Running training

• Num examples = 1000
• Num Epochs = 120
• Instantaneous batch size per device = 64
• Total train batch size = 64
• Gradient Accumulation steps = 1
• Total optimization steps = 1920

Train Output

• global_step = 1920
• train_runtime = 1849.2158
• train_samples_per_second = 64.892
• train_steps_per_second = 1.038
• train_loss = 0.04717305501302083
• epoch = 120.0

Dataset

• edbeeching/decision_transformer_gym_replay
  • halfcheetah-expert-v2