Update config.json
Browse files- config.json +23 -29
config.json
CHANGED
|
@@ -1,27 +1,22 @@
|
|
| 1 |
{
|
| 2 |
-
"
|
| 3 |
-
"
|
| 4 |
-
"
|
| 5 |
-
"
|
| 6 |
-
"
|
| 7 |
-
"
|
| 8 |
-
"
|
| 9 |
-
"is_encoder_decoder":false,
|
| 10 |
-
"layer_norm_eps":1e-
|
| 11 |
-
"
|
| 12 |
-
"link_tower_type":"add",
|
| 13 |
-
"max_text_len":50,
|
| 14 |
-
"
|
| 15 |
-
"
|
| 16 |
-
"
|
| 17 |
-
"num_hidden_layers":6,
|
| 18 |
-
"stop_gradient":false,
|
| 19 |
-
"tie_word_embeddings":false,
|
| 20 |
-
"vocab_size":50265,
|
| 21 |
"text_config_dict": null,
|
| 22 |
"text_config":{
|
| 23 |
"architectures": ["BridgeTowerTextModel"],
|
| 24 |
-
"classifier_dropout": null,
|
| 25 |
"vocab_size": 50265,
|
| 26 |
"hidden_size": 768,
|
| 27 |
"num_hidden_layers": 12,
|
|
@@ -38,20 +33,19 @@
|
|
| 38 |
"bos_token_id": 0,
|
| 39 |
"eos_token_id": 2,
|
| 40 |
"position_embedding_type": "absolute",
|
| 41 |
-
"use_cache": true
|
|
|
|
| 42 |
},
|
| 43 |
"vision_config_dict": null,
|
| 44 |
"vision_config":{
|
| 45 |
"architectures": ["BridgeTowerVisionModel"],
|
| 46 |
-
"
|
| 47 |
-
"
|
| 48 |
-
"width": 768,
|
| 49 |
-
"layers": 12,
|
| 50 |
"patch_size": 16,
|
| 51 |
-
"
|
| 52 |
-
"output_resolution": 288,
|
| 53 |
"stop_gradient": false,
|
| 54 |
-
"
|
| 55 |
-
"vit_remove_last": false
|
|
|
|
| 56 |
}
|
| 57 |
}
|
|
|
|
| 1 |
{
|
| 2 |
+
"share_cross_modal_transformer_layers": true,
|
| 3 |
+
"drop_rate": 0.1,
|
| 4 |
+
"head_hidden_scale": 2,
|
| 5 |
+
"hidden_act": "gelu",
|
| 6 |
+
"hidden_size": 768,
|
| 7 |
+
"input_image_embed_size": 768,
|
| 8 |
+
"input_text_embed_size": 768,
|
| 9 |
+
"is_encoder_decoder": false,
|
| 10 |
+
"layer_norm_eps": 1e-05,
|
| 11 |
+
"share_link_tower_layers": false,
|
| 12 |
+
"link_tower_type": "add",
|
| 13 |
+
"max_text_len": 50,
|
| 14 |
+
"mlp_ratio=4,
|
| 15 |
+
"num_attention_heads=12,
|
| 16 |
+
"num_hidden_layers=6,
|
|
|
|
|
|
|
|
|
|
|
|
|
| 17 |
"text_config_dict": null,
|
| 18 |
"text_config":{
|
| 19 |
"architectures": ["BridgeTowerTextModel"],
|
|
|
|
| 20 |
"vocab_size": 50265,
|
| 21 |
"hidden_size": 768,
|
| 22 |
"num_hidden_layers": 12,
|
|
|
|
| 33 |
"bos_token_id": 0,
|
| 34 |
"eos_token_id": 2,
|
| 35 |
"position_embedding_type": "absolute",
|
| 36 |
+
"use_cache": true,
|
| 37 |
+
"classifier_dropout": null
|
| 38 |
},
|
| 39 |
"vision_config_dict": null,
|
| 40 |
"vision_config":{
|
| 41 |
"architectures": ["BridgeTowerVisionModel"],
|
| 42 |
+
"hidden_size": 768,
|
| 43 |
+
"num_hidden_layers": 12,
|
|
|
|
|
|
|
| 44 |
"patch_size": 16,
|
| 45 |
+
"image_size": 288,
|
|
|
|
| 46 |
"stop_gradient": false,
|
| 47 |
+
"share_layernorm": true,
|
| 48 |
+
"vit_remove_last": false,
|
| 49 |
+
"init_layernorm_from_vit": false,
|
| 50 |
}
|
| 51 |
}
|