| { | |
| "version": 1, | |
| "common": { | |
| "max_seq_len": 8192, | |
| "video_size": [ | |
| 256, | |
| 448 | |
| ], | |
| "pool_enable": true, | |
| "pool_spatial": 2, | |
| "pool_mode": "avg", | |
| "pool_stage": "post_vit", | |
| "native_resolution": false, | |
| "native_max_pixels": 1048576 | |
| }, | |
| "tasks": { | |
| "vlnce": { | |
| "num_history_frames": 64, | |
| "predict_horizon": 10, | |
| "action_tokenizer": { | |
| "method": "rvq", | |
| "bundle_path": "action_tokenizer" | |
| }, | |
| "slowfast_tiers": [ | |
| { | |
| "name": "current", | |
| "age_lo": 0, | |
| "age_hi": 1, | |
| "mode": "dense", | |
| "pool_spatial": 1 | |
| }, | |
| { | |
| "name": "fast", | |
| "age_lo": 2, | |
| "age_hi": 33, | |
| "mode": "dense", | |
| "pool_spatial": 2 | |
| }, | |
| { | |
| "name": "mid", | |
| "age_lo": 34, | |
| "age_hi": 89, | |
| "mode": "burst", | |
| "pair_stride": 6, | |
| "pool_spatial": 2 | |
| }, | |
| { | |
| "name": "long", | |
| "age_lo": 90, | |
| "age_hi": 1000000, | |
| "mode": "span", | |
| "num_pairs": 14, | |
| "pool_spatial": 4 | |
| }, | |
| { | |
| "name": "anchor", | |
| "age_lo": 0, | |
| "age_hi": 0, | |
| "mode": "anchor", | |
| "num_frames": 2, | |
| "pool_spatial": 4 | |
| } | |
| ], | |
| "prompt_style": "unified_traj", | |
| "timestamp_relative": true, | |
| "video_fps": 4 | |
| }, | |
| "trackvla": { | |
| "num_history_frames": 64, | |
| "predict_horizon": 10, | |
| "action_tokenizer": { | |
| "method": "rvq", | |
| "bundle_path": "action_tokenizer" | |
| }, | |
| "slowfast_tiers": [ | |
| { | |
| "name": "current", | |
| "age_lo": 0, | |
| "age_hi": 1, | |
| "mode": "dense", | |
| "pool_spatial": 1 | |
| }, | |
| { | |
| "name": "fast", | |
| "age_lo": 2, | |
| "age_hi": 33, | |
| "mode": "dense", | |
| "pool_spatial": 2 | |
| }, | |
| { | |
| "name": "mid", | |
| "age_lo": 34, | |
| "age_hi": 89, | |
| "mode": "burst", | |
| "pair_stride": 6, | |
| "pool_spatial": 2 | |
| }, | |
| { | |
| "name": "long", | |
| "age_lo": 90, | |
| "age_hi": 1000000, | |
| "mode": "span", | |
| "num_pairs": 14, | |
| "pool_spatial": 4 | |
| }, | |
| { | |
| "name": "anchor", | |
| "age_lo": 0, | |
| "age_hi": 0, | |
| "mode": "anchor", | |
| "num_frames": 2, | |
| "pool_spatial": 4 | |
| } | |
| ], | |
| "prompt_style": "unified_traj", | |
| "timestamp_relative": true, | |
| "video_fps": 4 | |
| } | |
| } | |
| } |