Upload config.json with huggingface_hub
Browse files- config.json +7 -7
config.json
CHANGED
|
@@ -20,11 +20,11 @@
|
|
| 20 |
]
|
| 21 |
},
|
| 22 |
"architectures": [
|
| 23 |
-
"
|
| 24 |
],
|
| 25 |
"auto_map": {
|
| 26 |
-
"AutoConfig": "
|
| 27 |
-
"AutoModelForImageTextToText": "
|
| 28 |
},
|
| 29 |
"image_patch_id": 152066,
|
| 30 |
"initializer_range": 0.02,
|
|
@@ -39,7 +39,7 @@
|
|
| 39 |
"intermediate_size": 18944,
|
| 40 |
"layer_norm_eps": 1e-06,
|
| 41 |
"max_position_embeddings": 4096,
|
| 42 |
-
"model_type": "
|
| 43 |
"norm_after": false,
|
| 44 |
"num_attention_heads": 28,
|
| 45 |
"num_hidden_layers": 28,
|
|
@@ -53,10 +53,10 @@
|
|
| 53 |
"use_qk_norm": false,
|
| 54 |
"vocab_size": 152064
|
| 55 |
},
|
| 56 |
-
"model_type": "
|
| 57 |
"n_action_bins": 256,
|
| 58 |
"norm_stats": {
|
| 59 |
-
"
|
| 60 |
"action": {
|
| 61 |
"max": [
|
| 62 |
0.06042003631591797,
|
|
@@ -263,7 +263,7 @@
|
|
| 263 |
"initializer_range": 0.02,
|
| 264 |
"intermediate_size": 4304,
|
| 265 |
"layer_norm_eps": 1e-06,
|
| 266 |
-
"model_type": "
|
| 267 |
"num_attention_heads": 16,
|
| 268 |
"num_hidden_layers": 27,
|
| 269 |
"num_key_value_heads": 16,
|
|
|
|
| 20 |
]
|
| 21 |
},
|
| 22 |
"architectures": [
|
| 23 |
+
"SPRVLAForActionReasoning"
|
| 24 |
],
|
| 25 |
"auto_map": {
|
| 26 |
+
"AutoConfig": "configuration_sprvla.SPRVLAConfig",
|
| 27 |
+
"AutoModelForImageTextToText": "modeling_sprvla.SPRVLAForActionReasoning"
|
| 28 |
},
|
| 29 |
"image_patch_id": 152066,
|
| 30 |
"initializer_range": 0.02,
|
|
|
|
| 39 |
"intermediate_size": 18944,
|
| 40 |
"layer_norm_eps": 1e-06,
|
| 41 |
"max_position_embeddings": 4096,
|
| 42 |
+
"model_type": "sprvla_llm",
|
| 43 |
"norm_after": false,
|
| 44 |
"num_attention_heads": 28,
|
| 45 |
"num_hidden_layers": 28,
|
|
|
|
| 53 |
"use_qk_norm": false,
|
| 54 |
"vocab_size": 152064
|
| 55 |
},
|
| 56 |
+
"model_type": "sprvla",
|
| 57 |
"n_action_bins": 256,
|
| 58 |
"norm_stats": {
|
| 59 |
+
"sprvla": {
|
| 60 |
"action": {
|
| 61 |
"max": [
|
| 62 |
0.06042003631591797,
|
|
|
|
| 263 |
"initializer_range": 0.02,
|
| 264 |
"intermediate_size": 4304,
|
| 265 |
"layer_norm_eps": 1e-06,
|
| 266 |
+
"model_type": "sprvla_vit",
|
| 267 |
"num_attention_heads": 16,
|
| 268 |
"num_hidden_layers": 27,
|
| 269 |
"num_key_value_heads": 16,
|