Uploading model files
Browse files- model.safetensors +2 -2
- vla_config.json +21 -9
model.safetensors
CHANGED
|
@@ -1,3 +1,3 @@
|
|
| 1 |
version https://git-lfs.github.com/spec/v1
|
| 2 |
-
oid sha256:
|
| 3 |
-
size
|
|
|
|
| 1 |
version https://git-lfs.github.com/spec/v1
|
| 2 |
+
oid sha256:b833cd1cee8d6070d162883e211cb13cdb455ed74633b5b06de199b2035be5ef
|
| 3 |
+
size 1219514848
|
vla_config.json
CHANGED
|
@@ -10,7 +10,7 @@
|
|
| 10 |
"difficulty",
|
| 11 |
"horizon"
|
| 12 |
],
|
| 13 |
-
"backbone_id_for_provenance": "facebook/dinov3-
|
| 14 |
"difficulty_condition_dropout": 0.45,
|
| 15 |
"difficulty_to_id": {
|
| 16 |
"low": 0,
|
|
@@ -53,7 +53,7 @@
|
|
| 53 |
"text_feature_version": "signed_hash_subtokens_v2",
|
| 54 |
"unknown_condition_strategy": "trained_fallback_embedding",
|
| 55 |
"vision_config": {
|
| 56 |
-
"_name_or_path": "facebook/dinov3-
|
| 57 |
"apply_layernorm": true,
|
| 58 |
"architectures": [
|
| 59 |
"DINOv3ViTModel"
|
|
@@ -63,14 +63,14 @@
|
|
| 63 |
"drop_path_rate": 0.0,
|
| 64 |
"dtype": "float32",
|
| 65 |
"hidden_act": "gelu",
|
| 66 |
-
"hidden_size":
|
| 67 |
"id2label": {
|
| 68 |
"0": "LABEL_0",
|
| 69 |
"1": "LABEL_1"
|
| 70 |
},
|
| 71 |
"image_size": 224,
|
| 72 |
"initializer_range": 0.02,
|
| 73 |
-
"intermediate_size":
|
| 74 |
"is_encoder_decoder": false,
|
| 75 |
"key_bias": false,
|
| 76 |
"label2id": {
|
|
@@ -81,15 +81,15 @@
|
|
| 81 |
"layerscale_value": 1.0,
|
| 82 |
"mlp_bias": true,
|
| 83 |
"model_type": "dinov3_vit",
|
| 84 |
-
"num_attention_heads":
|
| 85 |
"num_channels": 3,
|
| 86 |
-
"num_hidden_layers":
|
| 87 |
"num_register_tokens": 4,
|
| 88 |
"out_features": [
|
| 89 |
-
"
|
| 90 |
],
|
| 91 |
"out_indices": [
|
| 92 |
-
|
| 93 |
],
|
| 94 |
"output_attentions": false,
|
| 95 |
"output_hidden_states": false,
|
|
@@ -116,7 +116,19 @@
|
|
| 116 |
"stage9",
|
| 117 |
"stage10",
|
| 118 |
"stage11",
|
| 119 |
-
"stage12"
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
| 120 |
],
|
| 121 |
"transformers_version": "5.4.0",
|
| 122 |
"use_gated_mlp": false,
|
|
|
|
| 10 |
"difficulty",
|
| 11 |
"horizon"
|
| 12 |
],
|
| 13 |
+
"backbone_id_for_provenance": "facebook/dinov3-vitl16-pretrain-lvd1689m",
|
| 14 |
"difficulty_condition_dropout": 0.45,
|
| 15 |
"difficulty_to_id": {
|
| 16 |
"low": 0,
|
|
|
|
| 53 |
"text_feature_version": "signed_hash_subtokens_v2",
|
| 54 |
"unknown_condition_strategy": "trained_fallback_embedding",
|
| 55 |
"vision_config": {
|
| 56 |
+
"_name_or_path": "facebook/dinov3-vitl16-pretrain-lvd1689m",
|
| 57 |
"apply_layernorm": true,
|
| 58 |
"architectures": [
|
| 59 |
"DINOv3ViTModel"
|
|
|
|
| 63 |
"drop_path_rate": 0.0,
|
| 64 |
"dtype": "float32",
|
| 65 |
"hidden_act": "gelu",
|
| 66 |
+
"hidden_size": 1024,
|
| 67 |
"id2label": {
|
| 68 |
"0": "LABEL_0",
|
| 69 |
"1": "LABEL_1"
|
| 70 |
},
|
| 71 |
"image_size": 224,
|
| 72 |
"initializer_range": 0.02,
|
| 73 |
+
"intermediate_size": 4096,
|
| 74 |
"is_encoder_decoder": false,
|
| 75 |
"key_bias": false,
|
| 76 |
"label2id": {
|
|
|
|
| 81 |
"layerscale_value": 1.0,
|
| 82 |
"mlp_bias": true,
|
| 83 |
"model_type": "dinov3_vit",
|
| 84 |
+
"num_attention_heads": 16,
|
| 85 |
"num_channels": 3,
|
| 86 |
+
"num_hidden_layers": 24,
|
| 87 |
"num_register_tokens": 4,
|
| 88 |
"out_features": [
|
| 89 |
+
"stage24"
|
| 90 |
],
|
| 91 |
"out_indices": [
|
| 92 |
+
24
|
| 93 |
],
|
| 94 |
"output_attentions": false,
|
| 95 |
"output_hidden_states": false,
|
|
|
|
| 116 |
"stage9",
|
| 117 |
"stage10",
|
| 118 |
"stage11",
|
| 119 |
+
"stage12",
|
| 120 |
+
"stage13",
|
| 121 |
+
"stage14",
|
| 122 |
+
"stage15",
|
| 123 |
+
"stage16",
|
| 124 |
+
"stage17",
|
| 125 |
+
"stage18",
|
| 126 |
+
"stage19",
|
| 127 |
+
"stage20",
|
| 128 |
+
"stage21",
|
| 129 |
+
"stage22",
|
| 130 |
+
"stage23",
|
| 131 |
+
"stage24"
|
| 132 |
],
|
| 133 |
"transformers_version": "5.4.0",
|
| 134 |
"use_gated_mlp": false,
|