Hugging Face's logo

ashutoshj01
/
blip-vqa-lora-finetune-model

Visual Question Answering
Transformers
Safetensors
blip
Model card Files Community
ashutoshj01 commited on
Commit
9a4340e
·
verified ·
1 Parent(s): 4c0fc31

Upload BlipForQuestionAnswering

Browse files
Files changed (2) hide show
  1. config.json +54 -54
  2. generation_config.json +7 -7
config.json CHANGED
@@ -1,54 +1,54 @@
1
- {
2
- "architectures": [
3
- "BlipForQuestionAnswering"
4
- ],
5
- "image_text_hidden_size": 256,
6
- "initializer_factor": 1.0,
7
- "initializer_range": 0.02,
8
- "label_smoothing": 0.0,
9
- "logit_scale_init_value": 2.6592,
10
- "model_type": "blip",
11
- "projection_dim": 512,
12
- "text_config": {
13
- "_attn_implementation_autoset": true,
14
- "attention_probs_dropout_prob": 0.0,
15
- "encoder_hidden_size": 768,
16
- "hidden_act": "gelu",
17
- "hidden_dropout_prob": 0.0,
18
- "hidden_size": 768,
19
- "initializer_factor": 1.0,
20
- "initializer_range": 0.02,
21
- "intermediate_size": 3072,
22
- "label_smoothing": 0.0,
23
- "layer_norm_eps": 1e-12,
24
- "max_position_embeddings": 512,
25
- "model_type": "blip_text_model",
26
- "num_attention_heads": 12,
27
- "num_hidden_layers": 12,
28
- "projection_dim": 768,
29
- "torch_dtype": "float32",
30
- "use_cache": true,
31
- "vocab_size": 30524
32
- },
33
- "torch_dtype": "float32",
34
- "transformers_version": "4.51.3",
35
- "vision_config": {
36
- "_attn_implementation_autoset": true,
37
- "attention_dropout": 0.0,
38
- "dropout": 0.0,
39
- "hidden_act": "gelu",
40
- "hidden_size": 768,
41
- "image_size": 384,
42
- "initializer_factor": 1.0,
43
- "initializer_range": 0.02,
44
- "intermediate_size": 3072,
45
- "layer_norm_eps": 1e-05,
46
- "model_type": "blip_vision_model",
47
- "num_attention_heads": 12,
48
- "num_channels": 3,
49
- "num_hidden_layers": 12,
50
- "patch_size": 16,
51
- "projection_dim": 512,
52
- "torch_dtype": "float32"
53
- }
54
- }
 
1
+ {
2
+ "architectures": [
3
+ "BlipForQuestionAnswering"
4
+ ],
5
+ "image_text_hidden_size": 256,
6
+ "initializer_factor": 1.0,
7
+ "initializer_range": 0.02,
8
+ "label_smoothing": 0.0,
9
+ "logit_scale_init_value": 2.6592,
10
+ "model_type": "blip",
11
+ "projection_dim": 512,
12
+ "text_config": {
13
+ "_attn_implementation_autoset": true,
14
+ "attention_probs_dropout_prob": 0.0,
15
+ "encoder_hidden_size": 768,
16
+ "hidden_act": "gelu",
17
+ "hidden_dropout_prob": 0.0,
18
+ "hidden_size": 768,
19
+ "initializer_factor": 1.0,
20
+ "initializer_range": 0.02,
21
+ "intermediate_size": 3072,
22
+ "label_smoothing": 0.0,
23
+ "layer_norm_eps": 1e-12,
24
+ "max_position_embeddings": 512,
25
+ "model_type": "blip_text_model",
26
+ "num_attention_heads": 12,
27
+ "num_hidden_layers": 12,
28
+ "projection_dim": 768,
29
+ "torch_dtype": "float32",
30
+ "use_cache": true,
31
+ "vocab_size": 30524
32
+ },
33
+ "torch_dtype": "float32",
34
+ "transformers_version": "4.51.3",
35
+ "vision_config": {
36
+ "_attn_implementation_autoset": true,
37
+ "attention_dropout": 0.0,
38
+ "dropout": 0.0,
39
+ "hidden_act": "gelu",
40
+ "hidden_size": 768,
41
+ "image_size": 384,
42
+ "initializer_factor": 1.0,
43
+ "initializer_range": 0.02,
44
+ "intermediate_size": 3072,
45
+ "layer_norm_eps": 1e-05,
46
+ "model_type": "blip_vision_model",
47
+ "num_attention_heads": 12,
48
+ "num_channels": 3,
49
+ "num_hidden_layers": 12,
50
+ "patch_size": 16,
51
+ "projection_dim": 512,
52
+ "torch_dtype": "float32"
53
+ }
54
+ }
generation_config.json CHANGED
@@ -1,7 +1,7 @@
1
- {
2
- "_from_model_config": true,
3
- "bos_token_id": 30522,
4
- "eos_token_id": 2,
5
- "pad_token_id": 0,
6
- "transformers_version": "4.51.3"
7
- }
 
1
+ {
2
+ "_from_model_config": true,
3
+ "bos_token_id": 30522,
4
+ "eos_token_id": 2,
5
+ "pad_token_id": 0,
6
+ "transformers_version": "4.51.3"
7
+ }