HF RL Explorer

Task: Demonstrate the Effect of Steps-per-Epoch vs Number of Epochs on Model Training

Task: Demonstrate the Effect of Steps-per-Epoch vs Number of Epochs on Model Training: a task in Terminal-Lego-15k (Harbor dataset). Both configurations process the same total number of batches (500 total steps). Your script must demonstrate and document the differences in training behavior…

The task

Both configurations process the same total number of batches (500 total steps). Your script must demonstrate and document the differences in training behavior, including how metrics are recorded, how callbacks (like learning rate schedulers and early stopping) behave, and how the training history granularity differs.

Part of PrimeIntellect/Terminal-Lego-15k.