fix(mimo): adapt training layer for MCore submodule bump#2979
Merged
yaoyu-33 merged 1 commit intoMar 25, 2026
Conversation
339f237 to
8959b16
Compare
cd3f7fc to
e4d2fdf
Compare
8959b16 to
339d640
Compare
yaoyu-33
approved these changes
Mar 25, 2026
28aa989
into
NVIDIA-NeMo:mimo/phase5-checkpointing-rebuild
2 checks passed
Contributor
|
cross_entropy_loss_fusion=True — Is this required by the new MCore, or is it an independent improvement? It's only added in the checkpoint resume test's _make_language_config(), not in the other e2e tests (test_mimo_training_e2e.py, test_mimo_training_llava.py). Should those tests also get this flag for consistency? |
QA RCCA Analysis1. Fix Reference
2. Root CauseMCore submodule bump required MIMO training layer adaptations:
3. Trigger Configuration
4. Nature of the BugClassification: CODE BUG - MCore API compatibility issues in MIMO training 5. Existing Test CoverageIn Fix PR: YES - 5 test files:
In NMFW Tests: Not specifically covering MIMO MCore compatibility 6. Coverage Assessment
7. New Regression TestNOT NEEDED - Fix PR includes comprehensive unit and e2e tests for MIMO training. 8. ConclusionVerdict: ADEQUATE COVERAGE - Fix PR includes 5 test files covering MIMO training layer. |
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
Sign up for free
to join this conversation on GitHub.
Already have an account?
Sign in to comment
4 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Outdated suggestions cannot be applied.This suggestion has been applied or marked resolved.Suggestions cannot be applied from pending reviews.Suggestions cannot be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
module_output_ndimtoMultiModulePipelineCommunicatorfor correct 2D/3D tensor routing (vision encoders produce 2D[S, H], LLM produces 3D[S, B, H])MIMO_LANGUAGE_MODULE_KEYinstead of removedrole.language_module_nameattribute inmimo_step.pyandtrain_mimo.pylanguage_module_keyassertion frompretrain_mimo.py(removed fromMimoModelConfig)language_module_key/language_module_namereferences in test mocksDepends on: #2978 (phase 4 model layer fixes)
Test plan
torchrun --nproc_per_node=8 tests/e2e/mimo/test_mimo_training_e2e.py)🤖 Generated with Claude Code