| { | |
| "version": "1.0", | |
| "description": "Distillation data collected from PPO training", | |
| "fields": { | |
| "sample_id": { | |
| "type": "str", | |
| "description": "Unique sample identifier" | |
| }, | |
| "global_step": { | |
| "type": "int", | |
| "description": "Global training step" | |
| }, | |
| "actor_rank": { | |
| "type": "int", | |
| "description": "Actor process rank" | |
| }, | |
| "env_rank": { | |
| "type": "int", | |
| "description": "Environment rank" | |
| }, | |
| "episode_id": { | |
| "type": "str", | |
| "description": "Episode identifier" | |
| }, | |
| "t": { | |
| "type": "int", | |
| "description": "Time step index (0..T-1)" | |
| }, | |
| "b": { | |
| "type": "int", | |
| "description": "Batch index (0..B-1, where B=rollout_epoch*batch)" | |
| }, | |
| "main_image": { | |
| "type": "ndarray", | |
| "dtype": "uint8", | |
| "shape": "[H, W, C]", | |
| "nullable": true | |
| }, | |
| "wrist_image": { | |
| "type": "ndarray", | |
| "dtype": "uint8", | |
| "nullable": true | |
| }, | |
| "proprio": { | |
| "type": "ndarray", | |
| "dtype": "float32", | |
| "nullable": true | |
| }, | |
| "prompt_text": { | |
| "type": "str", | |
| "nullable": true | |
| }, | |
| "prompt_token_ids": { | |
| "type": "ndarray", | |
| "dtype": "int64", | |
| "nullable": true | |
| }, | |
| "prompt_attention_mask": { | |
| "type": "ndarray", | |
| "dtype": "int64", | |
| "nullable": true | |
| }, | |
| "action_chunk": { | |
| "type": "ndarray", | |
| "dtype": "float32", | |
| "shape": "[5, 7]" | |
| }, | |
| "value_t": { | |
| "type": "float", | |
| "nullable": true | |
| }, | |
| "value_t1": { | |
| "type": "float", | |
| "nullable": true | |
| }, | |
| "delta_v": { | |
| "type": "float", | |
| "nullable": true | |
| }, | |
| "has_valid_delta_v": { | |
| "type": "bool" | |
| }, | |
| "reward_t": { | |
| "type": "float", | |
| "nullable": true | |
| }, | |
| "done_t": { | |
| "type": "bool" | |
| }, | |
| "success": { | |
| "type": "bool" | |
| } | |
| } | |
| } |