Add libero_caddy/world_model/config.json
Browse files
libero_caddy/world_model/config.json
ADDED
|
@@ -0,0 +1,53 @@
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
| 1 |
+
{
|
| 2 |
+
"_target_": "stable_worldmodel.wm.lewm.MultiViewLeWM",
|
| 3 |
+
"encoder": {
|
| 4 |
+
"_target_": "stable_pretraining.backbone.utils.vit_hf",
|
| 5 |
+
"size": "tiny",
|
| 6 |
+
"patch_size": 14,
|
| 7 |
+
"image_size": 112,
|
| 8 |
+
"pretrained": false,
|
| 9 |
+
"use_mask_token": false
|
| 10 |
+
},
|
| 11 |
+
"predictor": {
|
| 12 |
+
"_target_": "stable_worldmodel.wm.lewm.module.Predictor",
|
| 13 |
+
"num_frames": 3,
|
| 14 |
+
"input_dim": 768,
|
| 15 |
+
"hidden_dim": 512,
|
| 16 |
+
"output_dim": 768,
|
| 17 |
+
"depth": 7,
|
| 18 |
+
"heads": 8,
|
| 19 |
+
"mlp_dim": 2048,
|
| 20 |
+
"dim_head": 64,
|
| 21 |
+
"dropout": 0.1,
|
| 22 |
+
"emb_dropout": 0.0
|
| 23 |
+
},
|
| 24 |
+
"action_encoder": {
|
| 25 |
+
"_target_": "stable_worldmodel.wm.lewm.module.Embedder",
|
| 26 |
+
"input_dim": 35,
|
| 27 |
+
"emb_dim": 768
|
| 28 |
+
},
|
| 29 |
+
"projector": {
|
| 30 |
+
"_target_": "stable_worldmodel.wm.lewm.module.MLP",
|
| 31 |
+
"input_dim": 192,
|
| 32 |
+
"output_dim": 384,
|
| 33 |
+
"hidden_dim": 2048,
|
| 34 |
+
"norm_fn": {
|
| 35 |
+
"_target_": "torch.nn.BatchNorm1d",
|
| 36 |
+
"_partial_": true
|
| 37 |
+
}
|
| 38 |
+
},
|
| 39 |
+
"pred_proj": {
|
| 40 |
+
"_target_": "stable_worldmodel.wm.lewm.module.MLP",
|
| 41 |
+
"input_dim": 768,
|
| 42 |
+
"output_dim": 768,
|
| 43 |
+
"hidden_dim": 2048,
|
| 44 |
+
"norm_fn": {
|
| 45 |
+
"_target_": "torch.nn.BatchNorm1d",
|
| 46 |
+
"_partial_": true
|
| 47 |
+
}
|
| 48 |
+
},
|
| 49 |
+
"view_keys": [
|
| 50 |
+
"pixels",
|
| 51 |
+
"wrist_pixels"
|
| 52 |
+
]
|
| 53 |
+
}
|