davidheineman/tiny-rl-refined-tasks-v5 download history

davidheineman/tiny-rl-refined-tasks-v5 is a text generation dataset on the Hugging Face Hub. In the last 30 days it was downloaded 24,853 times (24,460 in the last 7 days), and 24,853 times in total. It ranks #1,597 among datasets by monthly downloads.

Tiny RL refined tasks v5 Development study with 16M, 32M and 64M policies. Each initial task has exactly 2,097,152 SFT input prediction tokens followed by 65,536 RL samples. SFT supervision depends on the recipe; masks, weights and exact budgets are recorded in the training profiles.

Open davidheineman/tiny-rl-refined-tasks-v5 on Hugging Face

Sister project: Paper Pulse, the upvote history of every Hugging Face Daily Paper.