Hugging Face's logo

weizhiwang
/
llava-video-llama-3.1-8b-siglip-so-384-aapool-144-projector

llava_llama
Model card Files Community
weizhiwang commited on
Commit
91d15e7
•
1 Parent(s): f32383a

Upload folder using huggingface_hub

Browse files
Files changed (5) hide show
  1. checkpoint-2180/config.json +51 -0
  2. checkpoint-2180/mm_projector.bin +3 -0
  3. config.json +51 -0
  4. mm_projector.bin +3 -0
  5. trainer_state.json +0 -0
checkpoint-2180/config.json ADDED
@@ -0,0 +1,51 @@
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
1
+ {
2
+ "_name_or_path": "/mnt/raid5/weizhi/models/Meta-Llama-3.1-8B",
3
+ "architectures": [
4
+ "LlamaForCausalLM"
5
+ ],
6
+ "attention_bias": false,
7
+ "attention_dropout": 0.0,
8
+ "bos_token_id": 128000,
9
+ "eos_token_id": 128001,
10
+ "freeze_mm_mlp_adapter": false,
11
+ "hidden_act": "silu",
12
+ "hidden_size": 4096,
13
+ "image_aspect_ratio": "square",
14
+ "initializer_range": 0.02,
15
+ "intermediate_size": 14336,
16
+ "max_position_embeddings": 131072,
17
+ "mlp_bias": false,
18
+ "mm_hidden_size": 1152,
19
+ "mm_num_image_tokens": 144,
20
+ "mm_patch_merge_type": "flat",
21
+ "mm_projector_lr": null,
22
+ "mm_projector_type": "aapool_mlp",
23
+ "mm_use_im_patch_token": false,
24
+ "mm_use_im_start_end": false,
25
+ "mm_vision_select_feature": "patch",
26
+ "mm_vision_select_layer": -1,
27
+ "mm_vision_tower": "google/siglip-so400m-patch14-384",
28
+ "model_type": "llava_llama",
29
+ "num_attention_heads": 32,
30
+ "num_hidden_layers": 32,
31
+ "num_key_value_heads": 8,
32
+ "pretraining_tp": 1,
33
+ "rms_norm_eps": 1e-05,
34
+ "rope_scaling": {
35
+ "factor": 8.0,
36
+ "high_freq_factor": 4.0,
37
+ "low_freq_factor": 1.0,
38
+ "original_max_position_embeddings": 8192,
39
+ "rope_type": "llama3"
40
+ },
41
+ "rope_theta": 500000.0,
42
+ "tie_word_embeddings": false,
43
+ "tokenizer_model_max_length": 2048,
44
+ "tokenizer_padding_side": "right",
45
+ "torch_dtype": "bfloat16",
46
+ "transformers_version": "4.44.0",
47
+ "tune_mm_mlp_adapter": true,
48
+ "use_cache": false,
49
+ "use_mm_proj": true,
50
+ "vocab_size": 128256
51
+ }
checkpoint-2180/mm_projector.bin ADDED
@@ -0,0 +1,3 @@
 
 
 
 
1
+ version https://git-lfs.github.com/spec/v1
2
+ oid sha256:37cf6a4c3ff44f9d61fae1629db38c56ca99d58c4f8fe51ee6e1c0699eb4644d
3
+ size 43010168
config.json ADDED
@@ -0,0 +1,51 @@
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
1
+ {
2
+ "_name_or_path": "/mnt/raid5/weizhi/models/Meta-Llama-3.1-8B",
3
+ "architectures": [
4
+ "LlamaForCausalLM"
5
+ ],
6
+ "attention_bias": false,
7
+ "attention_dropout": 0.0,
8
+ "bos_token_id": 128000,
9
+ "eos_token_id": 128001,
10
+ "freeze_mm_mlp_adapter": false,
11
+ "hidden_act": "silu",
12
+ "hidden_size": 4096,
13
+ "image_aspect_ratio": "square",
14
+ "initializer_range": 0.02,
15
+ "intermediate_size": 14336,
16
+ "max_position_embeddings": 131072,
17
+ "mlp_bias": false,
18
+ "mm_hidden_size": 1152,
19
+ "mm_num_image_tokens": 144,
20
+ "mm_patch_merge_type": "flat",
21
+ "mm_projector_lr": null,
22
+ "mm_projector_type": "aapool_mlp",
23
+ "mm_use_im_patch_token": false,
24
+ "mm_use_im_start_end": false,
25
+ "mm_vision_select_feature": "patch",
26
+ "mm_vision_select_layer": -1,
27
+ "mm_vision_tower": "google/siglip-so400m-patch14-384",
28
+ "model_type": "llava_llama",
29
+ "num_attention_heads": 32,
30
+ "num_hidden_layers": 32,
31
+ "num_key_value_heads": 8,
32
+ "pretraining_tp": 1,
33
+ "rms_norm_eps": 1e-05,
34
+ "rope_scaling": {
35
+ "factor": 8.0,
36
+ "high_freq_factor": 4.0,
37
+ "low_freq_factor": 1.0,
38
+ "original_max_position_embeddings": 8192,
39
+ "rope_type": "llama3"
40
+ },
41
+ "rope_theta": 500000.0,
42
+ "tie_word_embeddings": false,
43
+ "tokenizer_model_max_length": 2048,
44
+ "tokenizer_padding_side": "right",
45
+ "torch_dtype": "bfloat16",
46
+ "transformers_version": "4.44.0",
47
+ "tune_mm_mlp_adapter": true,
48
+ "use_cache": true,
49
+ "use_mm_proj": true,
50
+ "vocab_size": 128256
51
+ }
mm_projector.bin ADDED
@@ -0,0 +1,3 @@
 
 
 
 
1
+ version https://git-lfs.github.com/spec/v1
2
+ oid sha256:37cf6a4c3ff44f9d61fae1629db38c56ca99d58c4f8fe51ee6e1c0699eb4644d
3
+ size 43010168
trainer_state.json ADDED
The diff for this file is too large to render. See raw diff