Upload ./robotwin_wan_vlm_mask_stage2_15w/checkpoint_step_40000/config.json with huggingface_hub
Browse files
robotwin_wan_vlm_mask_stage2_15w/checkpoint_step_40000/config.json
ADDED
|
@@ -0,0 +1,30 @@
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
| 1 |
+
{
|
| 2 |
+
"common": {
|
| 3 |
+
"action_dim": 14,
|
| 4 |
+
"state_dim": 14,
|
| 5 |
+
"num_video_frames": 8,
|
| 6 |
+
"video_height": 384,
|
| 7 |
+
"video_width": 320,
|
| 8 |
+
"global_downsample_rate": 3,
|
| 9 |
+
"video_action_freq_ratio": 2,
|
| 10 |
+
"action_chunk_size": 16
|
| 11 |
+
},
|
| 12 |
+
"action_expert": {
|
| 13 |
+
"hidden_size": 1024,
|
| 14 |
+
"ffn_dim_multiplier": 4,
|
| 15 |
+
"norm_eps": 1e-05
|
| 16 |
+
},
|
| 17 |
+
"qwen3_expert": {
|
| 18 |
+
"vlm_dim": 2048,
|
| 19 |
+
"head_dim": 128,
|
| 20 |
+
"num_heads": 24,
|
| 21 |
+
"num_layers": 30,
|
| 22 |
+
"norm_eps": 1e-05
|
| 23 |
+
},
|
| 24 |
+
"time_distribution": {
|
| 25 |
+
"timestep_sample_method": "logit_normal",
|
| 26 |
+
"sigmoid_scale": 1.0,
|
| 27 |
+
"min_t": 0.0,
|
| 28 |
+
"max_t": 1.0
|
| 29 |
+
}
|
| 30 |
+
}
|