mirror of
https://github.com/NVIDIA/Model-Optimizer.git
synced 2026-10-02 03:14:52 +08:00
Daniel Korzekwaandcoderabbitai[bot]
4d19ca1a08
### What does this PR do? Create a tool for data blend preparation to enable fast experimentation with distillation ### Usage - examples/researcher_guide/README.md (## Prepare token-budgeted data blends) - examples/dataset/prepare_data_blend.py ### Testing - tests/examples/dataset/test_prepare_data_blend.py - tests/gpu_megatron/torch/utils/plugins/test_megatron_preprocess_data.py ### Before your PR is "*Ready for review*" - Is this change backward compatible?: ✅ - Did you write any new necessary tests?: ✅ - Did you update [Changelog](https://github.com/NVIDIA/Model-Optimizer/blob/main/CHANGELOG.rst)?: ✅ ### Additional Information <!-- E.g. related issue. --> <!-- This is an auto-generated comment: release notes by coderabbit.ai --> ## Summary by CodeRabbit * **New Features** * Added `max_tokens` support for preprocessing to token-cap outputs with safe early stopping and distinct capped-run artifacts (including a matching CLI flag). * Added a YAML-driven workflow to generate weighted Megatron data blends with per-source token allocation and reproducible output metadata. * **Documentation** * Added a researcher fast-experimentation guide for iterative, token-limited evaluation and token-budgeted distillation blends. * Updated dataset-prep and tutorial instructions to recommend `target_tokens`/token-budgeted subsets. * **Tests** * Added CPU/GPU tests covering `max_tokens` stopping behavior, HF streaming caching, and the data-blend YAML workflow. <!-- end of auto-generated comment: release notes by coderabbit.ai --> --------- Signed-off-by: Daniel Korzekwa <dkorzekwa@nvidia.com> Co-authored-by: coderabbitai[bot] <136622811+coderabbitai[bot]@users.noreply.github.com>
Minitron Pruning — End-to-End Tutorials
End-to-end tutorials for Minitron structured pruning followed by knowledge distillation, quantization, evaluation,and vLLM deployment.
Each subdirectory covers a specific source model and target size, including the full data blend, pruning config, distillation hyperparameters, evaluation results, and throughput benchmarks.