initialized = False | ||
|
|
||
| def dump_graph(self, mod: nn.Module, transform_name: str, stage: str) -> None: |
Member
There was a problem hiding this comment.
how is this different from this feature: https://github.com/NVIDIA/TensorRT-LLM/pull/10117/changes#r2676526174
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.
Summary by CodeRabbit
New Features
Improvements
Tests
✏️ Tip: You can customize this high-level summary in your review settings.
Description
Fixes #10245
AD_DUMP_GRAPHS_DIR=<dir to dump graphs>Test Coverage
PR Checklist
Please review the following before submitting your PR:
PR description clearly explains what and why. If using CodeRabbit's summary, please make sure it makes sense.
PR Follows TRT-LLM CODING GUIDELINES to the best of your knowledge.
Test cases are provided for new code paths (see test instructions)
Any new dependencies have been scanned for license and vulnerabilities
CODEOWNERS updated if ownership changes
Documentation updated as needed
Update tava architecture diagram if there is a significant design change in PR.
The reviewers assigned automatically/manually are appropriate for the PR.
Please check this after reviewing the above items as appropriate for this PR.
GitHub Bot Help
/bot [-h] ['run', 'kill', 'skip', 'reuse-pipeline'] ...Provide a user friendly way for developers to interact with a Jenkins server.
Run
/bot [-h|--help]to print this help message.See details below for each supported subcommand.
Details
run [--reuse-test (optional)pipeline-id --disable-fail-fast --skip-test --stage-list "A10-PyTorch-1, xxx" --gpu-type "A30, H100_PCIe" --test-backend "pytorch, cpp" --add-multi-gpu-test --only-multi-gpu-test --disable-multi-gpu-test --post-merge --extra-stage "H100_PCIe-TensorRT-Post-Merge-1, xxx" --detailed-log --debug(experimental)]Launch build/test pipelines. All previously running jobs will be killed.
--reuse-test (optional)pipeline-id(OPTIONAL) : Allow the new pipeline to reuse build artifacts and skip successful test stages from a specified pipeline or the last pipeline if no pipeline-id is indicated. If the Git commit ID has changed, this option will be always ignored. The DEFAULT behavior of the bot is to reuse build artifacts and successful test results from the last pipeline.--disable-reuse-test(OPTIONAL) : Explicitly prevent the pipeline from reusing build artifacts and skipping successful test stages from a previous pipeline. Ensure that all builds and tests are run regardless of previous successes.--disable-fail-fast(OPTIONAL) : Disable fail fast on build/tests/infra failures.--skip-test(OPTIONAL) : Skip all test stages, but still run build stages, package stages and sanity check stages. Note: Does NOT update GitHub check status.--stage-list "A10-PyTorch-1, xxx"(OPTIONAL) : Only run the specified test stages. Examples: "A10-PyTorch-1, xxx". Note: Does NOT update GitHub check status.--gpu-type "A30, H100_PCIe"(OPTIONAL) : Only run the test stages on the specified GPU types. Examples: "A30, H100_PCIe". Note: Does NOT update GitHub check status.--test-backend "pytorch, cpp"(OPTIONAL) : Skip test stages which don't match the specified backends. Only support [pytorch, cpp, tensorrt, triton]. Examples: "pytorch, cpp" (does not run test stages with tensorrt or triton backend). Note: Does NOT update GitHub pipeline status.--only-multi-gpu-test(OPTIONAL) : Only run the multi-GPU tests. Note: Does NOT update GitHub check status.--disable-multi-gpu-test(OPTIONAL) : Disable the multi-GPU tests. Note: Does NOT update GitHub check status.--add-multi-gpu-test(OPTIONAL) : Force run the multi-GPU tests in addition to running L0 pre-merge pipeline.--post-merge(OPTIONAL) : Run the L0 post-merge pipeline instead of the ordinary L0 pre-merge pipeline.--extra-stage "H100_PCIe-TensorRT-Post-Merge-1, xxx"(OPTIONAL) : Run the ordinary L0 pre-merge pipeline and specified test stages. Examples: --extra-stage "H100_PCIe-TensorRT-Post-Merge-1, xxx".--detailed-log(OPTIONAL) : Enable flushing out all logs to the Jenkins console. This will significantly increase the log volume and may slow down the job.--debug(OPTIONAL) : Experimental feature. Enable access to the CI container for debugging purpose. Note: Specify exactly one stage in thestage-listparameter to access the appropriate container environment. Note: Does NOT update GitHub check status.For guidance on mapping tests to stage names, see
docs/source/reference/ci-overview.mdand the
scripts/test_to_stage_mapping.pyhelper.kill
killKill all running builds associated with pull request.
skip
skip --comment COMMENTSkip testing for latest commit on pull request.
--comment "Reason for skipping build/test"is required. IMPORTANT NOTE: This is dangerous since lack of user care and validation can cause top of tree to break.reuse-pipeline
reuse-pipelineReuse a previous pipeline to validate current commit. This action will also kill all currently running builds associated with the pull request. IMPORTANT NOTE: This is dangerous since lack of user care and validation can cause top of tree to break.