Hugging Face's logo

binxia
/
LLMGA-pretrained-mlp

Model card Files Community
binxia commited on
Commit
086b9f5
•
1 Parent(s): e099719

Upload folder using huggingface_hub

Browse files
Files changed (27) hide show
  1. llmga-Phi-3-mini-128k-pretrain/config.json +153 -0
  2. llmga-Phi-3-mini-128k-pretrain/mm_projector.bin +3 -0
  3. llmga-Phi-3-mini-128k-pretrain/trainer_state.json +0 -0
  4. llmga-Qwen2-0.5B-pretrain/config.json +43 -0
  5. llmga-Qwen2-0.5B-pretrain/mm_projector.bin +3 -0
  6. llmga-Qwen2-0.5B-pretrain/trainer_state.json +0 -0
  7. llmga-Qwen2-1.5B-pretrain/config.json +43 -0
  8. llmga-Qwen2-1.5B-pretrain/mm_projector.bin +3 -0
  9. llmga-Qwen2-1.5B-pretrain/trainer_state.json +0 -0
  10. llmga-Qwen2-7B-pretrain/config.json +43 -0
  11. llmga-Qwen2-7B-pretrain/mm_projector.bin +3 -0
  12. llmga-Qwen2-7B-pretrain/trainer_state.json +0 -0
  13. llmga-gemma-pretrain/config.json +44 -0
  14. llmga-gemma-pretrain/mm_projector.bin +3 -0
  15. llmga-gemma-pretrain/trainer_state.json +0 -0
  16. llmga-llama3-8b-pretrain/config.json +43 -0
  17. llmga-llama3-8b-pretrain/mm_projector.bin +3 -0
  18. llmga-llama3-8b-pretrain/trainer_state.json +0 -0
  19. llmga-mistral-pretrain/config.json +41 -0
  20. llmga-mistral-pretrain/mm_projector.bin +3 -0
  21. llmga-mistral-pretrain/trainer_state.json +0 -0
  22. llmga-vicuna-13b-v1.5-pretrain/config.json +41 -0
  23. llmga-vicuna-13b-v1.5-pretrain/mm_projector.bin +3 -0
  24. llmga-vicuna-13b-v1.5-pretrain/trainer_state.json +0 -0
  25. llmga-vicuna-7b-v1.5-pretrain/config.json +40 -0
  26. llmga-vicuna-7b-v1.5-pretrain/mm_projector.bin +3 -0
  27. llmga-vicuna-7b-v1.5-pretrain/trainer_state.json +0 -0
llmga-Phi-3-mini-128k-pretrain/config.json ADDED
@@ -0,0 +1,153 @@
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
1
+ {
2
+ "_name_or_path": "/mnt/bn/wyt-large-dataset/xiabin-model/LLM/Phi-3-mini-128k-instruct",
3
+ "architectures": [
4
+ "Phi3ForCausalLM"
5
+ ],
6
+ "attention_bias": false,
7
+ "attention_dropout": 0.0,
8
+ "auto_map": {
9
+ "AutoConfig": "configuration_phi3.Phi3Config",
10
+ "AutoModelForCausalLM": "modeling_phi3.Phi3ForCausalLM"
11
+ },
12
+ "bos_token_id": 1,
13
+ "embd_pdrop": 0.0,
14
+ "eos_token_id": 32000,
15
+ "freeze_mm_mlp_adapter": false,
16
+ "hidden_act": "silu",
17
+ "hidden_size": 3072,
18
+ "image_aspect_ratio": "resizesquare",
19
+ "initializer_range": 0.02,
20
+ "intermediate_size": 8192,
21
+ "max_position_embeddings": 131072,
22
+ "mm_hidden_size": 1024,
23
+ "mm_patch_merge_type": "flat",
24
+ "mm_projector_lr": null,
25
+ "mm_projector_type": "mlp2x_gelu",
26
+ "mm_use_im_patch_token": false,
27
+ "mm_use_im_start_end": false,
28
+ "mm_vision_select_feature": "patch",
29
+ "mm_vision_select_layer": -2,
30
+ "mm_vision_tower": "openai/clip-vit-large-patch14-336",
31
+ "model_type": "llava_Phi3",
32
+ "num_attention_heads": 32,
33
+ "num_hidden_layers": 32,
34
+ "num_key_value_heads": 32,
35
+ "original_max_position_embeddings": 4096,
36
+ "pad_token_id": 32000,
37
+ "resid_pdrop": 0.0,
38
+ "rms_norm_eps": 1e-05,
39
+ "rope_scaling": {
40
+ "long_factor": [
41
+ 1.0299999713897705,
42
+ 1.0499999523162842,
43
+ 1.0499999523162842,
44
+ 1.0799999237060547,
45
+ 1.2299998998641968,
46
+ 1.2299998998641968,
47
+ 1.2999999523162842,
48
+ 1.4499999284744263,
49
+ 1.5999999046325684,
50
+ 1.6499998569488525,
51
+ 1.8999998569488525,
52
+ 2.859999895095825,
53
+ 3.68999981880188,
54
+ 5.419999599456787,
55
+ 5.489999771118164,
56
+ 5.489999771118164,
57
+ 9.09000015258789,
58
+ 11.579999923706055,
59
+ 15.65999984741211,
60
+ 15.769999504089355,
61
+ 15.789999961853027,
62
+ 18.360000610351562,
63
+ 21.989999771118164,
64
+ 23.079999923706055,
65
+ 30.009998321533203,
66
+ 32.35000228881836,
67
+ 32.590003967285156,
68
+ 35.56000518798828,
69
+ 39.95000457763672,
70
+ 53.840003967285156,
71
+ 56.20000457763672,
72
+ 57.95000457763672,
73
+ 59.29000473022461,
74
+ 59.77000427246094,
75
+ 59.920005798339844,
76
+ 61.190006256103516,
77
+ 61.96000671386719,
78
+ 62.50000762939453,
79
+ 63.3700065612793,
80
+ 63.48000717163086,
81
+ 63.48000717163086,
82
+ 63.66000747680664,
83
+ 63.850006103515625,
84
+ 64.08000946044922,
85
+ 64.760009765625,
86
+ 64.80001068115234,
87
+ 64.81001281738281,
88
+ 64.81001281738281
89
+ ],
90
+ "short_factor": [
91
+ 1.05,
92
+ 1.05,
93
+ 1.05,
94
+ 1.1,
95
+ 1.1,
96
+ 1.1500000000000001,
97
+ 1.2000000000000002,
98
+ 1.2500000000000002,
99
+ 1.3000000000000003,
100
+ 1.3500000000000003,
101
+ 1.5000000000000004,
102
+ 2.000000000000001,
103
+ 2.000000000000001,
104
+ 2.000000000000001,
105
+ 2.000000000000001,
106
+ 2.000000000000001,
107
+ 2.000000000000001,
108
+ 2.000000000000001,
109
+ 2.000000000000001,
110
+ 2.000000000000001,
111
+ 2.000000000000001,
112
+ 2.000000000000001,
113
+ 2.000000000000001,
114
+ 2.000000000000001,
115
+ 2.000000000000001,
116
+ 2.000000000000001,
117
+ 2.000000000000001,
118
+ 2.000000000000001,
119
+ 2.000000000000001,
120
+ 2.000000000000001,
121
+ 2.000000000000001,
122
+ 2.000000000000001,
123
+ 2.0500000000000007,
124
+ 2.0500000000000007,
125
+ 2.0500000000000007,
126
+ 2.1000000000000005,
127
+ 2.1000000000000005,
128
+ 2.1000000000000005,
129
+ 2.1500000000000004,
130
+ 2.1500000000000004,
131
+ 2.3499999999999996,
132
+ 2.549999999999999,
133
+ 2.5999999999999988,
134
+ 2.5999999999999988,
135
+ 2.7499999999999982,
136
+ 2.849999999999998,
137
+ 2.849999999999998,
138
+ 2.9499999999999975
139
+ ],
140
+ "type": "su"
141
+ },
142
+ "rope_theta": 10000.0,
143
+ "sliding_window": 262144,
144
+ "tie_word_embeddings": false,
145
+ "tokenizer_model_max_length": 2048,
146
+ "tokenizer_padding_side": "right",
147
+ "torch_dtype": "bfloat16",
148
+ "transformers_version": "4.41.2",
149
+ "tune_mm_mlp_adapter": true,
150
+ "use_cache": true,
151
+ "use_mm_proj": true,
152
+ "vocab_size": 32064
153
+ }
llmga-Phi-3-mini-128k-pretrain/mm_projector.bin ADDED
@@ -0,0 +1,3 @@
 
 
 
 
1
+ version https://git-lfs.github.com/spec/v1
2
+ oid sha256:40154d390cb770ac648ab5b484d3c41687f819ad4c2993af018f501fb830f0a0
3
+ size 25180216
llmga-Phi-3-mini-128k-pretrain/trainer_state.json ADDED
The diff for this file is too large to render. See raw diff
 
llmga-Qwen2-0.5B-pretrain/config.json ADDED
@@ -0,0 +1,43 @@
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
1
+ {
2
+ "_name_or_path": "/mnt/bn/wyt-large-dataset/xiabin-model/LLM/Qwen2-0.5B-Instruct",
3
+ "architectures": [
4
+ "Qwen2ForCausalLM"
5
+ ],
6
+ "attention_dropout": 0.0,
7
+ "bos_token_id": 151643,
8
+ "eos_token_id": 151645,
9
+ "freeze_mm_mlp_adapter": false,
10
+ "hidden_act": "silu",
11
+ "hidden_size": 896,
12
+ "image_aspect_ratio": "resizesquare",
13
+ "initializer_range": 0.02,
14
+ "intermediate_size": 4864,
15
+ "max_position_embeddings": 32768,
16
+ "max_window_layers": 24,
17
+ "mm_hidden_size": 1024,
18
+ "mm_patch_merge_type": "flat",
19
+ "mm_projector_lr": null,
20
+ "mm_projector_type": "mlp2x_gelu",
21
+ "mm_use_im_patch_token": false,
22
+ "mm_use_im_start_end": false,
23
+ "mm_vision_select_feature": "patch",
24
+ "mm_vision_select_layer": -2,
25
+ "mm_vision_tower": "openai/clip-vit-large-patch14-336",
26
+ "model_type": "llava_Qwen2",
27
+ "num_attention_heads": 14,
28
+ "num_hidden_layers": 24,
29
+ "num_key_value_heads": 2,
30
+ "rms_norm_eps": 1e-06,
31
+ "rope_theta": 1000000.0,
32
+ "sliding_window": 32768,
33
+ "tie_word_embeddings": true,
34
+ "tokenizer_model_max_length": 2048,
35
+ "tokenizer_padding_side": "right",
36
+ "torch_dtype": "bfloat16",
37
+ "transformers_version": "4.41.2",
38
+ "tune_mm_mlp_adapter": true,
39
+ "use_cache": true,
40
+ "use_mm_proj": true,
41
+ "use_sliding_window": false,
42
+ "vocab_size": 151936
43
+ }
llmga-Qwen2-0.5B-pretrain/mm_projector.bin ADDED
@@ -0,0 +1,3 @@
 
 
 
 
1
+ version https://git-lfs.github.com/spec/v1
2
+ oid sha256:91ff9f53028ff3205fdf3cd0033d0c84a2ceae28fc0e6adcc7a770336d5ae607
3
+ size 3446328
llmga-Qwen2-0.5B-pretrain/trainer_state.json ADDED
The diff for this file is too large to render. See raw diff
 
llmga-Qwen2-1.5B-pretrain/config.json ADDED
@@ -0,0 +1,43 @@
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
1
+ {
2
+ "_name_or_path": "/mnt/bn/wyt-large-dataset/xiabin-model/LLM/Qwen2-1.5B-Instruct",
3
+ "architectures": [
4
+ "Qwen2ForCausalLM"
5
+ ],
6
+ "attention_dropout": 0.0,
7
+ "bos_token_id": 151643,
8
+ "eos_token_id": 151645,
9
+ "freeze_mm_mlp_adapter": false,
10
+ "hidden_act": "silu",
11
+ "hidden_size": 1536,
12
+ "image_aspect_ratio": "resizesquare",
13
+ "initializer_range": 0.02,
14
+ "intermediate_size": 8960,
15
+ "max_position_embeddings": 32768,
16
+ "max_window_layers": 28,
17
+ "mm_hidden_size": 1024,
18
+ "mm_patch_merge_type": "flat",
19
+ "mm_projector_lr": null,
20
+ "mm_projector_type": "mlp2x_gelu",
21
+ "mm_use_im_patch_token": false,
22
+ "mm_use_im_start_end": false,
23
+ "mm_vision_select_feature": "patch",
24
+ "mm_vision_select_layer": -2,
25
+ "mm_vision_tower": "openai/clip-vit-large-patch14-336",
26
+ "model_type": "llava_Qwen2",
27
+ "num_attention_heads": 12,
28
+ "num_hidden_layers": 28,
29
+ "num_key_value_heads": 2,
30
+ "rms_norm_eps": 1e-06,
31
+ "rope_theta": 1000000.0,
32
+ "sliding_window": 32768,
33
+ "tie_word_embeddings": true,
34
+ "tokenizer_model_max_length": 2048,
35
+ "tokenizer_padding_side": "right",
36
+ "torch_dtype": "bfloat16",
37
+ "transformers_version": "4.41.2",
38
+ "tune_mm_mlp_adapter": true,
39
+ "use_cache": true,
40
+ "use_mm_proj": true,
41
+ "use_sliding_window": false,
42
+ "vocab_size": 151936
43
+ }
llmga-Qwen2-1.5B-pretrain/mm_projector.bin ADDED
@@ -0,0 +1,3 @@
 
 
 
 
1
+ version https://git-lfs.github.com/spec/v1
2
+ oid sha256:e7b3f2a6f26615b8d036b34f82dea91ceedc319ddc08649817cc94cc91319962
3
+ size 7872568
llmga-Qwen2-1.5B-pretrain/trainer_state.json ADDED
The diff for this file is too large to render. See raw diff
 
llmga-Qwen2-7B-pretrain/config.json ADDED
@@ -0,0 +1,43 @@
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
1
+ {
2
+ "_name_or_path": "/mnt/bn/wyt-large-dataset/xiabin-model/LLM/Qwen2-7B-Instruct",
3
+ "architectures": [
4
+ "Qwen2ForCausalLM"
5
+ ],
6
+ "attention_dropout": 0.0,
7
+ "bos_token_id": 151643,
8
+ "eos_token_id": 151645,
9
+ "freeze_mm_mlp_adapter": false,
10
+ "hidden_act": "silu",
11
+ "hidden_size": 3584,
12
+ "image_aspect_ratio": "resizesquare",
13
+ "initializer_range": 0.02,
14
+ "intermediate_size": 18944,
15
+ "max_position_embeddings": 32768,
16
+ "max_window_layers": 28,
17
+ "mm_hidden_size": 1024,
18
+ "mm_patch_merge_type": "flat",
19
+ "mm_projector_lr": null,
20
+ "mm_projector_type": "mlp2x_gelu",
21
+ "mm_use_im_patch_token": false,
22
+ "mm_use_im_start_end": false,
23
+ "mm_vision_select_feature": "patch",
24
+ "mm_vision_select_layer": -2,
25
+ "mm_vision_tower": "openai/clip-vit-large-patch14-336",
26
+ "model_type": "llava_Qwen2",
27
+ "num_attention_heads": 28,
28
+ "num_hidden_layers": 28,
29
+ "num_key_value_heads": 4,
30
+ "rms_norm_eps": 1e-06,
31
+ "rope_theta": 1000000.0,
32
+ "sliding_window": 131072,
33
+ "tie_word_embeddings": false,
34
+ "tokenizer_model_max_length": 2048,
35
+ "tokenizer_padding_side": "right",
36
+ "torch_dtype": "bfloat16",
37
+ "transformers_version": "4.41.2",
38
+ "tune_mm_mlp_adapter": true,
39
+ "use_cache": true,
40
+ "use_mm_proj": true,
41
+ "use_sliding_window": false,
42
+ "vocab_size": 152064
43
+ }
llmga-Qwen2-7B-pretrain/mm_projector.bin ADDED
@@ -0,0 +1,3 @@
 
 
 
 
1
+ version https://git-lfs.github.com/spec/v1
2
+ oid sha256:a07f1846da6840beca110eb0b79a6fc5deaf630196e87c006d8bfc8965a6d6db
3
+ size 33046584
llmga-Qwen2-7B-pretrain/trainer_state.json ADDED
The diff for this file is too large to render. See raw diff
 
llmga-gemma-pretrain/config.json ADDED
@@ -0,0 +1,44 @@
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
1
+ {
2
+ "_name_or_path": "/mnt/bn/xiabinpaintv2/gemma-2b-it",
3
+ "architectures": [
4
+ "GemmaForCausalLM"
5
+ ],
6
+ "attention_bias": false,
7
+ "attention_dropout": 0.0,
8
+ "bos_token_id": 2,
9
+ "eos_token_id": 1,
10
+ "freeze_mm_mlp_adapter": false,
11
+ "head_dim": 256,
12
+ "hidden_act": "gelu",
13
+ "hidden_activation": null,
14
+ "hidden_size": 2048,
15
+ "image_aspect_ratio": "resizesquare",
16
+ "initializer_range": 0.02,
17
+ "intermediate_size": 16384,
18
+ "max_position_embeddings": 8192,
19
+ "mm_hidden_size": 1024,
20
+ "mm_patch_merge_type": "flat",
21
+ "mm_projector_lr": null,
22
+ "mm_projector_type": "mlp2x_gelu",
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": -2,
27
+ "mm_vision_tower": "openai/clip-vit-large-patch14-336",
28
+ "model_type": "llava_gemma",
29
+ "num_attention_heads": 8,
30
+ "num_hidden_layers": 18,
31
+ "num_key_value_heads": 1,
32
+ "pad_token_id": 0,
33
+ "rms_norm_eps": 1e-06,
34
+ "rope_scaling": null,
35
+ "rope_theta": 10000.0,
36
+ "tokenizer_model_max_length": 2048,
37
+ "tokenizer_padding_side": "right",
38
+ "torch_dtype": "bfloat16",
39
+ "transformers_version": "4.39.3",
40
+ "tune_mm_mlp_adapter": true,
41
+ "use_cache": true,
42
+ "use_mm_proj": true,
43
+ "vocab_size": 256000
44
+ }
llmga-gemma-pretrain/mm_projector.bin ADDED
@@ -0,0 +1,3 @@
 
 
 
 
1
+ version https://git-lfs.github.com/spec/v1
2
+ oid sha256:25af5b2f31dab5a0564b20c4b654081eca9f9241c216d313c095de13fe3d522f
3
+ size 12593208
llmga-gemma-pretrain/trainer_state.json ADDED
The diff for this file is too large to render. See raw diff
 
llmga-llama3-8b-pretrain/config.json ADDED
@@ -0,0 +1,43 @@
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
1
+ {
2
+ "_name_or_path": "/mnt/bn/wyt-large-dataset/xiabin-dataset/Meta-Llama-3-8B-Instruct",
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": "resizesquare",
14
+ "initializer_range": 0.02,
15
+ "intermediate_size": 14336,
16
+ "max_position_embeddings": 8192,
17
+ "mm_hidden_size": 1024,
18
+ "mm_patch_merge_type": "flat",
19
+ "mm_projector_lr": null,
20
+ "mm_projector_type": "mlp2x_gelu",
21
+ "mm_use_im_patch_token": false,
22
+ "mm_use_im_start_end": false,
23
+ "mm_vision_select_feature": "patch",
24
+ "mm_vision_select_layer": -2,
25
+ "mm_vision_tower": "openai/clip-vit-large-patch14-336",
26
+ "model_type": "llava_llama",
27
+ "num_attention_heads": 32,
28
+ "num_hidden_layers": 32,
29
+ "num_key_value_heads": 8,
30
+ "pretraining_tp": 1,
31
+ "rms_norm_eps": 1e-05,
32
+ "rope_scaling": null,
33
+ "rope_theta": 500000.0,
34
+ "tie_word_embeddings": false,
35
+ "tokenizer_model_max_length": 2048,
36
+ "tokenizer_padding_side": "right",
37
+ "torch_dtype": "bfloat16",
38
+ "transformers_version": "4.40.1",
39
+ "tune_mm_mlp_adapter": true,
40
+ "use_cache": true,
41
+ "use_mm_proj": true,
42
+ "vocab_size": 128256
43
+ }
llmga-llama3-8b-pretrain/mm_projector.bin ADDED
@@ -0,0 +1,3 @@
 
 
 
 
1
+ version https://git-lfs.github.com/spec/v1
2
+ oid sha256:2ea4ae9b00c7ff66349b74372673f50829ffc3c0c6817584856792f5d3f43123
3
+ size 41961528
llmga-llama3-8b-pretrain/trainer_state.json ADDED
The diff for this file is too large to render. See raw diff
 
llmga-mistral-pretrain/config.json ADDED
@@ -0,0 +1,41 @@
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
1
+ {
2
+ "_name_or_path": "/mnt/bn/wyt-large-dataset/xiabin-dataset/Mistral-7B-Instruct-v0.2",
3
+ "architectures": [
4
+ "MistralForCausalLM"
5
+ ],
6
+ "attention_dropout": 0.0,
7
+ "bos_token_id": 1,
8
+ "eos_token_id": 2,
9
+ "freeze_mm_mlp_adapter": false,
10
+ "hidden_act": "silu",
11
+ "hidden_size": 4096,
12
+ "image_aspect_ratio": "resizesquare",
13
+ "initializer_range": 0.02,
14
+ "intermediate_size": 14336,
15
+ "max_position_embeddings": 32768,
16
+ "mm_hidden_size": 1024,
17
+ "mm_patch_merge_type": "flat",
18
+ "mm_projector_lr": null,
19
+ "mm_projector_type": "mlp2x_gelu",
20
+ "mm_use_im_patch_token": false,
21
+ "mm_use_im_start_end": false,
22
+ "mm_vision_select_feature": "patch",
23
+ "mm_vision_select_layer": -2,
24
+ "mm_vision_tower": "openai/clip-vit-large-patch14-336",
25
+ "model_type": "llava_mistral",
26
+ "num_attention_heads": 32,
27
+ "num_hidden_layers": 32,
28
+ "num_key_value_heads": 8,
29
+ "rms_norm_eps": 1e-05,
30
+ "rope_theta": 1000000.0,
31
+ "sliding_window": null,
32
+ "tie_word_embeddings": false,
33
+ "tokenizer_model_max_length": 2048,
34
+ "tokenizer_padding_side": "right",
35
+ "torch_dtype": "bfloat16",
36
+ "transformers_version": "4.41.2",
37
+ "tune_mm_mlp_adapter": true,
38
+ "use_cache": true,
39
+ "use_mm_proj": true,
40
+ "vocab_size": 32000
41
+ }
llmga-mistral-pretrain/mm_projector.bin ADDED
@@ -0,0 +1,3 @@
 
 
 
 
1
+ version https://git-lfs.github.com/spec/v1
2
+ oid sha256:9d1abee01ac5eef59c437e08e16735d38d43a24729bc713c6f315cac0d304c1a
3
+ size 41961528
llmga-mistral-pretrain/trainer_state.json ADDED
The diff for this file is too large to render. See raw diff
 
llmga-vicuna-13b-v1.5-pretrain/config.json ADDED
@@ -0,0 +1,41 @@
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
1
+ {
2
+ "_name_or_path": "/mnt/bn/inpainting-bytenas-lq/xiabin/LLMGA-modelv2/vicuna-v1.5/vicuna-13b-v1.5",
3
+ "architectures": [
4
+ "LlamaForCausalLM"
5
+ ],
6
+ "bos_token_id": 1,
7
+ "eos_token_id": 2,
8
+ "freeze_mm_mlp_adapter": false,
9
+ "hidden_act": "silu",
10
+ "hidden_size": 5120,
11
+ "image_aspect_ratio": "square",
12
+ "initializer_range": 0.02,
13
+ "intermediate_size": 13824,
14
+ "max_length": 4096,
15
+ "max_position_embeddings": 4096,
16
+ "mm_hidden_size": 1024,
17
+ "mm_projector_lr": null,
18
+ "mm_projector_type": "mlp2x_gelu",
19
+ "mm_use_im_patch_token": false,
20
+ "mm_use_im_start_end": false,
21
+ "mm_vision_select_feature": "patch",
22
+ "mm_vision_select_layer": -2,
23
+ "mm_vision_tower": "openai/clip-vit-large-patch14-336",
24
+ "model_type": "llava",
25
+ "num_attention_heads": 40,
26
+ "num_hidden_layers": 40,
27
+ "num_key_value_heads": 40,
28
+ "pad_token_id": 0,
29
+ "pretraining_tp": 1,
30
+ "rms_norm_eps": 1e-05,
31
+ "rope_scaling": null,
32
+ "tie_word_embeddings": false,
33
+ "tokenizer_model_max_length": 2048,
34
+ "tokenizer_padding_side": "right",
35
+ "torch_dtype": "float16",
36
+ "transformers_version": "4.31.0",
37
+ "tune_mm_mlp_adapter": true,
38
+ "use_cache": true,
39
+ "use_mm_proj": true,
40
+ "vocab_size": 32000
41
+ }
llmga-vicuna-13b-v1.5-pretrain/mm_projector.bin ADDED
@@ -0,0 +1,3 @@
 
 
 
 
1
+ version https://git-lfs.github.com/spec/v1
2
+ oid sha256:a76159a881736588cc972c48ce7f588782513c035503c2196456d2ba3cdfe4b6
3
+ size 62936701
llmga-vicuna-13b-v1.5-pretrain/trainer_state.json ADDED
The diff for this file is too large to render. See raw diff
 
llmga-vicuna-7b-v1.5-pretrain/config.json ADDED
@@ -0,0 +1,40 @@
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
1
+ {
2
+ "_name_or_path": "/mnt/bn/inpainting-bytenas-lq/xiabin/LLMGA-modelv2/vicuna-v1.5/vicuna-7b-v1.5",
3
+ "architectures": [
4
+ "LlamaForCausalLM"
5
+ ],
6
+ "bos_token_id": 1,
7
+ "eos_token_id": 2,
8
+ "freeze_mm_mlp_adapter": false,
9
+ "hidden_act": "silu",
10
+ "hidden_size": 4096,
11
+ "image_aspect_ratio": "square",
12
+ "initializer_range": 0.02,
13
+ "intermediate_size": 11008,
14
+ "max_position_embeddings": 4096,
15
+ "mm_hidden_size": 1024,
16
+ "mm_projector_lr": null,
17
+ "mm_projector_type": "mlp2x_gelu",
18
+ "mm_use_im_patch_token": false,
19
+ "mm_use_im_start_end": false,
20
+ "mm_vision_select_feature": "patch",
21
+ "mm_vision_select_layer": -2,
22
+ "mm_vision_tower": "openai/clip-vit-large-patch14-336",
23
+ "model_type": "llava",
24
+ "num_attention_heads": 32,
25
+ "num_hidden_layers": 32,
26
+ "num_key_value_heads": 32,
27
+ "pad_token_id": 0,
28
+ "pretraining_tp": 1,
29
+ "rms_norm_eps": 1e-05,
30
+ "rope_scaling": null,
31
+ "tie_word_embeddings": false,
32
+ "tokenizer_model_max_length": 2048,
33
+ "tokenizer_padding_side": "right",
34
+ "torch_dtype": "float16",
35
+ "transformers_version": "4.31.0",
36
+ "tune_mm_mlp_adapter": true,
37
+ "use_cache": true,
38
+ "use_mm_proj": true,
39
+ "vocab_size": 32000
40
+ }
llmga-vicuna-7b-v1.5-pretrain/mm_projector.bin ADDED
@@ -0,0 +1,3 @@
 
 
 
 
1
+ version https://git-lfs.github.com/spec/v1
2
+ oid sha256:ce13fa83b5415ebd74c64d1eb3683f49bcf0975aa699658d94972b3af4a659fc
3
+ size 41961085
llmga-vicuna-7b-v1.5-pretrain/trainer_state.json ADDED
The diff for this file is too large to render. See raw diff