Load MIMO-trained VLM checkpoints into LLaVAModel for dynamic inference

[mathemakitten]

• I, the PR author, have personally reviewed every line of this PR.

What does this PR do?

Use tools/run_dynamic_text_generation_server.py to serve VLMs trained with MIMO, so image + text evals run on Megatron's own inference engine.

MIMO checkpoints nest each module under its own prefix (language_model.module.module., modality_submodules..module.module.encoders.., ...input_projections.0.*). Instead of building a MIMO model for inference, the server builds a LLaVAModel from the same parts and remaps its sharded state-dict keys onto the checkpoint's names.

New flags:

--vision-num-layers, --vision-hidden-size, --vision-ffn-hidden-size, --vision-num-attention-heads, --vision-num-query-groups, --vision-kv-channels (override the registry sizes of --vision-model-type)- --vision-projection-activation {gelu,fast_gelu} - --image-token-id

Needs #7799, #7752, #7772 to merge.

Issue tracking

For PRs from open-source community contributors:

• New features: a linked issue is required. Please open a feature request and reference it here before submitting the PR.
• Small updates (bug fixes, minor improvements): a linked issue is recommended and will accelerate the PR review process.

Linked issue:

Contribution process

Pre-checks

• I have added relevant unit tests
• I have added relevant functional tests
• If this PR adds or changes a GPU kernel (Triton, jit_fuser/torch.compile, CUDA extension, TE or external-library dispatch, or a scatter/index accumulation), I have added or updated its bit-exact determinism test and registered it in tests/unit_tests/determinism/kernels/manifest.py (guide)
• I have added proper typing to my code Typing guidelines
• I have added relevant documentation
• I have run the autoformatter.sh on my PR

Code review

Feel free to message or comment @NVIDIA/mcore-oncall to help accelerate your merge into main. The less complex your PR is, the faster it will be approved and merged!

All PRs start as draft. If you open a non-draft PR, it will be automatically converted to draft.

Step 1: Mark PR as "Ready for Review"

1. When your PR is ready, click Ready for Review.
2. An oncall reviewer is auto-assigned and expert reviewers are notified based on your changes.
   • Some PRs may jump straight to step 2. This is determined by .github/CODEOWNERS.

⚠️ Only mark as ready once merge-conflicts are resolved and the CI is passing.
Final Review might get declined if these requirements are not fulfilled.

Step 2: Final Review

For PRs that change megatron/core, once all expert reviewers have approved, the Final Review label is applied automatically and final reviewers are assigned.

For PRs outside megatron/core, this step is skipped.

Step 3: Approved

Once all required reviewers have approved, the Approved label is applied automatically.

Merge

Any member of mcore-engineers will be able to merge your PR.

[copy-pr-bot]

Auto-sync is disabled for draft pull requests in this repository. Workflows must be run manually.

Contributors can view more details about this message here.

[wdykas]

In case it helps we do have mimo to llava remapping stuff here https://github.com/NVIDIA/Megatron-LM/pull/7367/changes#diff-585c880dcd23609d8158a7da63e68841621f9630ec33e8614fd22742b1707571R106

[nemo-automation-bot]

🔄 Merge queue validation started!

You can track the progress here: https://github.com/NVIDIA/Megatron-LM/actions/runs/37515456785

[nemo-automation-bot]

🔄 Merge queue validation started!

You can track the progress here: https://github.com/NVIDIA/Megatron-LM/actions/runs/37521041312