NVIDIA / NVIDIA/Model-Optimizer

[Puzzletron] Standalone export fails for heterogeneous distilled checkpoints

Open
#1,848 4 comments 0 reactions 1 assignee View on GitHub

@kevalmorabia97 is already working on this.

Since Jul 5, 2026.

bug investigating torch.pruning
Dominant language
Python
Stars
3.8k
Forks
604
Avg merge
2d 8h
Merged PRs (30d)
142

Description

modelopt branch: main
container: nemo 26.02

The documented command for exporting an intermediate distillation checkpoint fails for heterogeneous Puzzletron models:

uv run python /opt/Megatron-Bridge/examples/conversion/convert_checkpoints.py export
--hf-model <pruned_hf_checkpoint>
--megatron-path <distillation_dir>/checkpoints/iter_0000060
--hf-path <output_hf_checkpoint>

Exporting the final checkpoint inline through distill.py --hf_export_path --student_hf_model works.

The standalone exporter appears not to register Puzzletron’s custom Megatron-Bridge adapters or reproduce the full-model-template export path used by
distill.py.

Contributor guide

Open the contributing guide

First steps

  1. Read the whole issue, then the project's contributing guide.
  2. Comment on the issue to say you are picking it up — it saves two people doing the same work.
  3. Fork the repository and make your change on a branch.
  4. Open a pull request that references the issue number.

Assessment

This issue has not been assessed yet.

Get new issues in your inbox

A short digest of beginner-friendly GitHub issues.