Hugging Face's logo

Xenova
/
tiny-random-PaliGemmaForConditionalGeneration

Image-Text-to-Text
Transformers
ONNX
Safetensors
Transformers.js
paligemma
pretraining
text-generation-inference
Inference Endpoints
Model card Files Community
2
Xenova HF staff commited on
Commit
c0ac708
•
1 Parent(s): 439caf3

Upload PaliGemmaForConditionalGeneration

Browse files
Files changed (2) hide show
  1. config.json +15 -10
  2. model.safetensors +2 -2
config.json CHANGED
@@ -2,32 +2,37 @@
2
  "architectures": [
3
  "PaliGemmaForConditionalGeneration"
4
  ],
 
 
5
  "hidden_size": 16,
6
  "ignore_index": -100,
7
  "image_token_index": 257152,
8
  "model_type": "paligemma",
9
- "projection_dim": 32,
 
10
  "text_config": {
11
- "head_dim": 16,
12
- "hidden_size": 32,
13
- "intermediate_size": 32,
14
  "model_type": "gemma",
15
  "num_attention_heads": 4,
16
  "num_hidden_layers": 2,
17
- "num_image_tokens": 196,
18
- "num_key_value_heads": 4,
19
  "vocab_size": 257216
20
  },
21
  "torch_dtype": "float32",
22
  "transformers_version": "4.42.0.dev0",
23
  "vision_config": {
24
- "hidden_size": 32,
25
  "intermediate_size": 32,
26
  "model_type": "siglip_vision_model",
27
- "num_attention_heads": 4,
28
  "num_hidden_layers": 2,
29
- "patch_size": 16,
30
- "projection_dim": 32
 
 
 
31
  },
32
  "vocab_size": 257152
33
  }
 
2
  "architectures": [
3
  "PaliGemmaForConditionalGeneration"
4
  ],
5
+ "bos_token_id": 2,
6
+ "eos_token_id": 1,
7
  "hidden_size": 16,
8
  "ignore_index": -100,
9
  "image_token_index": 257152,
10
  "model_type": "paligemma",
11
+ "pad_token_id": 0,
12
+ "projection_dim": 16,
13
  "text_config": {
14
+ "hidden_size": 16,
15
+ "intermediate_size": 16,
 
16
  "model_type": "gemma",
17
  "num_attention_heads": 4,
18
  "num_hidden_layers": 2,
19
+ "num_image_tokens": 256,
20
+ "num_key_value_heads": 1,
21
  "vocab_size": 257216
22
  },
23
  "torch_dtype": "float32",
24
  "transformers_version": "4.42.0.dev0",
25
  "vision_config": {
26
+ "hidden_size": 16,
27
  "intermediate_size": 32,
28
  "model_type": "siglip_vision_model",
29
+ "num_attention_heads": 16,
30
  "num_hidden_layers": 2,
31
+ "num_image_tokens": 256,
32
+ "patch_size": 14,
33
+ "projection_dim": 16,
34
+ "projector_hidden_act": "gelu_fast",
35
+ "vision_use_head": false
36
  },
37
  "vocab_size": 257152
38
  }
model.safetensors CHANGED
@@ -1,3 +1,3 @@
1
  version https://git-lfs.github.com/spec/v1
2
- oid sha256:c2aff118cf16a794a2831fd73c9fb7924e9489e37a1bb9aa40eee474c9e05c65
3
- size 33228536
 
1
  version https://git-lfs.github.com/spec/v1
2
+ oid sha256:e0412c1ad39be16b8f8485ccaf603dea2cc0acee2c0fba18118cab1f8d2cebc2
3
+ size 16876488