| { | |
| "common": { | |
| "action_dim": 14, | |
| "state_dim": 14, | |
| "num_video_frames": 8, | |
| "video_height": 384, | |
| "video_width": 320, | |
| "global_downsample_rate": 3, | |
| "video_action_freq_ratio": 2 | |
| }, | |
| "action_expert": { | |
| "hidden_size": 1024, | |
| "ffn_dim_multiplier": 4, | |
| "norm_eps": 1e-05 | |
| }, | |
| "und_expert": { | |
| "hidden_size": 512, | |
| "ffn_dim_multiplier": 4, | |
| "norm_eps": 1e-05, | |
| "vlm": { | |
| "input_dim": 2048, | |
| "projector_type": "mlp3x_silu" | |
| } | |
| }, | |
| "time_distribution": { | |
| "timestep_sample_method": "logit_normal", | |
| "sigmoid_scale": 1.0, | |
| "min_t": 0.0, | |
| "max_t": 1.0 | |
| }, | |
| "ema": { | |
| "enabled": false, | |
| "update_after_step": 0, | |
| "inv_gamma": 1.0, | |
| "power": 0.75, | |
| "min_value": 0.0, | |
| "max_value": 0.9999 | |
| } | |
| } |