HF RL Explorer

Evaluates the ability to reverse-engineer a PyTorch Transformer model architecture from a state dictionary…

Evaluates the ability to reverse-engineer a PyTorch Transformer model architecture from a state dictionary…: a task in Terminal-Bench 2.1 (Harbor git-repos dataset) (Harbor dataset). Success Criteria: - The TorchScript model at /app/model.pt must be able to load the original weights from…

The task

Success Criteria: - The TorchScript model at /app/model.pt must be able to load the original weights from /app/weights.pt with no errors. - The only difference between the state dicts of /app/model.pt and /app/weights.pt should be in the weights of the output_layer. - The MSE loss using the updated output_layer must…

Part of harborframework/terminal-bench-2.1.