Upload config.json with huggingface_hub
Browse files- config.json +49 -0
config.json
ADDED
|
@@ -0,0 +1,49 @@
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
| 1 |
+
{
|
| 2 |
+
"action_dim": 2,
|
| 3 |
+
"ema": {
|
| 4 |
+
"inv_gamma": 1.0,
|
| 5 |
+
"max_value": 0.9999,
|
| 6 |
+
"min_value": 0.0,
|
| 7 |
+
"power": 0.75,
|
| 8 |
+
"update_after_step": 0
|
| 9 |
+
},
|
| 10 |
+
"img_adaptor": "mlp2x_gelu",
|
| 11 |
+
"img_cond_len": 1458,
|
| 12 |
+
"img_pos_embed_config": [
|
| 13 |
+
[
|
| 14 |
+
"image",
|
| 15 |
+
[
|
| 16 |
+
2,
|
| 17 |
+
1,
|
| 18 |
+
-729
|
| 19 |
+
]
|
| 20 |
+
]
|
| 21 |
+
],
|
| 22 |
+
"img_token_dim": 1152,
|
| 23 |
+
"lang_adaptor": "mlp2x_gelu",
|
| 24 |
+
"lang_pos_embed_config": [
|
| 25 |
+
[
|
| 26 |
+
"lang",
|
| 27 |
+
-1024
|
| 28 |
+
]
|
| 29 |
+
],
|
| 30 |
+
"lang_token_dim": 3584,
|
| 31 |
+
"max_lang_cond_len": 1024,
|
| 32 |
+
"noise_scheduler": {
|
| 33 |
+
"beta_schedule": "squaredcos_cap_v2",
|
| 34 |
+
"clip_sample": false,
|
| 35 |
+
"num_inference_timesteps": 5,
|
| 36 |
+
"num_train_timesteps": 1000,
|
| 37 |
+
"prediction_type": "sample",
|
| 38 |
+
"type": "ddpm"
|
| 39 |
+
},
|
| 40 |
+
"pred_horizon": 5,
|
| 41 |
+
"rdt": {
|
| 42 |
+
"cond_pos_embed_type": "multimodal",
|
| 43 |
+
"depth": 28,
|
| 44 |
+
"hidden_size": 2048,
|
| 45 |
+
"num_heads": 32
|
| 46 |
+
},
|
| 47 |
+
"state_adaptor": "mlp3x_gelu",
|
| 48 |
+
"state_token_dim": 2
|
| 49 |
+
}
|