Transformers
PyTorch
English
bridgetower
Inference Endpoints
shaoyent commited on
Commit
323700b
1 Parent(s): 52dfd71

Update config.json

Browse files
Files changed (1) hide show
  1. config.json +11 -11
config.json CHANGED
@@ -1,31 +1,27 @@
1
  {
2
-
3
- "cache_dir":"/tmp",
4
  "drop_rate":0.1,
5
  "freeze_RoBERTa":false,
6
  "freeze_ViT":false,
7
  "freeze_layer_count_roberta":false,
8
  "freeze_layer_count_vit":false,
9
  "head_hidden_scale":2,
 
10
  "hidden_size":768,
11
  "image_size":288,
12
  "input_text_embed_size":768,
 
 
13
  "layer_norm_eps":1e-5,
14
  "link_tower_shared":false,
15
  "link_tower_type":"add",
16
- "log_dir":"log_dir",
17
  "max_text_len":50,
18
  "mlp_ratio":4,
19
- "model_type":"bridgetower",
20
- "num_heads":12,
21
- "num_layers":6,
22
- "num_nodes":1,
23
- "only_load_cross_modal_from_meter":false,
24
- "patch_size":16,
25
  "resolution_before":224,
26
  "stop_gradient":false,
27
  "task_head_layers":2,
28
- "test_only":false,
29
  "tokenizer":"roberta-base",
30
  "unfreeze_RoBERTa_attention":false,
31
  "unfreeze_RoBERTa_embeddings":false,
@@ -33,9 +29,13 @@
33
  "unfreeze_RoBERTa_layernorm":false,
34
  "unfreeze_ViT_attention":false,
35
  "unfreeze_ViT_layernorm":false,
36
- "vit":"ViT-B/16",
 
37
  "vit_layernorm_init_from_vit":false,
38
  "vit_layernorm_shared":true,
 
39
  "vit_remove_last":false,
 
 
40
  "vocab_size":50265
41
  }
 
1
  {
 
 
2
  "drop_rate":0.1,
3
  "freeze_RoBERTa":false,
4
  "freeze_ViT":false,
5
  "freeze_layer_count_roberta":false,
6
  "freeze_layer_count_vit":false,
7
  "head_hidden_scale":2,
8
+ "hidden_act":"gelu",
9
  "hidden_size":768,
10
  "image_size":288,
11
  "input_text_embed_size":768,
12
+ "input_image_embed_size":768,
13
+ "is_encoder_decoder":false,
14
  "layer_norm_eps":1e-5,
15
  "link_tower_shared":false,
16
  "link_tower_type":"add",
 
17
  "max_text_len":50,
18
  "mlp_ratio":4,
19
+ "num_attention_heads":12,
20
+ "num_hidden_layers":6,
 
 
 
 
21
  "resolution_before":224,
22
  "stop_gradient":false,
23
  "task_head_layers":2,
24
+ "tie_word_embeddings":false,
25
  "tokenizer":"roberta-base",
26
  "unfreeze_RoBERTa_attention":false,
27
  "unfreeze_RoBERTa_embeddings":false,
 
29
  "unfreeze_RoBERTa_layernorm":false,
30
  "unfreeze_ViT_attention":false,
31
  "unfreeze_ViT_layernorm":false,
32
+ "vit_embed_dim":512,
33
+ "vit_layers":12,
34
  "vit_layernorm_init_from_vit":false,
35
  "vit_layernorm_shared":true,
36
+ "vit_patch_size":16,
37
  "vit_remove_last":false,
38
+ "vit_transformer_width":512,
39
+ "vit_width":768,
40
  "vocab_size":50265
41
  }