Image-Text-to-Text
Transformers
Safetensors
Portuguese
tinyllava
text-generation
vision
custom_code
nicholasKluge commited on
Commit
ddbf708
·
verified ·
1 Parent(s): af585c3

Upload config.json with huggingface_hub

Browse files
Files changed (1) hide show
  1. config.json +75 -0
config.json ADDED
@@ -0,0 +1,75 @@
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
1
+ {
2
+ "architectures": [
3
+ "TinyLlavaForConditionalGeneration"
4
+ ],
5
+ "cache_dir": null,
6
+ "connector_type": "mlp2x_gelu",
7
+ "hidden_size": 2560,
8
+ "ignore_index": -100,
9
+ "image_aspect_ratio": "square",
10
+ "image_token_index": -200,
11
+ "llm_model_name_or_path": "TucanoBR/Tucano-2b4",
12
+ "model_type": "tinyllava",
13
+ "auto_map": {
14
+ "AutoConfig": "configuration.TinyLlavaConfig",
15
+ "AutoModelForCausalLM": "modeling_tinyllava_tucano.TinyLlavaForConditionalGeneration"
16
+ },
17
+ "num_queries": 128,
18
+ "num_resampler_layers": 3,
19
+ "pad_token": "<pad>",
20
+ "pad_token_id": 3,
21
+ "resampler_hidden_size": 768,
22
+ "text_config": {
23
+ "_name_or_path": "TucanoBR/Tucano-2b4",
24
+ "architectures": [
25
+ "LlamaForCausalLM"
26
+ ],
27
+ "auto_map": {
28
+ "AutoConfig": "TucanoBR/Tucano-2b4--configuration_llama.LlamaConfig",
29
+ "AutoModelForCausalLM": "TucanoBR/Tucano-2b4--modeling_llama.LlamaForCausalLM"
30
+ },
31
+ "head_dim": 160,
32
+ "hidden_size": 2560,
33
+ "intermediate_size": 10240,
34
+ "max_position_embeddings": 4096,
35
+ "mlp_bias": false,
36
+ "model_type": "llama",
37
+ "num_attention_heads": 16,
38
+ "num_hidden_layers": 24,
39
+ "num_key_value_heads": 4,
40
+ "pad_token_id": 3,
41
+ "rms_norm_eps": 1e-05,
42
+ "torch_dtype": "bfloat16",
43
+ "use_cache": false
44
+ },
45
+ "tokenizer_model_max_length": 4096,
46
+ "tokenizer_name_or_path": "TucanoBR/Tucano-2b4",
47
+ "tokenizer_padding_side": "right",
48
+ "tokenizer_use_fast": true,
49
+ "torch_dtype": "bfloat16",
50
+ "transformers_version": "4.40.1",
51
+ "tune_type_connector": "full",
52
+ "tune_type_llm": "full",
53
+ "tune_type_vision_tower": "frozen",
54
+ "tune_vision_tower_from_layer": 0,
55
+ "use_cache": true,
56
+ "vision_config": {
57
+ "hidden_act": "gelu_pytorch_tanh",
58
+ "hidden_size": 1152,
59
+ "image_size": 384,
60
+ "intermediate_size": 4304,
61
+ "layer_norm_eps": 1e-06,
62
+ "model_name_or_path": "google/siglip-so400m-patch14-384",
63
+ "model_name_or_path2": "",
64
+ "model_type": "siglip_vision_model",
65
+ "num_attention_heads": 16,
66
+ "num_hidden_layers": 27,
67
+ "patch_size": 14
68
+ },
69
+ "vision_feature_layer": -2,
70
+ "vision_feature_select_strategy": "patch",
71
+ "vision_hidden_size": 1152,
72
+ "vision_model_name_or_path": "google/siglip-so400m-patch14-384",
73
+ "vision_model_name_or_path2": "",
74
+ "vocab_size": 32000
75
+ }