Vitron commited on
Commit
c836a35
·
verified ·
1 Parent(s): 52016d4

Upload config.json with huggingface_hub

Browse files
Files changed (1) hide show
  1. config.json +50 -0
config.json ADDED
@@ -0,0 +1,50 @@
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
1
+ {
2
+ "X": [
3
+ "Video",
4
+ "Image"
5
+ ],
6
+ "_name_or_path": "checkpoints/Vitron-base",
7
+ "architectures": [
8
+ "LlavaLlamaForCausalLM"
9
+ ],
10
+ "bos_token_id": 1,
11
+ "eos_token_id": 2,
12
+ "freeze_mm_mlp_adapter": false,
13
+ "freeze_region_mlp_adapter": false,
14
+ "hidden_act": "silu",
15
+ "hidden_size": 4096,
16
+ "image_aspect_ratio": "pad",
17
+ "image_grid_pinpoints": null,
18
+ "initializer_range": 0.02,
19
+ "intermediate_size": 11008,
20
+ "max_position_embeddings": 4096,
21
+ "mm_hidden_size": 1024,
22
+ "mm_image_tower": "checkpoints/LanguageBind/LanguageBind_Image",
23
+ "mm_projector_lr": 1e-05,
24
+ "mm_projector_type": "mlp2x_gelu",
25
+ "mm_use_im_patch_token": false,
26
+ "mm_use_im_start_end": false,
27
+ "mm_use_x_patch_token": false,
28
+ "mm_use_x_start_end": false,
29
+ "mm_video_tower": "checkpoints/LanguageBind/LanguageBind_Video_merge",
30
+ "mm_vision_select_feature": "patch",
31
+ "mm_vision_select_layer": -2,
32
+ "model_type": "llava",
33
+ "num_attention_heads": 32,
34
+ "num_hidden_layers": 32,
35
+ "num_key_value_heads": 32,
36
+ "pad_token_id": 0,
37
+ "pretraining_tp": 1,
38
+ "rms_norm_eps": 1e-05,
39
+ "rope_scaling": null,
40
+ "tie_word_embeddings": false,
41
+ "tokenizer_model_max_length": 3072,
42
+ "tokenizer_padding_side": "right",
43
+ "torch_dtype": "bfloat16",
44
+ "transformers_version": "4.31.0",
45
+ "tune_mm_mlp_adapter": true,
46
+ "tune_region_mlp_adapter": true,
47
+ "use_cache": true,
48
+ "use_mm_proj": true,
49
+ "vocab_size": 32000
50
+ }