bxyu-nvidia comments

Results 5 comments of


                                            bxyu-nvidia

Pipeclean NeMo RL training with all environments

Please also check the Huggingface datasets themselves on HF Hub. We also want to fix issues like those in the screenshot below https://huggingface.co/datasets/nvidia/Nemotron-RL-math-OpenMathReasoning I believe the fix for this particular...

Pipeclean NeMo RL training with all environments

For QA, please ensure that all the rows in the HF dataset match what is expected

RL + Gym Rollout logs contains mostly nulls as attribute values

The fix for this is to upload responses_create_params as a string to HF since the apache parquet backing requires non-ragged data structure

single-step unsloth nemo gym notebook

Merged into Unsloth notebooks repo here instead https://github.com/unslothai/notebooks/blob/main/nb/nemo_gym_sudoku.ipynb

Workplace assistant has references to workbench

We should update the HuggingFace dataset description and resources server README with references and citations to the original Workbench https://github.com/olly-styles/WorkBench