Hugging Face's logo

1509cxt
/
Geo-TinyLLaVA

Safetensors
tinyllava
custom_code
Model card Files
xet
Community
1509cxt commited on
Commit
580dc8f
·
verified ·
1 Parent(s): 4e88625

init

Browse files
Files changed (35) hide show
  1. added_tokens.json +40 -0
  2. checkpoint-21520/added_tokens.json +40 -0
  3. checkpoint-21520/config.json +76 -0
  4. checkpoint-21520/configuration.py +116 -0
  5. checkpoint-21520/generation_config.json +7 -0
  6. checkpoint-21520/latest.txt +1 -0
  7. checkpoint-21520/merges.txt +0 -0
  8. checkpoint-21520/model-00001-of-00002.safetensors +3 -0
  9. checkpoint-21520/model-00002-of-00002.safetensors +3 -0
  10. checkpoint-21520/model.safetensors.index.json +912 -0
  11. checkpoint-21520/modeling_tinyllava_phi.py +624 -0
  12. checkpoint-21520/rng_state_0.pth +3 -0
  13. checkpoint-21520/rng_state_1.pth +3 -0
  14. checkpoint-21520/rng_state_2.pth +3 -0
  15. checkpoint-21520/rng_state_3.pth +3 -0
  16. checkpoint-21520/scheduler.pt +3 -0
  17. checkpoint-21520/special_tokens_map.json +30 -0
  18. checkpoint-21520/tokenizer_config.json +328 -0
  19. checkpoint-21520/trainer_state.json +0 -0
  20. checkpoint-21520/training_args.bin +3 -0
  21. checkpoint-21520/vocab.json +0 -0
  22. checkpoint-21520/zero_to_fp32.py +604 -0
  23. config.json +76 -0
  24. configuration.py +116 -0
  25. connector/pytorch_model.bin +3 -0
  26. language_model/config.json +34 -0
  27. language_model/pytorch_model.bin +3 -0
  28. log.txt +11 -0
  29. merges.txt +0 -0
  30. special_tokens_map.json +30 -0
  31. tokenizer_config.json +328 -0
  32. trainer_state.json +0 -0
  33. vision_tower/config.json +16 -0
  34. vision_tower/pytorch_model.bin +3 -0
  35. vocab.json +0 -0
added_tokens.json ADDED
@@ -0,0 +1,40 @@
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
1
+ {
2
+ "\t\t": 50294,
3
+ "\t\t\t": 50293,
4
+ "\t\t\t\t": 50292,
5
+ "\t\t\t\t\t": 50291,
6
+ "\t\t\t\t\t\t": 50290,
7
+ "\t\t\t\t\t\t\t": 50289,
8
+ "\t\t\t\t\t\t\t\t": 50288,
9
+ "\t\t\t\t\t\t\t\t\t": 50287,
10
+ " ": 50286,
11
+ " ": 50285,
12
+ " ": 50284,
13
+ " ": 50283,
14
+ " ": 50282,
15
+ " ": 50281,
16
+ " ": 50280,
17
+ " ": 50279,
18
+ " ": 50278,
19
+ " ": 50277,
20
+ " ": 50276,
21
+ " ": 50275,
22
+ " ": 50274,
23
+ " ": 50273,
24
+ " ": 50272,
25
+ " ": 50271,
26
+ " ": 50270,
27
+ " ": 50269,
28
+ " ": 50268,
29
+ " ": 50267,
30
+ " ": 50266,
31
+ " ": 50265,
32
+ " ": 50264,
33
+ " ": 50263,
34
+ " ": 50262,
35
+ " ": 50261,
36
+ " ": 50260,
37
+ " ": 50259,
38
+ " ": 50258,
39
+ " ": 50257
40
+ }
checkpoint-21520/added_tokens.json ADDED
@@ -0,0 +1,40 @@
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
1
+ {
2
+ "\t\t": 50294,
3
+ "\t\t\t": 50293,
4
+ "\t\t\t\t": 50292,
5
+ "\t\t\t\t\t": 50291,
6
+ "\t\t\t\t\t\t": 50290,
7
+ "\t\t\t\t\t\t\t": 50289,
8
+ "\t\t\t\t\t\t\t\t": 50288,
9
+ "\t\t\t\t\t\t\t\t\t": 50287,
10
+ " ": 50286,
11
+ " ": 50285,
12
+ " ": 50284,
13
+ " ": 50283,
14
+ " ": 50282,
15
+ " ": 50281,
16
+ " ": 50280,
17
+ " ": 50279,
18
+ " ": 50278,
19
+ " ": 50277,
20
+ " ": 50276,
21
+ " ": 50275,
22
+ " ": 50274,
23
+ " ": 50273,
24
+ " ": 50272,
25
+ " ": 50271,
26
+ " ": 50270,
27
+ " ": 50269,
28
+ " ": 50268,
29
+ " ": 50267,
30
+ " ": 50266,
31
+ " ": 50265,
32
+ " ": 50264,
33
+ " ": 50263,
34
+ " ": 50262,
35
+ " ": 50261,
36
+ " ": 50260,
37
+ " ": 50259,
38
+ " ": 50258,
39
+ " ": 50257
40
+ }
checkpoint-21520/config.json ADDED
@@ -0,0 +1,76 @@
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
1
+ {
2
+ "_name_or_path": "/home/work/workspace/checkpoints/tinyllava-phi2-siglip-3.1B",
3
+ "architectures": [
4
+ "TinyLlavaForConditionalGeneration"
5
+ ],
6
+ "auto_map": {
7
+ "AutoConfig": "configuration.TinyLlavaConfig",
8
+ "AutoModelForCausalLM": "modeling_tinyllava_phi.TinyLlavaForConditionalGeneration"
9
+ },
10
+ "cache_dir": null,
11
+ "connector_type": "mlp2x_gelu",
12
+ "hidden_size": 2560,
13
+ "ignore_index": -100,
14
+ "image_aspect_ratio": "square",
15
+ "image_token_index": -200,
16
+ "llm_model_name_or_path": "/home/work/workspace/checkpoints/phi-2",
17
+ "model_type": "tinyllava",
18
+ "num_queries": 128,
19
+ "num_resampler_layers": 3,
20
+ "pad_token": "<|endoftext|>",
21
+ "resampler_hidden_size": 768,
22
+ "text_config": {
23
+ "_name_or_path": "/home/work/workspace/checkpoints/phi-2",
24
+ "architectures": [
25
+ "PhiForCausalLM"
26
+ ],
27
+ "auto_map": {
28
+ "AutoConfig": "microsoft/phi-2--configuration_phi.PhiConfig",
29
+ "AutoModelForCausalLM": "microsoft/phi-2--modeling_phi.PhiForCausalLM"
30
+ },
31
+ "bos_token_id": 50256,
32
+ "embd_pdrop": 0.0,
33
+ "eos_token_id": 50256,
34
+ "hidden_act": "gelu_new",
35
+ "hidden_size": 2560,
36
+ "intermediate_size": 10240,
37
+ "layer_norm_eps": 1e-05,
38
+ "model_type": "phi",
39
+ "num_hidden_layers": 32,
40
+ "partial_rotary_factor": 0.4,
41
+ "qk_layernorm": false,
42
+ "resid_pdrop": 0.1,
43
+ "torch_dtype": "float16",
44
+ "vocab_size": 51200
45
+ },
46
+ "tokenizer_model_max_length": 3072,
47
+ "tokenizer_name_or_path": "/home/work/workspace/checkpoints/phi-2",
48
+ "tokenizer_padding_side": "right",
49
+ "tokenizer_use_fast": false,
50
+ "torch_dtype": "float16",
51
+ "transformers_version": "4.44.2",
52
+ "tune_type_connector": "full",
53
+ "tune_type_llm": "full",
54
+ "tune_type_vision_tower": "full",
55
+ "tune_vision_tower_from_layer": 0,
56
+ "use_cache": false,
57
+ "vision_config": {
58
+ "hidden_act": "gelu_pytorch_tanh",
59
+ "hidden_size": 1152,
60
+ "image_size": 384,
61
+ "intermediate_size": 4304,
62
+ "layer_norm_eps": 1e-06,
63
+ "model_name_or_path": "/home/work/workspace/checkpoints/siglip",
64
+ "model_name_or_path2": "",
65
+ "model_type": "siglip_vision_model",
66
+ "num_attention_heads": 16,
67
+ "num_hidden_layers": 27,
68
+ "patch_size": 14
69
+ },
70
+ "vision_feature_layer": -2,
71
+ "vision_feature_select_strategy": "patch",
72
+ "vision_hidden_size": 1152,
73
+ "vision_model_name_or_path": "/home/work/workspace/checkpoints/siglip",
74
+ "vision_model_name_or_path2": "",
75
+ "vocab_size": 51200
76
+ }
checkpoint-21520/configuration.py ADDED
@@ -0,0 +1,116 @@
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
1
+ from transformers import PretrainedConfig
2
+ from transformers import CONFIG_MAPPING
3
+ from transformers import AutoConfig
4
+
5
+ IGNORE_INDEX = -100
6
+ IMAGE_TOKEN_INDEX = -200
7
+ DEFAULT_IMAGE_TOKEN = "<image>"
8
+
9
+
10
+ class TinyLlavaConfig(PretrainedConfig):
11
+
12
+ model_type = "tinyllava"
13
+ def __init__(
14
+ self,
15
+ llm_model_name_or_path = '',
16
+ tokenizer_name_or_path = None,
17
+ vision_model_name_or_path = '',
18
+ vision_model_name_or_path2 = '',
19
+ connector_type = None,
20
+ text_config=None,
21
+ hidden_size=2048,
22
+ vocab_size=32000,
23
+ ignore_index=-100,
24
+ image_token_index=32000,
25
+ pad_token = None,
26
+ pad_token_id = None,
27
+ tokenizer_padding_side = 'right',
28
+ tokenizer_model_max_length = 2048,
29
+ vision_config = None,
30
+ vision_hidden_size = None,
31
+ vision_feature_layer = -2,
32
+ vision_feature_select_strategy = 'patch',
33
+ image_aspect_ratio = 'square',
34
+ resampler_hidden_size = None,
35
+ num_queries = None,
36
+ num_resampler_layers = None,
37
+ use_cache = False,
38
+ cache_dir = None,
39
+ tokenizer_use_fast = False,
40
+ tune_type_llm = 'frozen',
41
+ tune_type_connector = 'frozen',
42
+ tune_type_vision_tower = 'frozen',
43
+ tune_vision_tower_from_layer = -1,
44
+
45
+ **kwargs
46
+
47
+ ):
48
+ self.llm_model_name_or_path = llm_model_name_or_path
49
+ self.tokenizer_name_or_path = tokenizer_name_or_path or self.llm_model_name_or_path
50
+ self.vision_model_name_or_path = vision_model_name_or_path
51
+ self.vision_model_name_or_path2 = vision_model_name_or_path2
52
+ self.connector_type = connector_type
53
+ self.tune_type_llm = tune_type_llm
54
+ self.tune_type_connector = tune_type_connector
55
+ self.tune_type_vision_tower = tune_type_vision_tower
56
+ self.tune_vision_tower_from_layer = tune_vision_tower_from_layer
57
+
58
+ self.ignore_index = IGNORE_INDEX
59
+ self.image_token_index = IMAGE_TOKEN_INDEX
60
+ self.pad_token = pad_token
61
+ self.pad_token_id = pad_token_id
62
+ self.tokenizer_padding_side = tokenizer_padding_side
63
+ self.tokenizer_model_max_length = tokenizer_model_max_length
64
+ self.vision_feature_layer = vision_feature_layer
65
+ self.vision_feature_select_strategy = vision_feature_select_strategy
66
+ self.image_aspect_ratio = image_aspect_ratio
67
+ self.resampler_hidden_size = resampler_hidden_size
68
+ self.num_queries = num_queries
69
+ self.num_resampler_layers = num_resampler_layers
70
+ self.use_cache = use_cache
71
+ self.cache_dir = cache_dir
72
+ self.tokenizer_use_fast = tokenizer_use_fast
73
+ self._load_text_config(text_config)
74
+ self._load_vision_config(vision_config)
75
+
76
+ super().__init__(**kwargs)
77
+
78
+
79
+ def _load_text_config(self, text_config=None):
80
+ if self.llm_model_name_or_path is None or self.llm_model_name_or_path == '':
81
+ self.text_config = CONFIG_MAPPING['llama']()
82
+
83
+ else:
84
+ self.text_config = AutoConfig.from_pretrained(self.llm_model_name_or_path, trust_remote_code=True)
85
+ if text_config is not None:
86
+ self.text_config = self.text_config.from_dict(text_config)
87
+
88
+ self.hidden_size = getattr(self.text_config, 'hidden_size', getattr(self.text_config, 'model_dim', None))
89
+ self.vocab_size = getattr(self.text_config, 'vocab_size', None)
90
+
91
+
92
+
93
+ def _load_vision_config(self, vision_config=None):
94
+ if self.vision_model_name_or_path is None or self.vision_model_name_or_path == '':
95
+ self.vision_config = CONFIG_MAPPING['clip_vision_model'](
96
+ intermediate_size=4096,
97
+ hidden_size=1024,
98
+ patch_size=14,
99
+ image_size=336,
100
+ num_hidden_layers=24,
101
+ num_attention_heads=16,
102
+ vocab_size=32000,
103
+ projection_dim=768,
104
+ )
105
+
106
+ else:
107
+ self.vision_config = AutoConfig.from_pretrained(self.vision_model_name_or_path.split(':')[-1])
108
+ self.vision_config = getattr(self.vision_config, 'vision_config', self.vision_config)
109
+ if vision_config is not None:
110
+ self.vision_config = self.vision_config.from_dict(vision_config)
111
+
112
+ self.vision_config.model_name_or_path = self.vision_model_name_or_path.split(':')[-1]
113
+ self.vision_config.model_name_or_path2 = self.vision_model_name_or_path2.split(':')[-1]
114
+ self.vision_hidden_size = getattr(self.vision_config, 'hidden_size', None)
115
+
116
+
checkpoint-21520/generation_config.json ADDED
@@ -0,0 +1,7 @@
 
 
 
 
 
 
 
 
1
+ {
2
+ "_from_model_config": true,
3
+ "bos_token_id": 50256,
4
+ "eos_token_id": 50256,
5
+ "transformers_version": "4.44.2",
6
+ "use_cache": false
7
+ }
checkpoint-21520/latest.txt ADDED
@@ -0,0 +1 @@
 
 
1
+ global_step21520
checkpoint-21520/merges.txt ADDED
The diff for this file is too large to render. See raw diff
 
checkpoint-21520/model-00001-of-00002.safetensors ADDED
@@ -0,0 +1,3 @@
 
 
 
 
1
+ version https://git-lfs.github.com/spec/v1
2
+ oid sha256:48ca9b7f3ee7f9f0956ee0eed2b1f85eddd821598692b329aac4b69fba82b025
3
+ size 103022592
checkpoint-21520/model-00002-of-00002.safetensors ADDED
@@ -0,0 +1,3 @@
 
 
 
 
1
+ version https://git-lfs.github.com/spec/v1
2
+ oid sha256:bf92368a1b6f28c05c6e097181fb29fabee904b92fe334e308dddc4b939f141c
3
+ size 1439367824
checkpoint-21520/model.safetensors.index.json ADDED
@@ -0,0 +1,912 @@
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
1
+ {
2
+ "metadata": {
3
+ "total_size": 6434834560
4
+ },
5
+ "weight_map": {
6
+ "connector._connector.0.bias": "model-00002-of-00002.safetensors",
7
+ "connector._connector.0.weight": "model-00002-of-00002.safetensors",
8
+ "connector._connector.2.bias": "model-00002-of-00002.safetensors",
9
+ "connector._connector.2.weight": "model-00002-of-00002.safetensors",
10
+ "language_model.lm_head.bias": "model-00002-of-00002.safetensors",
11
+ "language_model.lm_head.weight": "model-00002-of-00002.safetensors",
12
+ "language_model.model.embed_tokens.weight": "model-00001-of-00002.safetensors",
13
+ "language_model.model.final_layernorm.bias": "model-00002-of-00002.safetensors",
14
+ "language_model.model.final_layernorm.weight": "model-00002-of-00002.safetensors",
15
+ "language_model.model.layers.0.input_layernorm.bias": "model-00001-of-00002.safetensors",
16
+ "language_model.model.layers.0.input_layernorm.weight": "model-00001-of-00002.safetensors",
17
+ "language_model.model.layers.0.mlp.fc1.bias": "model-00001-of-00002.safetensors",
18
+ "language_model.model.layers.0.mlp.fc1.weight": "model-00001-of-00002.safetensors",
19
+ "language_model.model.layers.0.mlp.fc2.bias": "model-00001-of-00002.safetensors",
20
+ "language_model.model.layers.0.mlp.fc2.weight": "model-00001-of-00002.safetensors",
21
+ "language_model.model.layers.0.self_attn.dense.bias": "model-00001-of-00002.safetensors",
22
+ "language_model.model.layers.0.self_attn.dense.weight": "model-00001-of-00002.safetensors",
23
+ "language_model.model.layers.0.self_attn.k_proj.bias": "model-00001-of-00002.safetensors",
24
+ "language_model.model.layers.0.self_attn.k_proj.weight": "model-00001-of-00002.safetensors",
25
+ "language_model.model.layers.0.self_attn.q_proj.bias": "model-00001-of-00002.safetensors",
26
+ "language_model.model.layers.0.self_attn.q_proj.weight": "model-00001-of-00002.safetensors",
27
+ "language_model.model.layers.0.self_attn.v_proj.bias": "model-00001-of-00002.safetensors",
28
+ "language_model.model.layers.0.self_attn.v_proj.weight": "model-00001-of-00002.safetensors",
29
+ "language_model.model.layers.1.input_layernorm.bias": "model-00001-of-00002.safetensors",
30
+ "language_model.model.layers.1.input_layernorm.weight": "model-00001-of-00002.safetensors",
31
+ "language_model.model.layers.1.mlp.fc1.bias": "model-00001-of-00002.safetensors",
32
+ "language_model.model.layers.1.mlp.fc1.weight": "model-00001-of-00002.safetensors",
33
+ "language_model.model.layers.1.mlp.fc2.bias": "model-00001-of-00002.safetensors",
34
+ "language_model.model.layers.1.mlp.fc2.weight": "model-00001-of-00002.safetensors",
35
+ "language_model.model.layers.1.self_attn.dense.bias": "model-00001-of-00002.safetensors",
36
+ "language_model.model.layers.1.self_attn.dense.weight": "model-00001-of-00002.safetensors",
37
+ "language_model.model.layers.1.self_attn.k_proj.bias": "model-00001-of-00002.safetensors",
38
+ "language_model.model.layers.1.self_attn.k_proj.weight": "model-00001-of-00002.safetensors",
39
+ "language_model.model.layers.1.self_attn.q_proj.bias": "model-00001-of-00002.safetensors",
40
+ "language_model.model.layers.1.self_attn.q_proj.weight": "model-00001-of-00002.safetensors",
41
+ "language_model.model.layers.1.self_attn.v_proj.bias": "model-00001-of-00002.safetensors",
42
+ "language_model.model.layers.1.self_attn.v_proj.weight": "model-00001-of-00002.safetensors",
43
+ "language_model.model.layers.10.input_layernorm.bias": "model-00001-of-00002.safetensors",
44
+ "language_model.model.layers.10.input_layernorm.weight": "model-00001-of-00002.safetensors",
45
+ "language_model.model.layers.10.mlp.fc1.bias": "model-00001-of-00002.safetensors",
46
+ "language_model.model.layers.10.mlp.fc1.weight": "model-00001-of-00002.safetensors",
47
+ "language_model.model.layers.10.mlp.fc2.bias": "model-00001-of-00002.safetensors",
48
+ "language_model.model.layers.10.mlp.fc2.weight": "model-00001-of-00002.safetensors",
49
+ "language_model.model.layers.10.self_attn.dense.bias": "model-00001-of-00002.safetensors",
50
+ "language_model.model.layers.10.self_attn.dense.weight": "model-00001-of-00002.safetensors",
51
+ "language_model.model.layers.10.self_attn.k_proj.bias": "model-00001-of-00002.safetensors",
52
+ "language_model.model.layers.10.self_attn.k_proj.weight": "model-00001-of-00002.safetensors",
53
+ "language_model.model.layers.10.self_attn.q_proj.bias": "model-00001-of-00002.safetensors",
54
+ "language_model.model.layers.10.self_attn.q_proj.weight": "model-00001-of-00002.safetensors",
55
+ "language_model.model.layers.10.self_attn.v_proj.bias": "model-00001-of-00002.safetensors",
56
+ "language_model.model.layers.10.self_attn.v_proj.weight": "model-00001-of-00002.safetensors",
57
+ "language_model.model.layers.11.input_layernorm.bias": "model-00001-of-00002.safetensors",
58
+ "language_model.model.layers.11.input_layernorm.weight": "model-00001-of-00002.safetensors",
59
+ "language_model.model.layers.11.mlp.fc1.bias": "model-00001-of-00002.safetensors",
60
+ "language_model.model.layers.11.mlp.fc1.weight": "model-00001-of-00002.safetensors",
61
+ "language_model.model.layers.11.mlp.fc2.bias": "model-00001-of-00002.safetensors",
62
+ "language_model.model.layers.11.mlp.fc2.weight": "model-00001-of-00002.safetensors",
63
+ "language_model.model.layers.11.self_attn.dense.bias": "model-00001-of-00002.safetensors",
64
+ "language_model.model.layers.11.self_attn.dense.weight": "model-00001-of-00002.safetensors",
65
+ "language_model.model.layers.11.self_attn.k_proj.bias": "model-00001-of-00002.safetensors",
66
+ "language_model.model.layers.11.self_attn.k_proj.weight": "model-00001-of-00002.safetensors",
67
+ "language_model.model.layers.11.self_attn.q_proj.bias": "model-00001-of-00002.safetensors",
68
+ "language_model.model.layers.11.self_attn.q_proj.weight": "model-00001-of-00002.safetensors",
69
+ "language_model.model.layers.11.self_attn.v_proj.bias": "model-00001-of-00002.safetensors",
70
+ "language_model.model.layers.11.self_attn.v_proj.weight": "model-00001-of-00002.safetensors",
71
+ "language_model.model.layers.12.input_layernorm.bias": "model-00001-of-00002.safetensors",
72
+ "language_model.model.layers.12.input_layernorm.weight": "model-00001-of-00002.safetensors",
73
+ "language_model.model.layers.12.mlp.fc1.bias": "model-00001-of-00002.safetensors",
74
+ "language_model.model.layers.12.mlp.fc1.weight": "model-00001-of-00002.safetensors",
75
+ "language_model.model.layers.12.mlp.fc2.bias": "model-00001-of-00002.safetensors",
76
+ "language_model.model.layers.12.mlp.fc2.weight": "model-00001-of-00002.safetensors",
77
+ "language_model.model.layers.12.self_attn.dense.bias": "model-00001-of-00002.safetensors",
78
+ "language_model.model.layers.12.self_attn.dense.weight": "model-00001-of-00002.safetensors",
79
+ "language_model.model.layers.12.self_attn.k_proj.bias": "model-00001-of-00002.safetensors",
80
+ "language_model.model.layers.12.self_attn.k_proj.weight": "model-00001-of-00002.safetensors",
81
+ "language_model.model.layers.12.self_attn.q_proj.bias": "model-00001-of-00002.safetensors",
82
+ "language_model.model.layers.12.self_attn.q_proj.weight": "model-00001-of-00002.safetensors",
83
+ "language_model.model.layers.12.self_attn.v_proj.bias": "model-00001-of-00002.safetensors",
84
+ "language_model.model.layers.12.self_attn.v_proj.weight": "model-00001-of-00002.safetensors",
85
+ "language_model.model.layers.13.input_layernorm.bias": "model-00001-of-00002.safetensors",
86
+ "language_model.model.layers.13.input_layernorm.weight": "model-00001-of-00002.safetensors",
87
+ "language_model.model.layers.13.mlp.fc1.bias": "model-00001-of-00002.safetensors",
88
+ "language_model.model.layers.13.mlp.fc1.weight": "model-00001-of-00002.safetensors",
89
+ "language_model.model.layers.13.mlp.fc2.bias": "model-00001-of-00002.safetensors",
90
+ "language_model.model.layers.13.mlp.fc2.weight": "model-00001-of-00002.safetensors",
91
+ "language_model.model.layers.13.self_attn.dense.bias": "model-00001-of-00002.safetensors",
92
+ "language_model.model.layers.13.self_attn.dense.weight": "model-00001-of-00002.safetensors",
93
+ "language_model.model.layers.13.self_attn.k_proj.bias": "model-00001-of-00002.safetensors",
94
+ "language_model.model.layers.13.self_attn.k_proj.weight": "model-00001-of-00002.safetensors",
95
+ "language_model.model.layers.13.self_attn.q_proj.bias": "model-00001-of-00002.safetensors",
96
+ "language_model.model.layers.13.self_attn.q_proj.weight": "model-00001-of-00002.safetensors",
97
+ "language_model.model.layers.13.self_attn.v_proj.bias": "model-00001-of-00002.safetensors",
98
+ "language_model.model.layers.13.self_attn.v_proj.weight": "model-00001-of-00002.safetensors",
99
+ "language_model.model.layers.14.input_layernorm.bias": "model-00001-of-00002.safetensors",
100
+ "language_model.model.layers.14.input_layernorm.weight": "model-00001-of-00002.safetensors",
101
+ "language_model.model.layers.14.mlp.fc1.bias": "model-00001-of-00002.safetensors",
102
+ "language_model.model.layers.14.mlp.fc1.weight": "model-00001-of-00002.safetensors",
103
+ "language_model.model.layers.14.mlp.fc2.bias": "model-00001-of-00002.safetensors",
104
+ "language_model.model.layers.14.mlp.fc2.weight": "model-00001-of-00002.safetensors",
105
+ "language_model.model.layers.14.self_attn.dense.bias": "model-00001-of-00002.safetensors",
106
+ "language_model.model.layers.14.self_attn.dense.weight": "model-00001-of-00002.safetensors",
107
+ "language_model.model.layers.14.self_attn.k_proj.bias": "model-00001-of-00002.safetensors",
108
+ "language_model.model.layers.14.self_attn.k_proj.weight": "model-00001-of-00002.safetensors",
109
+ "language_model.model.layers.14.self_attn.q_proj.bias": "model-00001-of-00002.safetensors",
110
+ "language_model.model.layers.14.self_attn.q_proj.weight": "model-00001-of-00002.safetensors",
111
+ "language_model.model.layers.14.self_attn.v_proj.bias": "model-00001-of-00002.safetensors",
112
+ "language_model.model.layers.14.self_attn.v_proj.weight": "model-00001-of-00002.safetensors",
113
+ "language_model.model.layers.15.input_layernorm.bias": "model-00001-of-00002.safetensors",
114
+ "language_model.model.layers.15.input_layernorm.weight": "model-00001-of-00002.safetensors",
115
+ "language_model.model.layers.15.mlp.fc1.bias": "model-00001-of-00002.safetensors",
116
+ "language_model.model.layers.15.mlp.fc1.weight": "model-00001-of-00002.safetensors",
117
+ "language_model.model.layers.15.mlp.fc2.bias": "model-00001-of-00002.safetensors",
118
+ "language_model.model.layers.15.mlp.fc2.weight": "model-00001-of-00002.safetensors",
119
+ "language_model.model.layers.15.self_attn.dense.bias": "model-00001-of-00002.safetensors",
120
+ "language_model.model.layers.15.self_attn.dense.weight": "model-00001-of-00002.safetensors",
121
+ "language_model.model.layers.15.self_attn.k_proj.bias": "model-00001-of-00002.safetensors",
122
+ "language_model.model.layers.15.self_attn.k_proj.weight": "model-00001-of-00002.safetensors",
123
+ "language_model.model.layers.15.self_attn.q_proj.bias": "model-00001-of-00002.safetensors",
124
+ "language_model.model.layers.15.self_attn.q_proj.weight": "model-00001-of-00002.safetensors",
125
+ "language_model.model.layers.15.self_attn.v_proj.bias": "model-00001-of-00002.safetensors",
126
+ "language_model.model.layers.15.self_attn.v_proj.weight": "model-00001-of-00002.safetensors",
127
+ "language_model.model.layers.16.input_layernorm.bias": "model-00001-of-00002.safetensors",
128
+ "language_model.model.layers.16.input_layernorm.weight": "model-00001-of-00002.safetensors",
129
+ "language_model.model.layers.16.mlp.fc1.bias": "model-00001-of-00002.safetensors",
130
+ "language_model.model.layers.16.mlp.fc1.weight": "model-00001-of-00002.safetensors",
131
+ "language_model.model.layers.16.mlp.fc2.bias": "model-00001-of-00002.safetensors",
132
+ "language_model.model.layers.16.mlp.fc2.weight": "model-00001-of-00002.safetensors",
133
+ "language_model.model.layers.16.self_attn.dense.bias": "model-00001-of-00002.safetensors",
134
+ "language_model.model.layers.16.self_attn.dense.weight": "model-00001-of-00002.safetensors",
135
+ "language_model.model.layers.16.self_attn.k_proj.bias": "model-00001-of-00002.safetensors",
136
+ "language_model.model.layers.16.self_attn.k_proj.weight": "model-00001-of-00002.safetensors",
137
+ "language_model.model.layers.16.self_attn.q_proj.bias": "model-00001-of-00002.safetensors",
138
+ "language_model.model.layers.16.self_attn.q_proj.weight": "model-00001-of-00002.safetensors",
139
+ "language_model.model.layers.16.self_attn.v_proj.bias": "model-00001-of-00002.safetensors",
140
+ "language_model.model.layers.16.self_attn.v_proj.weight": "model-00001-of-00002.safetensors",
141
+ "language_model.model.layers.17.input_layernorm.bias": "model-00001-of-00002.safetensors",
142
+ "language_model.model.layers.17.input_layernorm.weight": "model-00001-of-00002.safetensors",
143
+ "language_model.model.layers.17.mlp.fc1.bias": "model-00001-of-00002.safetensors",
144
+ "language_model.model.layers.17.mlp.fc1.weight": "model-00001-of-00002.safetensors",
145
+ "language_model.model.layers.17.mlp.fc2.bias": "model-00001-of-00002.safetensors",
146
+ "language_model.model.layers.17.mlp.fc2.weight": "model-00001-of-00002.safetensors",
147
+ "language_model.model.layers.17.self_attn.dense.bias": "model-00001-of-00002.safetensors",
148
+ "language_model.model.layers.17.self_attn.dense.weight": "model-00001-of-00002.safetensors",
149
+ "language_model.model.layers.17.self_attn.k_proj.bias": "model-00001-of-00002.safetensors",
150
+ "language_model.model.layers.17.self_attn.k_proj.weight": "model-00001-of-00002.safetensors",
151
+ "language_model.model.layers.17.self_attn.q_proj.bias": "model-00001-of-00002.safetensors",
152
+ "language_model.model.layers.17.self_attn.q_proj.weight": "model-00001-of-00002.safetensors",
153
+ "language_model.model.layers.17.self_attn.v_proj.bias": "model-00001-of-00002.safetensors",
154
+ "language_model.model.layers.17.self_attn.v_proj.weight": "model-00001-of-00002.safetensors",
155
+ "language_model.model.layers.18.input_layernorm.bias": "model-00001-of-00002.safetensors",
156
+ "language_model.model.layers.18.input_layernorm.weight": "model-00001-of-00002.safetensors",
157
+ "language_model.model.layers.18.mlp.fc1.bias": "model-00001-of-00002.safetensors",
158
+ "language_model.model.layers.18.mlp.fc1.weight": "model-00001-of-00002.safetensors",
159
+ "language_model.model.layers.18.mlp.fc2.bias": "model-00001-of-00002.safetensors",
160
+ "language_model.model.layers.18.mlp.fc2.weight": "model-00001-of-00002.safetensors",
161
+ "language_model.model.layers.18.self_attn.dense.bias": "model-00001-of-00002.safetensors",
162
+ "language_model.model.layers.18.self_attn.dense.weight": "model-00001-of-00002.safetensors",
163
+ "language_model.model.layers.18.self_attn.k_proj.bias": "model-00001-of-00002.safetensors",
164
+ "language_model.model.layers.18.self_attn.k_proj.weight": "model-00001-of-00002.safetensors",
165
+ "language_model.model.layers.18.self_attn.q_proj.bias": "model-00001-of-00002.safetensors",
166
+ "language_model.model.layers.18.self_attn.q_proj.weight": "model-00001-of-00002.safetensors",
167
+ "language_model.model.layers.18.self_attn.v_proj.bias": "model-00001-of-00002.safetensors",
168
+ "language_model.model.layers.18.self_attn.v_proj.weight": "model-00001-of-00002.safetensors",
169
+ "language_model.model.layers.19.input_layernorm.bias": "model-00001-of-00002.safetensors",
170
+ "language_model.model.layers.19.input_layernorm.weight": "model-00001-of-00002.safetensors",
171
+ "language_model.model.layers.19.mlp.fc1.bias": "model-00001-of-00002.safetensors",
172
+ "language_model.model.layers.19.mlp.fc1.weight": "model-00001-of-00002.safetensors",
173
+ "language_model.model.layers.19.mlp.fc2.bias": "model-00001-of-00002.safetensors",
174
+ "language_model.model.layers.19.mlp.fc2.weight": "model-00001-of-00002.safetensors",
175
+ "language_model.model.layers.19.self_attn.dense.bias": "model-00001-of-00002.safetensors",
176
+ "language_model.model.layers.19.self_attn.dense.weight": "model-00001-of-00002.safetensors",
177
+ "language_model.model.layers.19.self_attn.k_proj.bias": "model-00001-of-00002.safetensors",
178
+ "language_model.model.layers.19.self_attn.k_proj.weight": "model-00001-of-00002.safetensors",
179
+ "language_model.model.layers.19.self_attn.q_proj.bias": "model-00001-of-00002.safetensors",
180
+ "language_model.model.layers.19.self_attn.q_proj.weight": "model-00001-of-00002.safetensors",
181
+ "language_model.model.layers.19.self_attn.v_proj.bias": "model-00001-of-00002.safetensors",
182
+ "language_model.model.layers.19.self_attn.v_proj.weight": "model-00001-of-00002.safetensors",
183
+ "language_model.model.layers.2.input_layernorm.bias": "model-00001-of-00002.safetensors",
184
+ "language_model.model.layers.2.input_layernorm.weight": "model-00001-of-00002.safetensors",
185
+ "language_model.model.layers.2.mlp.fc1.bias": "model-00001-of-00002.safetensors",
186
+ "language_model.model.layers.2.mlp.fc1.weight": "model-00001-of-00002.safetensors",
187
+ "language_model.model.layers.2.mlp.fc2.bias": "model-00001-of-00002.safetensors",
188
+ "language_model.model.layers.2.mlp.fc2.weight": "model-00001-of-00002.safetensors",
189
+ "language_model.model.layers.2.self_attn.dense.bias": "model-00001-of-00002.safetensors",
190
+ "language_model.model.layers.2.self_attn.dense.weight": "model-00001-of-00002.safetensors",
191
+ "language_model.model.layers.2.self_attn.k_proj.bias": "model-00001-of-00002.safetensors",
192
+ "language_model.model.layers.2.self_attn.k_proj.weight": "model-00001-of-00002.safetensors",
193
+ "language_model.model.layers.2.self_attn.q_proj.bias": "model-00001-of-00002.safetensors",
194
+ "language_model.model.layers.2.self_attn.q_proj.weight": "model-00001-of-00002.safetensors",
195
+ "language_model.model.layers.2.self_attn.v_proj.bias": "model-00001-of-00002.safetensors",
196
+ "language_model.model.layers.2.self_attn.v_proj.weight": "model-00001-of-00002.safetensors",
197
+ "language_model.model.layers.20.input_layernorm.bias": "model-00001-of-00002.safetensors",
198
+ "language_model.model.layers.20.input_layernorm.weight": "model-00001-of-00002.safetensors",
199
+ "language_model.model.layers.20.mlp.fc1.bias": "model-00001-of-00002.safetensors",
200
+ "language_model.model.layers.20.mlp.fc1.weight": "model-00001-of-00002.safetensors",
201
+ "language_model.model.layers.20.mlp.fc2.bias": "model-00001-of-00002.safetensors",
202
+ "language_model.model.layers.20.mlp.fc2.weight": "model-00001-of-00002.safetensors",
203
+ "language_model.model.layers.20.self_attn.dense.bias": "model-00001-of-00002.safetensors",
204
+ "language_model.model.layers.20.self_attn.dense.weight": "model-00001-of-00002.safetensors",
205
+ "language_model.model.layers.20.self_attn.k_proj.bias": "model-00001-of-00002.safetensors",
206
+ "language_model.model.layers.20.self_attn.k_proj.weight": "model-00001-of-00002.safetensors",
207
+ "language_model.model.layers.20.self_attn.q_proj.bias": "model-00001-of-00002.safetensors",
208
+ "language_model.model.layers.20.self_attn.q_proj.weight": "model-00001-of-00002.safetensors",
209
+ "language_model.model.layers.20.self_attn.v_proj.bias": "model-00001-of-00002.safetensors",
210
+ "language_model.model.layers.20.self_attn.v_proj.weight": "model-00001-of-00002.safetensors",
211
+ "language_model.model.layers.21.input_layernorm.bias": "model-00001-of-00002.safetensors",
212
+ "language_model.model.layers.21.input_layernorm.weight": "model-00001-of-00002.safetensors",
213
+ "language_model.model.layers.21.mlp.fc1.bias": "model-00001-of-00002.safetensors",
214
+ "language_model.model.layers.21.mlp.fc1.weight": "model-00001-of-00002.safetensors",
215
+ "language_model.model.layers.21.mlp.fc2.bias": "model-00001-of-00002.safetensors",
216
+ "language_model.model.layers.21.mlp.fc2.weight": "model-00001-of-00002.safetensors",
217
+ "language_model.model.layers.21.self_attn.dense.bias": "model-00001-of-00002.safetensors",
218
+ "language_model.model.layers.21.self_attn.dense.weight": "model-00001-of-00002.safetensors",
219
+ "language_model.model.layers.21.self_attn.k_proj.bias": "model-00001-of-00002.safetensors",
220
+ "language_model.model.layers.21.self_attn.k_proj.weight": "model-00001-of-00002.safetensors",
221
+ "language_model.model.layers.21.self_attn.q_proj.bias": "model-00001-of-00002.safetensors",
222
+ "language_model.model.layers.21.self_attn.q_proj.weight": "model-00001-of-00002.safetensors",
223
+ "language_model.model.layers.21.self_attn.v_proj.bias": "model-00001-of-00002.safetensors",
224
+ "language_model.model.layers.21.self_attn.v_proj.weight": "model-00001-of-00002.safetensors",
225
+ "language_model.model.layers.22.input_layernorm.bias": "model-00001-of-00002.safetensors",
226
+ "language_model.model.layers.22.input_layernorm.weight": "model-00001-of-00002.safetensors",
227
+ "language_model.model.layers.22.mlp.fc1.bias": "model-00001-of-00002.safetensors",
228
+ "language_model.model.layers.22.mlp.fc1.weight": "model-00001-of-00002.safetensors",
229
+ "language_model.model.layers.22.mlp.fc2.bias": "model-00001-of-00002.safetensors",
230
+ "language_model.model.layers.22.mlp.fc2.weight": "model-00001-of-00002.safetensors",
231
+ "language_model.model.layers.22.self_attn.dense.bias": "model-00001-of-00002.safetensors",
232
+ "language_model.model.layers.22.self_attn.dense.weight": "model-00001-of-00002.safetensors",
233
+ "language_model.model.layers.22.self_attn.k_proj.bias": "model-00001-of-00002.safetensors",
234
+ "language_model.model.layers.22.self_attn.k_proj.weight": "model-00001-of-00002.safetensors",
235
+ "language_model.model.layers.22.self_attn.q_proj.bias": "model-00001-of-00002.safetensors",
236
+ "language_model.model.layers.22.self_attn.q_proj.weight": "model-00001-of-00002.safetensors",
237
+ "language_model.model.layers.22.self_attn.v_proj.bias": "model-00001-of-00002.safetensors",
238
+ "language_model.model.layers.22.self_attn.v_proj.weight": "model-00001-of-00002.safetensors",
239
+ "language_model.model.layers.23.input_layernorm.bias": "model-00001-of-00002.safetensors",
240
+ "language_model.model.layers.23.input_layernorm.weight": "model-00001-of-00002.safetensors",
241
+ "language_model.model.layers.23.mlp.fc1.bias": "model-00001-of-00002.safetensors",
242
+ "language_model.model.layers.23.mlp.fc1.weight": "model-00001-of-00002.safetensors",
243
+ "language_model.model.layers.23.mlp.fc2.bias": "model-00001-of-00002.safetensors",
244
+ "language_model.model.layers.23.mlp.fc2.weight": "model-00001-of-00002.safetensors",
245
+ "language_model.model.layers.23.self_attn.dense.bias": "model-00001-of-00002.safetensors",
246
+ "language_model.model.layers.23.self_attn.dense.weight": "model-00001-of-00002.safetensors",
247
+ "language_model.model.layers.23.self_attn.k_proj.bias": "model-00001-of-00002.safetensors",
248
+ "language_model.model.layers.23.self_attn.k_proj.weight": "model-00001-of-00002.safetensors",
249
+ "language_model.model.layers.23.self_attn.q_proj.bias": "model-00001-of-00002.safetensors",
250
+ "language_model.model.layers.23.self_attn.q_proj.weight": "model-00001-of-00002.safetensors",
251
+ "language_model.model.layers.23.self_attn.v_proj.bias": "model-00001-of-00002.safetensors",
252
+ "language_model.model.layers.23.self_attn.v_proj.weight": "model-00001-of-00002.safetensors",
253
+ "language_model.model.layers.24.input_layernorm.bias": "model-00001-of-00002.safetensors",
254
+ "language_model.model.layers.24.input_layernorm.weight": "model-00001-of-00002.safetensors",
255
+ "language_model.model.layers.24.mlp.fc1.bias": "model-00001-of-00002.safetensors",
256
+ "language_model.model.layers.24.mlp.fc1.weight": "model-00001-of-00002.safetensors",
257
+ "language_model.model.layers.24.mlp.fc2.bias": "model-00001-of-00002.safetensors",
258
+ "language_model.model.layers.24.mlp.fc2.weight": "model-00001-of-00002.safetensors",
259
+ "language_model.model.layers.24.self_attn.dense.bias": "model-00001-of-00002.safetensors",
260
+ "language_model.model.layers.24.self_attn.dense.weight": "model-00001-of-00002.safetensors",
261
+ "language_model.model.layers.24.self_attn.k_proj.bias": "model-00001-of-00002.safetensors",
262
+ "language_model.model.layers.24.self_attn.k_proj.weight": "model-00001-of-00002.safetensors",
263
+ "language_model.model.layers.24.self_attn.q_proj.bias": "model-00001-of-00002.safetensors",
264
+ "language_model.model.layers.24.self_attn.q_proj.weight": "model-00001-of-00002.safetensors",
265
+ "language_model.model.layers.24.self_attn.v_proj.bias": "model-00001-of-00002.safetensors",
266
+ "language_model.model.layers.24.self_attn.v_proj.weight": "model-00001-of-00002.safetensors",
267
+ "language_model.model.layers.25.input_layernorm.bias": "model-00001-of-00002.safetensors",
268
+ "language_model.model.layers.25.input_layernorm.weight": "model-00001-of-00002.safetensors",
269
+ "language_model.model.layers.25.mlp.fc1.bias": "model-00001-of-00002.safetensors",
270
+ "language_model.model.layers.25.mlp.fc1.weight": "model-00001-of-00002.safetensors",
271
+ "language_model.model.layers.25.mlp.fc2.bias": "model-00001-of-00002.safetensors",
272
+ "language_model.model.layers.25.mlp.fc2.weight": "model-00001-of-00002.safetensors",
273
+ "language_model.model.layers.25.self_attn.dense.bias": "model-00001-of-00002.safetensors",
274
+ "language_model.model.layers.25.self_attn.dense.weight": "model-00001-of-00002.safetensors",
275
+ "language_model.model.layers.25.self_attn.k_proj.bias": "model-00001-of-00002.safetensors",
276
+ "language_model.model.layers.25.self_attn.k_proj.weight": "model-00001-of-00002.safetensors",
277
+ "language_model.model.layers.25.self_attn.q_proj.bias": "model-00001-of-00002.safetensors",
278
+ "language_model.model.layers.25.self_attn.q_proj.weight": "model-00001-of-00002.safetensors",
279
+ "language_model.model.layers.25.self_attn.v_proj.bias": "model-00001-of-00002.safetensors",
280
+ "language_model.model.layers.25.self_attn.v_proj.weight": "model-00001-of-00002.safetensors",
281
+ "language_model.model.layers.26.input_layernorm.bias": "model-00001-of-00002.safetensors",
282
+ "language_model.model.layers.26.input_layernorm.weight": "model-00001-of-00002.safetensors",
283
+ "language_model.model.layers.26.mlp.fc1.bias": "model-00001-of-00002.safetensors",
284
+ "language_model.model.layers.26.mlp.fc1.weight": "model-00001-of-00002.safetensors",
285
+ "language_model.model.layers.26.mlp.fc2.bias": "model-00001-of-00002.safetensors",
286
+ "language_model.model.layers.26.mlp.fc2.weight": "model-00001-of-00002.safetensors",
287
+ "language_model.model.layers.26.self_attn.dense.bias": "model-00001-of-00002.safetensors",
288
+ "language_model.model.layers.26.self_attn.dense.weight": "model-00001-of-00002.safetensors",
289
+ "language_model.model.layers.26.self_attn.k_proj.bias": "model-00001-of-00002.safetensors",
290
+ "language_model.model.layers.26.self_attn.k_proj.weight": "model-00001-of-00002.safetensors",
291
+ "language_model.model.layers.26.self_attn.q_proj.bias": "model-00001-of-00002.safetensors",
292
+ "language_model.model.layers.26.self_attn.q_proj.weight": "model-00001-of-00002.safetensors",
293
+ "language_model.model.layers.26.self_attn.v_proj.bias": "model-00001-of-00002.safetensors",
294
+ "language_model.model.layers.26.self_attn.v_proj.weight": "model-00001-of-00002.safetensors",
295
+ "language_model.model.layers.27.input_layernorm.bias": "model-00001-of-00002.safetensors",
296
+ "language_model.model.layers.27.input_layernorm.weight": "model-00001-of-00002.safetensors",
297
+ "language_model.model.layers.27.mlp.fc1.bias": "model-00001-of-00002.safetensors",
298
+ "language_model.model.layers.27.mlp.fc1.weight": "model-00001-of-00002.safetensors",
299
+ "language_model.model.layers.27.mlp.fc2.bias": "model-00001-of-00002.safetensors",
300
+ "language_model.model.layers.27.mlp.fc2.weight": "model-00001-of-00002.safetensors",
301
+ "language_model.model.layers.27.self_attn.dense.bias": "model-00001-of-00002.safetensors",
302
+ "language_model.model.layers.27.self_attn.dense.weight": "model-00001-of-00002.safetensors",
303
+ "language_model.model.layers.27.self_attn.k_proj.bias": "model-00001-of-00002.safetensors",
304
+ "language_model.model.layers.27.self_attn.k_proj.weight": "model-00001-of-00002.safetensors",
305
+ "language_model.model.layers.27.self_attn.q_proj.bias": "model-00001-of-00002.safetensors",
306
+ "language_model.model.layers.27.self_attn.q_proj.weight": "model-00001-of-00002.safetensors",
307
+ "language_model.model.layers.27.self_attn.v_proj.bias": "model-00001-of-00002.safetensors",
308
+ "language_model.model.layers.27.self_attn.v_proj.weight": "model-00001-of-00002.safetensors",
309
+ "language_model.model.layers.28.input_layernorm.bias": "model-00001-of-00002.safetensors",
310
+ "language_model.model.layers.28.input_layernorm.weight": "model-00001-of-00002.safetensors",
311
+ "language_model.model.layers.28.mlp.fc1.bias": "model-00001-of-00002.safetensors",
312
+ "language_model.model.layers.28.mlp.fc1.weight": "model-00001-of-00002.safetensors",
313
+ "language_model.model.layers.28.mlp.fc2.bias": "model-00001-of-00002.safetensors",
314
+ "language_model.model.layers.28.mlp.fc2.weight": "model-00001-of-00002.safetensors",
315
+ "language_model.model.layers.28.self_attn.dense.bias": "model-00001-of-00002.safetensors",
316
+ "language_model.model.layers.28.self_attn.dense.weight": "model-00001-of-00002.safetensors",
317
+ "language_model.model.layers.28.self_attn.k_proj.bias": "model-00001-of-00002.safetensors",
318
+ "language_model.model.layers.28.self_attn.k_proj.weight": "model-00001-of-00002.safetensors",
319
+ "language_model.model.layers.28.self_attn.q_proj.bias": "model-00001-of-00002.safetensors",
320
+ "language_model.model.layers.28.self_attn.q_proj.weight": "model-00001-of-00002.safetensors",
321
+ "language_model.model.layers.28.self_attn.v_proj.bias": "model-00001-of-00002.safetensors",
322
+ "language_model.model.layers.28.self_attn.v_proj.weight": "model-00001-of-00002.safetensors",
323
+ "language_model.model.layers.29.input_layernorm.bias": "model-00001-of-00002.safetensors",
324
+ "language_model.model.layers.29.input_layernorm.weight": "model-00001-of-00002.safetensors",
325
+ "language_model.model.layers.29.mlp.fc1.bias": "model-00001-of-00002.safetensors",
326
+ "language_model.model.layers.29.mlp.fc1.weight": "model-00001-of-00002.safetensors",
327
+ "language_model.model.layers.29.mlp.fc2.bias": "model-00001-of-00002.safetensors",
328
+ "language_model.model.layers.29.mlp.fc2.weight": "model-00001-of-00002.safetensors",
329
+ "language_model.model.layers.29.self_attn.dense.bias": "model-00001-of-00002.safetensors",
330
+ "language_model.model.layers.29.self_attn.dense.weight": "model-00001-of-00002.safetensors",
331
+ "language_model.model.layers.29.self_attn.k_proj.bias": "model-00001-of-00002.safetensors",
332
+ "language_model.model.layers.29.self_attn.k_proj.weight": "model-00001-of-00002.safetensors",
333
+ "language_model.model.layers.29.self_attn.q_proj.bias": "model-00001-of-00002.safetensors",
334
+ "language_model.model.layers.29.self_attn.q_proj.weight": "model-00001-of-00002.safetensors",
335
+ "language_model.model.layers.29.self_attn.v_proj.bias": "model-00001-of-00002.safetensors",
336
+ "language_model.model.layers.29.self_attn.v_proj.weight": "model-00001-of-00002.safetensors",
337
+ "language_model.model.layers.3.input_layernorm.bias": "model-00001-of-00002.safetensors",
338
+ "language_model.model.layers.3.input_layernorm.weight": "model-00001-of-00002.safetensors",
339
+ "language_model.model.layers.3.mlp.fc1.bias": "model-00001-of-00002.safetensors",
340
+ "language_model.model.layers.3.mlp.fc1.weight": "model-00001-of-00002.safetensors",
341
+ "language_model.model.layers.3.mlp.fc2.bias": "model-00001-of-00002.safetensors",
342
+ "language_model.model.layers.3.mlp.fc2.weight": "model-00001-of-00002.safetensors",
343
+ "language_model.model.layers.3.self_attn.dense.bias": "model-00001-of-00002.safetensors",
344
+ "language_model.model.layers.3.self_attn.dense.weight": "model-00001-of-00002.safetensors",
345
+ "language_model.model.layers.3.self_attn.k_proj.bias": "model-00001-of-00002.safetensors",
346
+ "language_model.model.layers.3.self_attn.k_proj.weight": "model-00001-of-00002.safetensors",
347
+ "language_model.model.layers.3.self_attn.q_proj.bias": "model-00001-of-00002.safetensors",
348
+ "language_model.model.layers.3.self_attn.q_proj.weight": "model-00001-of-00002.safetensors",
349
+ "language_model.model.layers.3.self_attn.v_proj.bias": "model-00001-of-00002.safetensors",
350
+ "language_model.model.layers.3.self_attn.v_proj.weight": "model-00001-of-00002.safetensors",
351
+ "language_model.model.layers.30.input_layernorm.bias": "model-00002-of-00002.safetensors",
352
+ "language_model.model.layers.30.input_layernorm.weight": "model-00002-of-00002.safetensors",
353
+ "language_model.model.layers.30.mlp.fc1.bias": "model-00002-of-00002.safetensors",
354
+ "language_model.model.layers.30.mlp.fc1.weight": "model-00002-of-00002.safetensors",
355
+ "language_model.model.layers.30.mlp.fc2.bias": "model-00002-of-00002.safetensors",
356
+ "language_model.model.layers.30.mlp.fc2.weight": "model-00002-of-00002.safetensors",
357
+ "language_model.model.layers.30.self_attn.dense.bias": "model-00002-of-00002.safetensors",
358
+ "language_model.model.layers.30.self_attn.dense.weight": "model-00002-of-00002.safetensors",
359
+ "language_model.model.layers.30.self_attn.k_proj.bias": "model-00002-of-00002.safetensors",
360
+ "language_model.model.layers.30.self_attn.k_proj.weight": "model-00002-of-00002.safetensors",
361
+ "language_model.model.layers.30.self_attn.q_proj.bias": "model-00001-of-00002.safetensors",
362
+ "language_model.model.layers.30.self_attn.q_proj.weight": "model-00001-of-00002.safetensors",
363
+ "language_model.model.layers.30.self_attn.v_proj.bias": "model-00002-of-00002.safetensors",
364
+ "language_model.model.layers.30.self_attn.v_proj.weight": "model-00002-of-00002.safetensors",
365
+ "language_model.model.layers.31.input_layernorm.bias": "model-00002-of-00002.safetensors",
366
+ "language_model.model.layers.31.input_layernorm.weight": "model-00002-of-00002.safetensors",
367
+ "language_model.model.layers.31.mlp.fc1.bias": "model-00002-of-00002.safetensors",
368
+ "language_model.model.layers.31.mlp.fc1.weight": "model-00002-of-00002.safetensors",
369
+ "language_model.model.layers.31.mlp.fc2.bias": "model-00002-of-00002.safetensors",
370
+ "language_model.model.layers.31.mlp.fc2.weight": "model-00002-of-00002.safetensors",
371
+ "language_model.model.layers.31.self_attn.dense.bias": "model-00002-of-00002.safetensors",
372
+ "language_model.model.layers.31.self_attn.dense.weight": "model-00002-of-00002.safetensors",
373
+ "language_model.model.layers.31.self_attn.k_proj.bias": "model-00002-of-00002.safetensors",
374
+ "language_model.model.layers.31.self_attn.k_proj.weight": "model-00002-of-00002.safetensors",
375
+ "language_model.model.layers.31.self_attn.q_proj.bias": "model-00002-of-00002.safetensors",
376
+ "language_model.model.layers.31.self_attn.q_proj.weight": "model-00002-of-00002.safetensors",
377
+ "language_model.model.layers.31.self_attn.v_proj.bias": "model-00002-of-00002.safetensors",
378
+ "language_model.model.layers.31.self_attn.v_proj.weight": "model-00002-of-00002.safetensors",
379
+ "language_model.model.layers.4.input_layernorm.bias": "model-00001-of-00002.safetensors",
380
+ "language_model.model.layers.4.input_layernorm.weight": "model-00001-of-00002.safetensors",
381
+ "language_model.model.layers.4.mlp.fc1.bias": "model-00001-of-00002.safetensors",
382
+ "language_model.model.layers.4.mlp.fc1.weight": "model-00001-of-00002.safetensors",
383
+ "language_model.model.layers.4.mlp.fc2.bias": "model-00001-of-00002.safetensors",
384
+ "language_model.model.layers.4.mlp.fc2.weight": "model-00001-of-00002.safetensors",
385
+ "language_model.model.layers.4.self_attn.dense.bias": "model-00001-of-00002.safetensors",
386
+ "language_model.model.layers.4.self_attn.dense.weight": "model-00001-of-00002.safetensors",
387
+ "language_model.model.layers.4.self_attn.k_proj.bias": "model-00001-of-00002.safetensors",
388
+ "language_model.model.layers.4.self_attn.k_proj.weight": "model-00001-of-00002.safetensors",
389
+ "language_model.model.layers.4.self_attn.q_proj.bias": "model-00001-of-00002.safetensors",
390
+ "language_model.model.layers.4.self_attn.q_proj.weight": "model-00001-of-00002.safetensors",
391
+ "language_model.model.layers.4.self_attn.v_proj.bias": "model-00001-of-00002.safetensors",
392
+ "language_model.model.layers.4.self_attn.v_proj.weight": "model-00001-of-00002.safetensors",
393
+ "language_model.model.layers.5.input_layernorm.bias": "model-00001-of-00002.safetensors",
394
+ "language_model.model.layers.5.input_layernorm.weight": "model-00001-of-00002.safetensors",
395
+ "language_model.model.layers.5.mlp.fc1.bias": "model-00001-of-00002.safetensors",
396
+ "language_model.model.layers.5.mlp.fc1.weight": "model-00001-of-00002.safetensors",
397
+ "language_model.model.layers.5.mlp.fc2.bias": "model-00001-of-00002.safetensors",
398
+ "language_model.model.layers.5.mlp.fc2.weight": "model-00001-of-00002.safetensors",
399
+ "language_model.model.layers.5.self_attn.dense.bias": "model-00001-of-00002.safetensors",
400
+ "language_model.model.layers.5.self_attn.dense.weight": "model-00001-of-00002.safetensors",
401
+ "language_model.model.layers.5.self_attn.k_proj.bias": "model-00001-of-00002.safetensors",
402
+ "language_model.model.layers.5.self_attn.k_proj.weight": "model-00001-of-00002.safetensors",
403
+ "language_model.model.layers.5.self_attn.q_proj.bias": "model-00001-of-00002.safetensors",
404
+ "language_model.model.layers.5.self_attn.q_proj.weight": "model-00001-of-00002.safetensors",
405
+ "language_model.model.layers.5.self_attn.v_proj.bias": "model-00001-of-00002.safetensors",
406
+ "language_model.model.layers.5.self_attn.v_proj.weight": "model-00001-of-00002.safetensors",
407
+ "language_model.model.layers.6.input_layernorm.bias": "model-00001-of-00002.safetensors",
408
+ "language_model.model.layers.6.input_layernorm.weight": "model-00001-of-00002.safetensors",
409
+ "language_model.model.layers.6.mlp.fc1.bias": "model-00001-of-00002.safetensors",
410
+ "language_model.model.layers.6.mlp.fc1.weight": "model-00001-of-00002.safetensors",
411
+ "language_model.model.layers.6.mlp.fc2.bias": "model-00001-of-00002.safetensors",
412
+ "language_model.model.layers.6.mlp.fc2.weight": "model-00001-of-00002.safetensors",
413
+ "language_model.model.layers.6.self_attn.dense.bias": "model-00001-of-00002.safetensors",
414
+ "language_model.model.layers.6.self_attn.dense.weight": "model-00001-of-00002.safetensors",
415
+ "language_model.model.layers.6.self_attn.k_proj.bias": "model-00001-of-00002.safetensors",
416
+ "language_model.model.layers.6.self_attn.k_proj.weight": "model-00001-of-00002.safetensors",
417
+ "language_model.model.layers.6.self_attn.q_proj.bias": "model-00001-of-00002.safetensors",
418
+ "language_model.model.layers.6.self_attn.q_proj.weight": "model-00001-of-00002.safetensors",
419
+ "language_model.model.layers.6.self_attn.v_proj.bias": "model-00001-of-00002.safetensors",
420
+ "language_model.model.layers.6.self_attn.v_proj.weight": "model-00001-of-00002.safetensors",
421
+ "language_model.model.layers.7.input_layernorm.bias": "model-00001-of-00002.safetensors",
422
+ "language_model.model.layers.7.input_layernorm.weight": "model-00001-of-00002.safetensors",
423
+ "language_model.model.layers.7.mlp.fc1.bias": "model-00001-of-00002.safetensors",
424
+ "language_model.model.layers.7.mlp.fc1.weight": "model-00001-of-00002.safetensors",
425
+ "language_model.model.layers.7.mlp.fc2.bias": "model-00001-of-00002.safetensors",
426
+ "language_model.model.layers.7.mlp.fc2.weight": "model-00001-of-00002.safetensors",
427
+ "language_model.model.layers.7.self_attn.dense.bias": "model-00001-of-00002.safetensors",
428
+ "language_model.model.layers.7.self_attn.dense.weight": "model-00001-of-00002.safetensors",
429
+ "language_model.model.layers.7.self_attn.k_proj.bias": "model-00001-of-00002.safetensors",
430
+ "language_model.model.layers.7.self_attn.k_proj.weight": "model-00001-of-00002.safetensors",
431
+ "language_model.model.layers.7.self_attn.q_proj.bias": "model-00001-of-00002.safetensors",
432
+ "language_model.model.layers.7.self_attn.q_proj.weight": "model-00001-of-00002.safetensors",
433
+ "language_model.model.layers.7.self_attn.v_proj.bias": "model-00001-of-00002.safetensors",
434
+ "language_model.model.layers.7.self_attn.v_proj.weight": "model-00001-of-00002.safetensors",
435
+ "language_model.model.layers.8.input_layernorm.bias": "model-00001-of-00002.safetensors",
436
+ "language_model.model.layers.8.input_layernorm.weight": "model-00001-of-00002.safetensors",
437
+ "language_model.model.layers.8.mlp.fc1.bias": "model-00001-of-00002.safetensors",
438
+ "language_model.model.layers.8.mlp.fc1.weight": "model-00001-of-00002.safetensors",
439
+ "language_model.model.layers.8.mlp.fc2.bias": "model-00001-of-00002.safetensors",
440
+ "language_model.model.layers.8.mlp.fc2.weight": "model-00001-of-00002.safetensors",
441
+ "language_model.model.layers.8.self_attn.dense.bias": "model-00001-of-00002.safetensors",
442
+ "language_model.model.layers.8.self_attn.dense.weight": "model-00001-of-00002.safetensors",
443
+ "language_model.model.layers.8.self_attn.k_proj.bias": "model-00001-of-00002.safetensors",
444
+ "language_model.model.layers.8.self_attn.k_proj.weight": "model-00001-of-00002.safetensors",
445
+ "language_model.model.layers.8.self_attn.q_proj.bias": "model-00001-of-00002.safetensors",
446
+ "language_model.model.layers.8.self_attn.q_proj.weight": "model-00001-of-00002.safetensors",
447
+ "language_model.model.layers.8.self_attn.v_proj.bias": "model-00001-of-00002.safetensors",
448
+ "language_model.model.layers.8.self_attn.v_proj.weight": "model-00001-of-00002.safetensors",
449
+ "language_model.model.layers.9.input_layernorm.bias": "model-00001-of-00002.safetensors",
450
+ "language_model.model.layers.9.input_layernorm.weight": "model-00001-of-00002.safetensors",
451
+ "language_model.model.layers.9.mlp.fc1.bias": "model-00001-of-00002.safetensors",
452
+ "language_model.model.layers.9.mlp.fc1.weight": "model-00001-of-00002.safetensors",
453
+ "language_model.model.layers.9.mlp.fc2.bias": "model-00001-of-00002.safetensors",
454
+ "language_model.model.layers.9.mlp.fc2.weight": "model-00001-of-00002.safetensors",
455
+ "language_model.model.layers.9.self_attn.dense.bias": "model-00001-of-00002.safetensors",
456
+ "language_model.model.layers.9.self_attn.dense.weight": "model-00001-of-00002.safetensors",
457
+ "language_model.model.layers.9.self_attn.k_proj.bias": "model-00001-of-00002.safetensors",
458
+ "language_model.model.layers.9.self_attn.k_proj.weight": "model-00001-of-00002.safetensors",
459
+ "language_model.model.layers.9.self_attn.q_proj.bias": "model-00001-of-00002.safetensors",
460
+ "language_model.model.layers.9.self_attn.q_proj.weight": "model-00001-of-00002.safetensors",
461
+ "language_model.model.layers.9.self_attn.v_proj.bias": "model-00001-of-00002.safetensors",
462
+ "language_model.model.layers.9.self_attn.v_proj.weight": "model-00001-of-00002.safetensors",
463
+ "vision_tower._vision_tower.vision_model.embeddings.patch_embedding.bias": "model-00002-of-00002.safetensors",
464
+ "vision_tower._vision_tower.vision_model.embeddings.patch_embedding.weight": "model-00002-of-00002.safetensors",
465
+ "vision_tower._vision_tower.vision_model.embeddings.position_embedding.weight": "model-00002-of-00002.safetensors",
466
+ "vision_tower._vision_tower.vision_model.encoder.layers.0.layer_norm1.bias": "model-00002-of-00002.safetensors",
467
+ "vision_tower._vision_tower.vision_model.encoder.layers.0.layer_norm1.weight": "model-00002-of-00002.safetensors",
468
+ "vision_tower._vision_tower.vision_model.encoder.layers.0.layer_norm2.bias": "model-00002-of-00002.safetensors",
469
+ "vision_tower._vision_tower.vision_model.encoder.layers.0.layer_norm2.weight": "model-00002-of-00002.safetensors",
470
+ "vision_tower._vision_tower.vision_model.encoder.layers.0.mlp.fc1.bias": "model-00002-of-00002.safetensors",
471
+ "vision_tower._vision_tower.vision_model.encoder.layers.0.mlp.fc1.weight": "model-00002-of-00002.safetensors",
472
+ "vision_tower._vision_tower.vision_model.encoder.layers.0.mlp.fc2.bias": "model-00002-of-00002.safetensors",
473
+ "vision_tower._vision_tower.vision_model.encoder.layers.0.mlp.fc2.weight": "model-00002-of-00002.safetensors",
474
+ "vision_tower._vision_tower.vision_model.encoder.layers.0.self_attn.k_proj.bias": "model-00002-of-00002.safetensors",
475
+ "vision_tower._vision_tower.vision_model.encoder.layers.0.self_attn.k_proj.weight": "model-00002-of-00002.safetensors",
476
+ "vision_tower._vision_tower.vision_model.encoder.layers.0.self_attn.out_proj.bias": "model-00002-of-00002.safetensors",
477
+ "vision_tower._vision_tower.vision_model.encoder.layers.0.self_attn.out_proj.weight": "model-00002-of-00002.safetensors",
478
+ "vision_tower._vision_tower.vision_model.encoder.layers.0.self_attn.q_proj.bias": "model-00002-of-00002.safetensors",
479
+ "vision_tower._vision_tower.vision_model.encoder.layers.0.self_attn.q_proj.weight": "model-00002-of-00002.safetensors",
480
+ "vision_tower._vision_tower.vision_model.encoder.layers.0.self_attn.v_proj.bias": "model-00002-of-00002.safetensors",
481
+ "vision_tower._vision_tower.vision_model.encoder.layers.0.self_attn.v_proj.weight": "model-00002-of-00002.safetensors",
482
+ "vision_tower._vision_tower.vision_model.encoder.layers.1.layer_norm1.bias": "model-00002-of-00002.safetensors",
483
+ "vision_tower._vision_tower.vision_model.encoder.layers.1.layer_norm1.weight": "model-00002-of-00002.safetensors",
484
+ "vision_tower._vision_tower.vision_model.encoder.layers.1.layer_norm2.bias": "model-00002-of-00002.safetensors",
485
+ "vision_tower._vision_tower.vision_model.encoder.layers.1.layer_norm2.weight": "model-00002-of-00002.safetensors",
486
+ "vision_tower._vision_tower.vision_model.encoder.layers.1.mlp.fc1.bias": "model-00002-of-00002.safetensors",
487
+ "vision_tower._vision_tower.vision_model.encoder.layers.1.mlp.fc1.weight": "model-00002-of-00002.safetensors",
488
+ "vision_tower._vision_tower.vision_model.encoder.layers.1.mlp.fc2.bias": "model-00002-of-00002.safetensors",
489
+ "vision_tower._vision_tower.vision_model.encoder.layers.1.mlp.fc2.weight": "model-00002-of-00002.safetensors",
490
+ "vision_tower._vision_tower.vision_model.encoder.layers.1.self_attn.k_proj.bias": "model-00002-of-00002.safetensors",
491
+ "vision_tower._vision_tower.vision_model.encoder.layers.1.self_attn.k_proj.weight": "model-00002-of-00002.safetensors",
492
+ "vision_tower._vision_tower.vision_model.encoder.layers.1.self_attn.out_proj.bias": "model-00002-of-00002.safetensors",
493
+ "vision_tower._vision_tower.vision_model.encoder.layers.1.self_attn.out_proj.weight": "model-00002-of-00002.safetensors",
494
+ "vision_tower._vision_tower.vision_model.encoder.layers.1.self_attn.q_proj.bias": "model-00002-of-00002.safetensors",
495
+ "vision_tower._vision_tower.vision_model.encoder.layers.1.self_attn.q_proj.weight": "model-00002-of-00002.safetensors",
496
+ "vision_tower._vision_tower.vision_model.encoder.layers.1.self_attn.v_proj.bias": "model-00002-of-00002.safetensors",
497
+ "vision_tower._vision_tower.vision_model.encoder.layers.1.self_attn.v_proj.weight": "model-00002-of-00002.safetensors",
498
+ "vision_tower._vision_tower.vision_model.encoder.layers.10.layer_norm1.bias": "model-00002-of-00002.safetensors",
499
+ "vision_tower._vision_tower.vision_model.encoder.layers.10.layer_norm1.weight": "model-00002-of-00002.safetensors",
500
+ "vision_tower._vision_tower.vision_model.encoder.layers.10.layer_norm2.bias": "model-00002-of-00002.safetensors",
501
+ "vision_tower._vision_tower.vision_model.encoder.layers.10.layer_norm2.weight": "model-00002-of-00002.safetensors",
502
+ "vision_tower._vision_tower.vision_model.encoder.layers.10.mlp.fc1.bias": "model-00002-of-00002.safetensors",
503
+ "vision_tower._vision_tower.vision_model.encoder.layers.10.mlp.fc1.weight": "model-00002-of-00002.safetensors",
504
+ "vision_tower._vision_tower.vision_model.encoder.layers.10.mlp.fc2.bias": "model-00002-of-00002.safetensors",
505
+ "vision_tower._vision_tower.vision_model.encoder.layers.10.mlp.fc2.weight": "model-00002-of-00002.safetensors",
506
+ "vision_tower._vision_tower.vision_model.encoder.layers.10.self_attn.k_proj.bias": "model-00002-of-00002.safetensors",
507
+ "vision_tower._vision_tower.vision_model.encoder.layers.10.self_attn.k_proj.weight": "model-00002-of-00002.safetensors",
508
+ "vision_tower._vision_tower.vision_model.encoder.layers.10.self_attn.out_proj.bias": "model-00002-of-00002.safetensors",
509
+ "vision_tower._vision_tower.vision_model.encoder.layers.10.self_attn.out_proj.weight": "model-00002-of-00002.safetensors",
510
+ "vision_tower._vision_tower.vision_model.encoder.layers.10.self_attn.q_proj.bias": "model-00002-of-00002.safetensors",
511
+ "vision_tower._vision_tower.vision_model.encoder.layers.10.self_attn.q_proj.weight": "model-00002-of-00002.safetensors",
512
+ "vision_tower._vision_tower.vision_model.encoder.layers.10.self_attn.v_proj.bias": "model-00002-of-00002.safetensors",
513
+ "vision_tower._vision_tower.vision_model.encoder.layers.10.self_attn.v_proj.weight": "model-00002-of-00002.safetensors",
514
+ "vision_tower._vision_tower.vision_model.encoder.layers.11.layer_norm1.bias": "model-00002-of-00002.safetensors",
515
+ "vision_tower._vision_tower.vision_model.encoder.layers.11.layer_norm1.weight": "model-00002-of-00002.safetensors",
516
+ "vision_tower._vision_tower.vision_model.encoder.layers.11.layer_norm2.bias": "model-00002-of-00002.safetensors",
517
+ "vision_tower._vision_tower.vision_model.encoder.layers.11.layer_norm2.weight": "model-00002-of-00002.safetensors",
518
+ "vision_tower._vision_tower.vision_model.encoder.layers.11.mlp.fc1.bias": "model-00002-of-00002.safetensors",
519
+ "vision_tower._vision_tower.vision_model.encoder.layers.11.mlp.fc1.weight": "model-00002-of-00002.safetensors",
520
+ "vision_tower._vision_tower.vision_model.encoder.layers.11.mlp.fc2.bias": "model-00002-of-00002.safetensors",
521
+ "vision_tower._vision_tower.vision_model.encoder.layers.11.mlp.fc2.weight": "model-00002-of-00002.safetensors",
522
+ "vision_tower._vision_tower.vision_model.encoder.layers.11.self_attn.k_proj.bias": "model-00002-of-00002.safetensors",
523
+ "vision_tower._vision_tower.vision_model.encoder.layers.11.self_attn.k_proj.weight": "model-00002-of-00002.safetensors",
524
+ "vision_tower._vision_tower.vision_model.encoder.layers.11.self_attn.out_proj.bias": "model-00002-of-00002.safetensors",
525
+ "vision_tower._vision_tower.vision_model.encoder.layers.11.self_attn.out_proj.weight": "model-00002-of-00002.safetensors",
526
+ "vision_tower._vision_tower.vision_model.encoder.layers.11.self_attn.q_proj.bias": "model-00002-of-00002.safetensors",
527
+ "vision_tower._vision_tower.vision_model.encoder.layers.11.self_attn.q_proj.weight": "model-00002-of-00002.safetensors",
528
+ "vision_tower._vision_tower.vision_model.encoder.layers.11.self_attn.v_proj.bias": "model-00002-of-00002.safetensors",
529
+ "vision_tower._vision_tower.vision_model.encoder.layers.11.self_attn.v_proj.weight": "model-00002-of-00002.safetensors",
530
+ "vision_tower._vision_tower.vision_model.encoder.layers.12.layer_norm1.bias": "model-00002-of-00002.safetensors",
531
+ "vision_tower._vision_tower.vision_model.encoder.layers.12.layer_norm1.weight": "model-00002-of-00002.safetensors",
532
+ "vision_tower._vision_tower.vision_model.encoder.layers.12.layer_norm2.bias": "model-00002-of-00002.safetensors",
533
+ "vision_tower._vision_tower.vision_model.encoder.layers.12.layer_norm2.weight": "model-00002-of-00002.safetensors",
534
+ "vision_tower._vision_tower.vision_model.encoder.layers.12.mlp.fc1.bias": "model-00002-of-00002.safetensors",
535
+ "vision_tower._vision_tower.vision_model.encoder.layers.12.mlp.fc1.weight": "model-00002-of-00002.safetensors",
536
+ "vision_tower._vision_tower.vision_model.encoder.layers.12.mlp.fc2.bias": "model-00002-of-00002.safetensors",
537
+ "vision_tower._vision_tower.vision_model.encoder.layers.12.mlp.fc2.weight": "model-00002-of-00002.safetensors",
538
+ "vision_tower._vision_tower.vision_model.encoder.layers.12.self_attn.k_proj.bias": "model-00002-of-00002.safetensors",
539
+ "vision_tower._vision_tower.vision_model.encoder.layers.12.self_attn.k_proj.weight": "model-00002-of-00002.safetensors",
540
+ "vision_tower._vision_tower.vision_model.encoder.layers.12.self_attn.out_proj.bias": "model-00002-of-00002.safetensors",
541
+ "vision_tower._vision_tower.vision_model.encoder.layers.12.self_attn.out_proj.weight": "model-00002-of-00002.safetensors",
542
+ "vision_tower._vision_tower.vision_model.encoder.layers.12.self_attn.q_proj.bias": "model-00002-of-00002.safetensors",
543
+ "vision_tower._vision_tower.vision_model.encoder.layers.12.self_attn.q_proj.weight": "model-00002-of-00002.safetensors",
544
+ "vision_tower._vision_tower.vision_model.encoder.layers.12.self_attn.v_proj.bias": "model-00002-of-00002.safetensors",
545
+ "vision_tower._vision_tower.vision_model.encoder.layers.12.self_attn.v_proj.weight": "model-00002-of-00002.safetensors",
546
+ "vision_tower._vision_tower.vision_model.encoder.layers.13.layer_norm1.bias": "model-00002-of-00002.safetensors",
547
+ "vision_tower._vision_tower.vision_model.encoder.layers.13.layer_norm1.weight": "model-00002-of-00002.safetensors",
548
+ "vision_tower._vision_tower.vision_model.encoder.layers.13.layer_norm2.bias": "model-00002-of-00002.safetensors",
549
+ "vision_tower._vision_tower.vision_model.encoder.layers.13.layer_norm2.weight": "model-00002-of-00002.safetensors",
550
+ "vision_tower._vision_tower.vision_model.encoder.layers.13.mlp.fc1.bias": "model-00002-of-00002.safetensors",
551
+ "vision_tower._vision_tower.vision_model.encoder.layers.13.mlp.fc1.weight": "model-00002-of-00002.safetensors",
552
+ "vision_tower._vision_tower.vision_model.encoder.layers.13.mlp.fc2.bias": "model-00002-of-00002.safetensors",
553
+ "vision_tower._vision_tower.vision_model.encoder.layers.13.mlp.fc2.weight": "model-00002-of-00002.safetensors",
554
+ "vision_tower._vision_tower.vision_model.encoder.layers.13.self_attn.k_proj.bias": "model-00002-of-00002.safetensors",
555
+ "vision_tower._vision_tower.vision_model.encoder.layers.13.self_attn.k_proj.weight": "model-00002-of-00002.safetensors",
556
+ "vision_tower._vision_tower.vision_model.encoder.layers.13.self_attn.out_proj.bias": "model-00002-of-00002.safetensors",
557
+ "vision_tower._vision_tower.vision_model.encoder.layers.13.self_attn.out_proj.weight": "model-00002-of-00002.safetensors",
558
+ "vision_tower._vision_tower.vision_model.encoder.layers.13.self_attn.q_proj.bias": "model-00002-of-00002.safetensors",
559
+ "vision_tower._vision_tower.vision_model.encoder.layers.13.self_attn.q_proj.weight": "model-00002-of-00002.safetensors",
560
+ "vision_tower._vision_tower.vision_model.encoder.layers.13.self_attn.v_proj.bias": "model-00002-of-00002.safetensors",
561
+ "vision_tower._vision_tower.vision_model.encoder.layers.13.self_attn.v_proj.weight": "model-00002-of-00002.safetensors",
562
+ "vision_tower._vision_tower.vision_model.encoder.layers.14.layer_norm1.bias": "model-00002-of-00002.safetensors",
563
+ "vision_tower._vision_tower.vision_model.encoder.layers.14.layer_norm1.weight": "model-00002-of-00002.safetensors",
564
+ "vision_tower._vision_tower.vision_model.encoder.layers.14.layer_norm2.bias": "model-00002-of-00002.safetensors",
565
+ "vision_tower._vision_tower.vision_model.encoder.layers.14.layer_norm2.weight": "model-00002-of-00002.safetensors",
566
+ "vision_tower._vision_tower.vision_model.encoder.layers.14.mlp.fc1.bias": "model-00002-of-00002.safetensors",
567
+ "vision_tower._vision_tower.vision_model.encoder.layers.14.mlp.fc1.weight": "model-00002-of-00002.safetensors",
568
+ "vision_tower._vision_tower.vision_model.encoder.layers.14.mlp.fc2.bias": "model-00002-of-00002.safetensors",
569
+ "vision_tower._vision_tower.vision_model.encoder.layers.14.mlp.fc2.weight": "model-00002-of-00002.safetensors",
570
+ "vision_tower._vision_tower.vision_model.encoder.layers.14.self_attn.k_proj.bias": "model-00002-of-00002.safetensors",
571
+ "vision_tower._vision_tower.vision_model.encoder.layers.14.self_attn.k_proj.weight": "model-00002-of-00002.safetensors",
572
+ "vision_tower._vision_tower.vision_model.encoder.layers.14.self_attn.out_proj.bias": "model-00002-of-00002.safetensors",
573
+ "vision_tower._vision_tower.vision_model.encoder.layers.14.self_attn.out_proj.weight": "model-00002-of-00002.safetensors",
574
+ "vision_tower._vision_tower.vision_model.encoder.layers.14.self_attn.q_proj.bias": "model-00002-of-00002.safetensors",
575
+ "vision_tower._vision_tower.vision_model.encoder.layers.14.self_attn.q_proj.weight": "model-00002-of-00002.safetensors",
576
+ "vision_tower._vision_tower.vision_model.encoder.layers.14.self_attn.v_proj.bias": "model-00002-of-00002.safetensors",
577
+ "vision_tower._vision_tower.vision_model.encoder.layers.14.self_attn.v_proj.weight": "model-00002-of-00002.safetensors",
578
+ "vision_tower._vision_tower.vision_model.encoder.layers.15.layer_norm1.bias": "model-00002-of-00002.safetensors",
579
+ "vision_tower._vision_tower.vision_model.encoder.layers.15.layer_norm1.weight": "model-00002-of-00002.safetensors",
580
+ "vision_tower._vision_tower.vision_model.encoder.layers.15.layer_norm2.bias": "model-00002-of-00002.safetensors",
581
+ "vision_tower._vision_tower.vision_model.encoder.layers.15.layer_norm2.weight": "model-00002-of-00002.safetensors",
582
+ "vision_tower._vision_tower.vision_model.encoder.layers.15.mlp.fc1.bias": "model-00002-of-00002.safetensors",
583
+ "vision_tower._vision_tower.vision_model.encoder.layers.15.mlp.fc1.weight": "model-00002-of-00002.safetensors",
584
+ "vision_tower._vision_tower.vision_model.encoder.layers.15.mlp.fc2.bias": "model-00002-of-00002.safetensors",
585
+ "vision_tower._vision_tower.vision_model.encoder.layers.15.mlp.fc2.weight": "model-00002-of-00002.safetensors",
586
+ "vision_tower._vision_tower.vision_model.encoder.layers.15.self_attn.k_proj.bias": "model-00002-of-00002.safetensors",
587
+ "vision_tower._vision_tower.vision_model.encoder.layers.15.self_attn.k_proj.weight": "model-00002-of-00002.safetensors",
588
+ "vision_tower._vision_tower.vision_model.encoder.layers.15.self_attn.out_proj.bias": "model-00002-of-00002.safetensors",
589
+ "vision_tower._vision_tower.vision_model.encoder.layers.15.self_attn.out_proj.weight": "model-00002-of-00002.safetensors",
590
+ "vision_tower._vision_tower.vision_model.encoder.layers.15.self_attn.q_proj.bias": "model-00002-of-00002.safetensors",
591
+ "vision_tower._vision_tower.vision_model.encoder.layers.15.self_attn.q_proj.weight": "model-00002-of-00002.safetensors",
592
+ "vision_tower._vision_tower.vision_model.encoder.layers.15.self_attn.v_proj.bias": "model-00002-of-00002.safetensors",
593
+ "vision_tower._vision_tower.vision_model.encoder.layers.15.self_attn.v_proj.weight": "model-00002-of-00002.safetensors",
594
+ "vision_tower._vision_tower.vision_model.encoder.layers.16.layer_norm1.bias": "model-00002-of-00002.safetensors",
595
+ "vision_tower._vision_tower.vision_model.encoder.layers.16.layer_norm1.weight": "model-00002-of-00002.safetensors",
596
+ "vision_tower._vision_tower.vision_model.encoder.layers.16.layer_norm2.bias": "model-00002-of-00002.safetensors",
597
+ "vision_tower._vision_tower.vision_model.encoder.layers.16.layer_norm2.weight": "model-00002-of-00002.safetensors",
598
+ "vision_tower._vision_tower.vision_model.encoder.layers.16.mlp.fc1.bias": "model-00002-of-00002.safetensors",
599
+ "vision_tower._vision_tower.vision_model.encoder.layers.16.mlp.fc1.weight": "model-00002-of-00002.safetensors",
600
+ "vision_tower._vision_tower.vision_model.encoder.layers.16.mlp.fc2.bias": "model-00002-of-00002.safetensors",
601
+ "vision_tower._vision_tower.vision_model.encoder.layers.16.mlp.fc2.weight": "model-00002-of-00002.safetensors",
602
+ "vision_tower._vision_tower.vision_model.encoder.layers.16.self_attn.k_proj.bias": "model-00002-of-00002.safetensors",
603
+ "vision_tower._vision_tower.vision_model.encoder.layers.16.self_attn.k_proj.weight": "model-00002-of-00002.safetensors",
604
+ "vision_tower._vision_tower.vision_model.encoder.layers.16.self_attn.out_proj.bias": "model-00002-of-00002.safetensors",
605
+ "vision_tower._vision_tower.vision_model.encoder.layers.16.self_attn.out_proj.weight": "model-00002-of-00002.safetensors",
606
+ "vision_tower._vision_tower.vision_model.encoder.layers.16.self_attn.q_proj.bias": "model-00002-of-00002.safetensors",
607
+ "vision_tower._vision_tower.vision_model.encoder.layers.16.self_attn.q_proj.weight": "model-00002-of-00002.safetensors",
608
+ "vision_tower._vision_tower.vision_model.encoder.layers.16.self_attn.v_proj.bias": "model-00002-of-00002.safetensors",
609
+ "vision_tower._vision_tower.vision_model.encoder.layers.16.self_attn.v_proj.weight": "model-00002-of-00002.safetensors",
610
+ "vision_tower._vision_tower.vision_model.encoder.layers.17.layer_norm1.bias": "model-00002-of-00002.safetensors",
611
+ "vision_tower._vision_tower.vision_model.encoder.layers.17.layer_norm1.weight": "model-00002-of-00002.safetensors",
612
+ "vision_tower._vision_tower.vision_model.encoder.layers.17.layer_norm2.bias": "model-00002-of-00002.safetensors",
613
+ "vision_tower._vision_tower.vision_model.encoder.layers.17.layer_norm2.weight": "model-00002-of-00002.safetensors",
614
+ "vision_tower._vision_tower.vision_model.encoder.layers.17.mlp.fc1.bias": "model-00002-of-00002.safetensors",
615
+ "vision_tower._vision_tower.vision_model.encoder.layers.17.mlp.fc1.weight": "model-00002-of-00002.safetensors",
616
+ "vision_tower._vision_tower.vision_model.encoder.layers.17.mlp.fc2.bias": "model-00002-of-00002.safetensors",
617
+ "vision_tower._vision_tower.vision_model.encoder.layers.17.mlp.fc2.weight": "model-00002-of-00002.safetensors",
618
+ "vision_tower._vision_tower.vision_model.encoder.layers.17.self_attn.k_proj.bias": "model-00002-of-00002.safetensors",
619
+ "vision_tower._vision_tower.vision_model.encoder.layers.17.self_attn.k_proj.weight": "model-00002-of-00002.safetensors",
620
+ "vision_tower._vision_tower.vision_model.encoder.layers.17.self_attn.out_proj.bias": "model-00002-of-00002.safetensors",
621
+ "vision_tower._vision_tower.vision_model.encoder.layers.17.self_attn.out_proj.weight": "model-00002-of-00002.safetensors",
622
+ "vision_tower._vision_tower.vision_model.encoder.layers.17.self_attn.q_proj.bias": "model-00002-of-00002.safetensors",
623
+ "vision_tower._vision_tower.vision_model.encoder.layers.17.self_attn.q_proj.weight": "model-00002-of-00002.safetensors",
624
+ "vision_tower._vision_tower.vision_model.encoder.layers.17.self_attn.v_proj.bias": "model-00002-of-00002.safetensors",
625
+ "vision_tower._vision_tower.vision_model.encoder.layers.17.self_attn.v_proj.weight": "model-00002-of-00002.safetensors",
626
+ "vision_tower._vision_tower.vision_model.encoder.layers.18.layer_norm1.bias": "model-00002-of-00002.safetensors",
627
+ "vision_tower._vision_tower.vision_model.encoder.layers.18.layer_norm1.weight": "model-00002-of-00002.safetensors",
628
+ "vision_tower._vision_tower.vision_model.encoder.layers.18.layer_norm2.bias": "model-00002-of-00002.safetensors",
629
+ "vision_tower._vision_tower.vision_model.encoder.layers.18.layer_norm2.weight": "model-00002-of-00002.safetensors",
630
+ "vision_tower._vision_tower.vision_model.encoder.layers.18.mlp.fc1.bias": "model-00002-of-00002.safetensors",
631
+ "vision_tower._vision_tower.vision_model.encoder.layers.18.mlp.fc1.weight": "model-00002-of-00002.safetensors",
632
+ "vision_tower._vision_tower.vision_model.encoder.layers.18.mlp.fc2.bias": "model-00002-of-00002.safetensors",
633
+ "vision_tower._vision_tower.vision_model.encoder.layers.18.mlp.fc2.weight": "model-00002-of-00002.safetensors",
634
+ "vision_tower._vision_tower.vision_model.encoder.layers.18.self_attn.k_proj.bias": "model-00002-of-00002.safetensors",
635
+ "vision_tower._vision_tower.vision_model.encoder.layers.18.self_attn.k_proj.weight": "model-00002-of-00002.safetensors",
636
+ "vision_tower._vision_tower.vision_model.encoder.layers.18.self_attn.out_proj.bias": "model-00002-of-00002.safetensors",
637
+ "vision_tower._vision_tower.vision_model.encoder.layers.18.self_attn.out_proj.weight": "model-00002-of-00002.safetensors",
638
+ "vision_tower._vision_tower.vision_model.encoder.layers.18.self_attn.q_proj.bias": "model-00002-of-00002.safetensors",
639
+ "vision_tower._vision_tower.vision_model.encoder.layers.18.self_attn.q_proj.weight": "model-00002-of-00002.safetensors",
640
+ "vision_tower._vision_tower.vision_model.encoder.layers.18.self_attn.v_proj.bias": "model-00002-of-00002.safetensors",
641
+ "vision_tower._vision_tower.vision_model.encoder.layers.18.self_attn.v_proj.weight": "model-00002-of-00002.safetensors",
642
+ "vision_tower._vision_tower.vision_model.encoder.layers.19.layer_norm1.bias": "model-00002-of-00002.safetensors",
643
+ "vision_tower._vision_tower.vision_model.encoder.layers.19.layer_norm1.weight": "model-00002-of-00002.safetensors",
644
+ "vision_tower._vision_tower.vision_model.encoder.layers.19.layer_norm2.bias": "model-00002-of-00002.safetensors",
645
+ "vision_tower._vision_tower.vision_model.encoder.layers.19.layer_norm2.weight": "model-00002-of-00002.safetensors",
646
+ "vision_tower._vision_tower.vision_model.encoder.layers.19.mlp.fc1.bias": "model-00002-of-00002.safetensors",
647
+ "vision_tower._vision_tower.vision_model.encoder.layers.19.mlp.fc1.weight": "model-00002-of-00002.safetensors",
648
+ "vision_tower._vision_tower.vision_model.encoder.layers.19.mlp.fc2.bias": "model-00002-of-00002.safetensors",
649
+ "vision_tower._vision_tower.vision_model.encoder.layers.19.mlp.fc2.weight": "model-00002-of-00002.safetensors",
650
+ "vision_tower._vision_tower.vision_model.encoder.layers.19.self_attn.k_proj.bias": "model-00002-of-00002.safetensors",
651
+ "vision_tower._vision_tower.vision_model.encoder.layers.19.self_attn.k_proj.weight": "model-00002-of-00002.safetensors",
652
+ "vision_tower._vision_tower.vision_model.encoder.layers.19.self_attn.out_proj.bias": "model-00002-of-00002.safetensors",
653
+ "vision_tower._vision_tower.vision_model.encoder.layers.19.self_attn.out_proj.weight": "model-00002-of-00002.safetensors",
654
+ "vision_tower._vision_tower.vision_model.encoder.layers.19.self_attn.q_proj.bias": "model-00002-of-00002.safetensors",
655
+ "vision_tower._vision_tower.vision_model.encoder.layers.19.self_attn.q_proj.weight": "model-00002-of-00002.safetensors",
656
+ "vision_tower._vision_tower.vision_model.encoder.layers.19.self_attn.v_proj.bias": "model-00002-of-00002.safetensors",
657
+ "vision_tower._vision_tower.vision_model.encoder.layers.19.self_attn.v_proj.weight": "model-00002-of-00002.safetensors",
658
+ "vision_tower._vision_tower.vision_model.encoder.layers.2.layer_norm1.bias": "model-00002-of-00002.safetensors",
659
+ "vision_tower._vision_tower.vision_model.encoder.layers.2.layer_norm1.weight": "model-00002-of-00002.safetensors",
660
+ "vision_tower._vision_tower.vision_model.encoder.layers.2.layer_norm2.bias": "model-00002-of-00002.safetensors",
661
+ "vision_tower._vision_tower.vision_model.encoder.layers.2.layer_norm2.weight": "model-00002-of-00002.safetensors",
662
+ "vision_tower._vision_tower.vision_model.encoder.layers.2.mlp.fc1.bias": "model-00002-of-00002.safetensors",
663
+ "vision_tower._vision_tower.vision_model.encoder.layers.2.mlp.fc1.weight": "model-00002-of-00002.safetensors",
664
+ "vision_tower._vision_tower.vision_model.encoder.layers.2.mlp.fc2.bias": "model-00002-of-00002.safetensors",
665
+ "vision_tower._vision_tower.vision_model.encoder.layers.2.mlp.fc2.weight": "model-00002-of-00002.safetensors",
666
+ "vision_tower._vision_tower.vision_model.encoder.layers.2.self_attn.k_proj.bias": "model-00002-of-00002.safetensors",
667
+ "vision_tower._vision_tower.vision_model.encoder.layers.2.self_attn.k_proj.weight": "model-00002-of-00002.safetensors",
668
+ "vision_tower._vision_tower.vision_model.encoder.layers.2.self_attn.out_proj.bias": "model-00002-of-00002.safetensors",
669
+ "vision_tower._vision_tower.vision_model.encoder.layers.2.self_attn.out_proj.weight": "model-00002-of-00002.safetensors",
670
+ "vision_tower._vision_tower.vision_model.encoder.layers.2.self_attn.q_proj.bias": "model-00002-of-00002.safetensors",
671
+ "vision_tower._vision_tower.vision_model.encoder.layers.2.self_attn.q_proj.weight": "model-00002-of-00002.safetensors",
672
+ "vision_tower._vision_tower.vision_model.encoder.layers.2.self_attn.v_proj.bias": "model-00002-of-00002.safetensors",
673
+ "vision_tower._vision_tower.vision_model.encoder.layers.2.self_attn.v_proj.weight": "model-00002-of-00002.safetensors",
674
+ "vision_tower._vision_tower.vision_model.encoder.layers.20.layer_norm1.bias": "model-00002-of-00002.safetensors",
675
+ "vision_tower._vision_tower.vision_model.encoder.layers.20.layer_norm1.weight": "model-00002-of-00002.safetensors",
676
+ "vision_tower._vision_tower.vision_model.encoder.layers.20.layer_norm2.bias": "model-00002-of-00002.safetensors",
677
+ "vision_tower._vision_tower.vision_model.encoder.layers.20.layer_norm2.weight": "model-00002-of-00002.safetensors",
678
+ "vision_tower._vision_tower.vision_model.encoder.layers.20.mlp.fc1.bias": "model-00002-of-00002.safetensors",
679
+ "vision_tower._vision_tower.vision_model.encoder.layers.20.mlp.fc1.weight": "model-00002-of-00002.safetensors",
680
+ "vision_tower._vision_tower.vision_model.encoder.layers.20.mlp.fc2.bias": "model-00002-of-00002.safetensors",
681
+ "vision_tower._vision_tower.vision_model.encoder.layers.20.mlp.fc2.weight": "model-00002-of-00002.safetensors",
682
+ "vision_tower._vision_tower.vision_model.encoder.layers.20.self_attn.k_proj.bias": "model-00002-of-00002.safetensors",
683
+ "vision_tower._vision_tower.vision_model.encoder.layers.20.self_attn.k_proj.weight": "model-00002-of-00002.safetensors",
684
+ "vision_tower._vision_tower.vision_model.encoder.layers.20.self_attn.out_proj.bias": "model-00002-of-00002.safetensors",
685
+ "vision_tower._vision_tower.vision_model.encoder.layers.20.self_attn.out_proj.weight": "model-00002-of-00002.safetensors",
686
+ "vision_tower._vision_tower.vision_model.encoder.layers.20.self_attn.q_proj.bias": "model-00002-of-00002.safetensors",
687
+ "vision_tower._vision_tower.vision_model.encoder.layers.20.self_attn.q_proj.weight": "model-00002-of-00002.safetensors",
688
+ "vision_tower._vision_tower.vision_model.encoder.layers.20.self_attn.v_proj.bias": "model-00002-of-00002.safetensors",
689
+ "vision_tower._vision_tower.vision_model.encoder.layers.20.self_attn.v_proj.weight": "model-00002-of-00002.safetensors",
690
+ "vision_tower._vision_tower.vision_model.encoder.layers.21.layer_norm1.bias": "model-00002-of-00002.safetensors",
691
+ "vision_tower._vision_tower.vision_model.encoder.layers.21.layer_norm1.weight": "model-00002-of-00002.safetensors",
692
+ "vision_tower._vision_tower.vision_model.encoder.layers.21.layer_norm2.bias": "model-00002-of-00002.safetensors",
693
+ "vision_tower._vision_tower.vision_model.encoder.layers.21.layer_norm2.weight": "model-00002-of-00002.safetensors",
694
+ "vision_tower._vision_tower.vision_model.encoder.layers.21.mlp.fc1.bias": "model-00002-of-00002.safetensors",
695
+ "vision_tower._vision_tower.vision_model.encoder.layers.21.mlp.fc1.weight": "model-00002-of-00002.safetensors",
696
+ "vision_tower._vision_tower.vision_model.encoder.layers.21.mlp.fc2.bias": "model-00002-of-00002.safetensors",
697
+ "vision_tower._vision_tower.vision_model.encoder.layers.21.mlp.fc2.weight": "model-00002-of-00002.safetensors",
698
+ "vision_tower._vision_tower.vision_model.encoder.layers.21.self_attn.k_proj.bias": "model-00002-of-00002.safetensors",
699
+ "vision_tower._vision_tower.vision_model.encoder.layers.21.self_attn.k_proj.weight": "model-00002-of-00002.safetensors",
700
+ "vision_tower._vision_tower.vision_model.encoder.layers.21.self_attn.out_proj.bias": "model-00002-of-00002.safetensors",
701
+ "vision_tower._vision_tower.vision_model.encoder.layers.21.self_attn.out_proj.weight": "model-00002-of-00002.safetensors",
702
+ "vision_tower._vision_tower.vision_model.encoder.layers.21.self_attn.q_proj.bias": "model-00002-of-00002.safetensors",
703
+ "vision_tower._vision_tower.vision_model.encoder.layers.21.self_attn.q_proj.weight": "model-00002-of-00002.safetensors",
704
+ "vision_tower._vision_tower.vision_model.encoder.layers.21.self_attn.v_proj.bias": "model-00002-of-00002.safetensors",
705
+ "vision_tower._vision_tower.vision_model.encoder.layers.21.self_attn.v_proj.weight": "model-00002-of-00002.safetensors",
706
+ "vision_tower._vision_tower.vision_model.encoder.layers.22.layer_norm1.bias": "model-00002-of-00002.safetensors",
707
+ "vision_tower._vision_tower.vision_model.encoder.layers.22.layer_norm1.weight": "model-00002-of-00002.safetensors",
708
+ "vision_tower._vision_tower.vision_model.encoder.layers.22.layer_norm2.bias": "model-00002-of-00002.safetensors",
709
+ "vision_tower._vision_tower.vision_model.encoder.layers.22.layer_norm2.weight": "model-00002-of-00002.safetensors",
710
+ "vision_tower._vision_tower.vision_model.encoder.layers.22.mlp.fc1.bias": "model-00002-of-00002.safetensors",
711
+ "vision_tower._vision_tower.vision_model.encoder.layers.22.mlp.fc1.weight": "model-00002-of-00002.safetensors",
712
+ "vision_tower._vision_tower.vision_model.encoder.layers.22.mlp.fc2.bias": "model-00002-of-00002.safetensors",
713
+ "vision_tower._vision_tower.vision_model.encoder.layers.22.mlp.fc2.weight": "model-00002-of-00002.safetensors",
714
+ "vision_tower._vision_tower.vision_model.encoder.layers.22.self_attn.k_proj.bias": "model-00002-of-00002.safetensors",
715
+ "vision_tower._vision_tower.vision_model.encoder.layers.22.self_attn.k_proj.weight": "model-00002-of-00002.safetensors",
716
+ "vision_tower._vision_tower.vision_model.encoder.layers.22.self_attn.out_proj.bias": "model-00002-of-00002.safetensors",
717
+ "vision_tower._vision_tower.vision_model.encoder.layers.22.self_attn.out_proj.weight": "model-00002-of-00002.safetensors",
718
+ "vision_tower._vision_tower.vision_model.encoder.layers.22.self_attn.q_proj.bias": "model-00002-of-00002.safetensors",
719
+ "vision_tower._vision_tower.vision_model.encoder.layers.22.self_attn.q_proj.weight": "model-00002-of-00002.safetensors",
720
+ "vision_tower._vision_tower.vision_model.encoder.layers.22.self_attn.v_proj.bias": "model-00002-of-00002.safetensors",
721
+ "vision_tower._vision_tower.vision_model.encoder.layers.22.self_attn.v_proj.weight": "model-00002-of-00002.safetensors",
722
+ "vision_tower._vision_tower.vision_model.encoder.layers.23.layer_norm1.bias": "model-00002-of-00002.safetensors",
723
+ "vision_tower._vision_tower.vision_model.encoder.layers.23.layer_norm1.weight": "model-00002-of-00002.safetensors",
724
+ "vision_tower._vision_tower.vision_model.encoder.layers.23.layer_norm2.bias": "model-00002-of-00002.safetensors",
725
+ "vision_tower._vision_tower.vision_model.encoder.layers.23.layer_norm2.weight": "model-00002-of-00002.safetensors",
726
+ "vision_tower._vision_tower.vision_model.encoder.layers.23.mlp.fc1.bias": "model-00002-of-00002.safetensors",
727
+ "vision_tower._vision_tower.vision_model.encoder.layers.23.mlp.fc1.weight": "model-00002-of-00002.safetensors",
728
+ "vision_tower._vision_tower.vision_model.encoder.layers.23.mlp.fc2.bias": "model-00002-of-00002.safetensors",
729
+ "vision_tower._vision_tower.vision_model.encoder.layers.23.mlp.fc2.weight": "model-00002-of-00002.safetensors",
730
+ "vision_tower._vision_tower.vision_model.encoder.layers.23.self_attn.k_proj.bias": "model-00002-of-00002.safetensors",
731
+ "vision_tower._vision_tower.vision_model.encoder.layers.23.self_attn.k_proj.weight": "model-00002-of-00002.safetensors",
732
+ "vision_tower._vision_tower.vision_model.encoder.layers.23.self_attn.out_proj.bias": "model-00002-of-00002.safetensors",
733
+ "vision_tower._vision_tower.vision_model.encoder.layers.23.self_attn.out_proj.weight": "model-00002-of-00002.safetensors",
734
+ "vision_tower._vision_tower.vision_model.encoder.layers.23.self_attn.q_proj.bias": "model-00002-of-00002.safetensors",
735
+ "vision_tower._vision_tower.vision_model.encoder.layers.23.self_attn.q_proj.weight": "model-00002-of-00002.safetensors",
736
+ "vision_tower._vision_tower.vision_model.encoder.layers.23.self_attn.v_proj.bias": "model-00002-of-00002.safetensors",
737
+ "vision_tower._vision_tower.vision_model.encoder.layers.23.self_attn.v_proj.weight": "model-00002-of-00002.safetensors",
738
+ "vision_tower._vision_tower.vision_model.encoder.layers.24.layer_norm1.bias": "model-00002-of-00002.safetensors",
739
+ "vision_tower._vision_tower.vision_model.encoder.layers.24.layer_norm1.weight": "model-00002-of-00002.safetensors",
740
+ "vision_tower._vision_tower.vision_model.encoder.layers.24.layer_norm2.bias": "model-00002-of-00002.safetensors",
741
+ "vision_tower._vision_tower.vision_model.encoder.layers.24.layer_norm2.weight": "model-00002-of-00002.safetensors",
742
+ "vision_tower._vision_tower.vision_model.encoder.layers.24.mlp.fc1.bias": "model-00002-of-00002.safetensors",
743
+ "vision_tower._vision_tower.vision_model.encoder.layers.24.mlp.fc1.weight": "model-00002-of-00002.safetensors",
744
+ "vision_tower._vision_tower.vision_model.encoder.layers.24.mlp.fc2.bias": "model-00002-of-00002.safetensors",
745
+ "vision_tower._vision_tower.vision_model.encoder.layers.24.mlp.fc2.weight": "model-00002-of-00002.safetensors",
746
+ "vision_tower._vision_tower.vision_model.encoder.layers.24.self_attn.k_proj.bias": "model-00002-of-00002.safetensors",
747
+ "vision_tower._vision_tower.vision_model.encoder.layers.24.self_attn.k_proj.weight": "model-00002-of-00002.safetensors",
748
+ "vision_tower._vision_tower.vision_model.encoder.layers.24.self_attn.out_proj.bias": "model-00002-of-00002.safetensors",
749
+ "vision_tower._vision_tower.vision_model.encoder.layers.24.self_attn.out_proj.weight": "model-00002-of-00002.safetensors",
750
+ "vision_tower._vision_tower.vision_model.encoder.layers.24.self_attn.q_proj.bias": "model-00002-of-00002.safetensors",
751
+ "vision_tower._vision_tower.vision_model.encoder.layers.24.self_attn.q_proj.weight": "model-00002-of-00002.safetensors",
752
+ "vision_tower._vision_tower.vision_model.encoder.layers.24.self_attn.v_proj.bias": "model-00002-of-00002.safetensors",
753
+ "vision_tower._vision_tower.vision_model.encoder.layers.24.self_attn.v_proj.weight": "model-00002-of-00002.safetensors",
754
+ "vision_tower._vision_tower.vision_model.encoder.layers.25.layer_norm1.bias": "model-00002-of-00002.safetensors",
755
+ "vision_tower._vision_tower.vision_model.encoder.layers.25.layer_norm1.weight": "model-00002-of-00002.safetensors",
756
+ "vision_tower._vision_tower.vision_model.encoder.layers.25.layer_norm2.bias": "model-00002-of-00002.safetensors",
757
+ "vision_tower._vision_tower.vision_model.encoder.layers.25.layer_norm2.weight": "model-00002-of-00002.safetensors",
758
+ "vision_tower._vision_tower.vision_model.encoder.layers.25.mlp.fc1.bias": "model-00002-of-00002.safetensors",
759
+ "vision_tower._vision_tower.vision_model.encoder.layers.25.mlp.fc1.weight": "model-00002-of-00002.safetensors",
760
+ "vision_tower._vision_tower.vision_model.encoder.layers.25.mlp.fc2.bias": "model-00002-of-00002.safetensors",
761
+ "vision_tower._vision_tower.vision_model.encoder.layers.25.mlp.fc2.weight": "model-00002-of-00002.safetensors",
762
+ "vision_tower._vision_tower.vision_model.encoder.layers.25.self_attn.k_proj.bias": "model-00002-of-00002.safetensors",
763
+ "vision_tower._vision_tower.vision_model.encoder.layers.25.self_attn.k_proj.weight": "model-00002-of-00002.safetensors",
764
+ "vision_tower._vision_tower.vision_model.encoder.layers.25.self_attn.out_proj.bias": "model-00002-of-00002.safetensors",
765
+ "vision_tower._vision_tower.vision_model.encoder.layers.25.self_attn.out_proj.weight": "model-00002-of-00002.safetensors",
766
+ "vision_tower._vision_tower.vision_model.encoder.layers.25.self_attn.q_proj.bias": "model-00002-of-00002.safetensors",
767
+ "vision_tower._vision_tower.vision_model.encoder.layers.25.self_attn.q_proj.weight": "model-00002-of-00002.safetensors",
768
+ "vision_tower._vision_tower.vision_model.encoder.layers.25.self_attn.v_proj.bias": "model-00002-of-00002.safetensors",
769
+ "vision_tower._vision_tower.vision_model.encoder.layers.25.self_attn.v_proj.weight": "model-00002-of-00002.safetensors",
770
+ "vision_tower._vision_tower.vision_model.encoder.layers.26.layer_norm1.bias": "model-00002-of-00002.safetensors",
771
+ "vision_tower._vision_tower.vision_model.encoder.layers.26.layer_norm1.weight": "model-00002-of-00002.safetensors",
772
+ "vision_tower._vision_tower.vision_model.encoder.layers.26.layer_norm2.bias": "model-00002-of-00002.safetensors",
773
+ "vision_tower._vision_tower.vision_model.encoder.layers.26.layer_norm2.weight": "model-00002-of-00002.safetensors",
774
+ "vision_tower._vision_tower.vision_model.encoder.layers.26.mlp.fc1.bias": "model-00002-of-00002.safetensors",
775
+ "vision_tower._vision_tower.vision_model.encoder.layers.26.mlp.fc1.weight": "model-00002-of-00002.safetensors",
776
+ "vision_tower._vision_tower.vision_model.encoder.layers.26.mlp.fc2.bias": "model-00002-of-00002.safetensors",
777
+ "vision_tower._vision_tower.vision_model.encoder.layers.26.mlp.fc2.weight": "model-00002-of-00002.safetensors",
778
+ "vision_tower._vision_tower.vision_model.encoder.layers.26.self_attn.k_proj.bias": "model-00002-of-00002.safetensors",
779
+ "vision_tower._vision_tower.vision_model.encoder.layers.26.self_attn.k_proj.weight": "model-00002-of-00002.safetensors",
780
+ "vision_tower._vision_tower.vision_model.encoder.layers.26.self_attn.out_proj.bias": "model-00002-of-00002.safetensors",
781
+ "vision_tower._vision_tower.vision_model.encoder.layers.26.self_attn.out_proj.weight": "model-00002-of-00002.safetensors",
782
+ "vision_tower._vision_tower.vision_model.encoder.layers.26.self_attn.q_proj.bias": "model-00002-of-00002.safetensors",
783
+ "vision_tower._vision_tower.vision_model.encoder.layers.26.self_attn.q_proj.weight": "model-00002-of-00002.safetensors",
784
+ "vision_tower._vision_tower.vision_model.encoder.layers.26.self_attn.v_proj.bias": "model-00002-of-00002.safetensors",
785
+ "vision_tower._vision_tower.vision_model.encoder.layers.26.self_attn.v_proj.weight": "model-00002-of-00002.safetensors",
786
+ "vision_tower._vision_tower.vision_model.encoder.layers.3.layer_norm1.bias": "model-00002-of-00002.safetensors",
787
+ "vision_tower._vision_tower.vision_model.encoder.layers.3.layer_norm1.weight": "model-00002-of-00002.safetensors",
788
+ "vision_tower._vision_tower.vision_model.encoder.layers.3.layer_norm2.bias": "model-00002-of-00002.safetensors",
789
+ "vision_tower._vision_tower.vision_model.encoder.layers.3.layer_norm2.weight": "model-00002-of-00002.safetensors",
790
+ "vision_tower._vision_tower.vision_model.encoder.layers.3.mlp.fc1.bias": "model-00002-of-00002.safetensors",
791
+ "vision_tower._vision_tower.vision_model.encoder.layers.3.mlp.fc1.weight": "model-00002-of-00002.safetensors",
792
+ "vision_tower._vision_tower.vision_model.encoder.layers.3.mlp.fc2.bias": "model-00002-of-00002.safetensors",
793
+ "vision_tower._vision_tower.vision_model.encoder.layers.3.mlp.fc2.weight": "model-00002-of-00002.safetensors",
794
+ "vision_tower._vision_tower.vision_model.encoder.layers.3.self_attn.k_proj.bias": "model-00002-of-00002.safetensors",
795
+ "vision_tower._vision_tower.vision_model.encoder.layers.3.self_attn.k_proj.weight": "model-00002-of-00002.safetensors",
796
+ "vision_tower._vision_tower.vision_model.encoder.layers.3.self_attn.out_proj.bias": "model-00002-of-00002.safetensors",
797
+ "vision_tower._vision_tower.vision_model.encoder.layers.3.self_attn.out_proj.weight": "model-00002-of-00002.safetensors",
798
+ "vision_tower._vision_tower.vision_model.encoder.layers.3.self_attn.q_proj.bias": "model-00002-of-00002.safetensors",
799
+ "vision_tower._vision_tower.vision_model.encoder.layers.3.self_attn.q_proj.weight": "model-00002-of-00002.safetensors",
800
+ "vision_tower._vision_tower.vision_model.encoder.layers.3.self_attn.v_proj.bias": "model-00002-of-00002.safetensors",
801
+ "vision_tower._vision_tower.vision_model.encoder.layers.3.self_attn.v_proj.weight": "model-00002-of-00002.safetensors",
802
+ "vision_tower._vision_tower.vision_model.encoder.layers.4.layer_norm1.bias": "model-00002-of-00002.safetensors",
803
+ "vision_tower._vision_tower.vision_model.encoder.layers.4.layer_norm1.weight": "model-00002-of-00002.safetensors",
804
+ "vision_tower._vision_tower.vision_model.encoder.layers.4.layer_norm2.bias": "model-00002-of-00002.safetensors",
805
+ "vision_tower._vision_tower.vision_model.encoder.layers.4.layer_norm2.weight": "model-00002-of-00002.safetensors",
806
+ "vision_tower._vision_tower.vision_model.encoder.layers.4.mlp.fc1.bias": "model-00002-of-00002.safetensors",
807
+ "vision_tower._vision_tower.vision_model.encoder.layers.4.mlp.fc1.weight": "model-00002-of-00002.safetensors",
808
+ "vision_tower._vision_tower.vision_model.encoder.layers.4.mlp.fc2.bias": "model-00002-of-00002.safetensors",
809
+ "vision_tower._vision_tower.vision_model.encoder.layers.4.mlp.fc2.weight": "model-00002-of-00002.safetensors",
810
+ "vision_tower._vision_tower.vision_model.encoder.layers.4.self_attn.k_proj.bias": "model-00002-of-00002.safetensors",
811
+ "vision_tower._vision_tower.vision_model.encoder.layers.4.self_attn.k_proj.weight": "model-00002-of-00002.safetensors",
812
+ "vision_tower._vision_tower.vision_model.encoder.layers.4.self_attn.out_proj.bias": "model-00002-of-00002.safetensors",
813
+ "vision_tower._vision_tower.vision_model.encoder.layers.4.self_attn.out_proj.weight": "model-00002-of-00002.safetensors",
814
+ "vision_tower._vision_tower.vision_model.encoder.layers.4.self_attn.q_proj.bias": "model-00002-of-00002.safetensors",
815
+ "vision_tower._vision_tower.vision_model.encoder.layers.4.self_attn.q_proj.weight": "model-00002-of-00002.safetensors",
816
+ "vision_tower._vision_tower.vision_model.encoder.layers.4.self_attn.v_proj.bias": "model-00002-of-00002.safetensors",
817
+ "vision_tower._vision_tower.vision_model.encoder.layers.4.self_attn.v_proj.weight": "model-00002-of-00002.safetensors",
818
+ "vision_tower._vision_tower.vision_model.encoder.layers.5.layer_norm1.bias": "model-00002-of-00002.safetensors",
819
+ "vision_tower._vision_tower.vision_model.encoder.layers.5.layer_norm1.weight": "model-00002-of-00002.safetensors",
820
+ "vision_tower._vision_tower.vision_model.encoder.layers.5.layer_norm2.bias": "model-00002-of-00002.safetensors",
821
+ "vision_tower._vision_tower.vision_model.encoder.layers.5.layer_norm2.weight": "model-00002-of-00002.safetensors",
822
+ "vision_tower._vision_tower.vision_model.encoder.layers.5.mlp.fc1.bias": "model-00002-of-00002.safetensors",
823
+ "vision_tower._vision_tower.vision_model.encoder.layers.5.mlp.fc1.weight": "model-00002-of-00002.safetensors",
824
+ "vision_tower._vision_tower.vision_model.encoder.layers.5.mlp.fc2.bias": "model-00002-of-00002.safetensors",
825
+ "vision_tower._vision_tower.vision_model.encoder.layers.5.mlp.fc2.weight": "model-00002-of-00002.safetensors",
826
+ "vision_tower._vision_tower.vision_model.encoder.layers.5.self_attn.k_proj.bias": "model-00002-of-00002.safetensors",
827
+ "vision_tower._vision_tower.vision_model.encoder.layers.5.self_attn.k_proj.weight": "model-00002-of-00002.safetensors",
828
+ "vision_tower._vision_tower.vision_model.encoder.layers.5.self_attn.out_proj.bias": "model-00002-of-00002.safetensors",
829
+ "vision_tower._vision_tower.vision_model.encoder.layers.5.self_attn.out_proj.weight": "model-00002-of-00002.safetensors",
830
+ "vision_tower._vision_tower.vision_model.encoder.layers.5.self_attn.q_proj.bias": "model-00002-of-00002.safetensors",
831
+ "vision_tower._vision_tower.vision_model.encoder.layers.5.self_attn.q_proj.weight": "model-00002-of-00002.safetensors",
832
+ "vision_tower._vision_tower.vision_model.encoder.layers.5.self_attn.v_proj.bias": "model-00002-of-00002.safetensors",
833
+ "vision_tower._vision_tower.vision_model.encoder.layers.5.self_attn.v_proj.weight": "model-00002-of-00002.safetensors",
834
+ "vision_tower._vision_tower.vision_model.encoder.layers.6.layer_norm1.bias": "model-00002-of-00002.safetensors",
835
+ "vision_tower._vision_tower.vision_model.encoder.layers.6.layer_norm1.weight": "model-00002-of-00002.safetensors",
836
+ "vision_tower._vision_tower.vision_model.encoder.layers.6.layer_norm2.bias": "model-00002-of-00002.safetensors",
837
+ "vision_tower._vision_tower.vision_model.encoder.layers.6.layer_norm2.weight": "model-00002-of-00002.safetensors",
838
+ "vision_tower._vision_tower.vision_model.encoder.layers.6.mlp.fc1.bias": "model-00002-of-00002.safetensors",
839
+ "vision_tower._vision_tower.vision_model.encoder.layers.6.mlp.fc1.weight": "model-00002-of-00002.safetensors",
840
+ "vision_tower._vision_tower.vision_model.encoder.layers.6.mlp.fc2.bias": "model-00002-of-00002.safetensors",
841
+ "vision_tower._vision_tower.vision_model.encoder.layers.6.mlp.fc2.weight": "model-00002-of-00002.safetensors",
842
+ "vision_tower._vision_tower.vision_model.encoder.layers.6.self_attn.k_proj.bias": "model-00002-of-00002.safetensors",
843
+ "vision_tower._vision_tower.vision_model.encoder.layers.6.self_attn.k_proj.weight": "model-00002-of-00002.safetensors",
844
+ "vision_tower._vision_tower.vision_model.encoder.layers.6.self_attn.out_proj.bias": "model-00002-of-00002.safetensors",
845
+ "vision_tower._vision_tower.vision_model.encoder.layers.6.self_attn.out_proj.weight": "model-00002-of-00002.safetensors",
846
+ "vision_tower._vision_tower.vision_model.encoder.layers.6.self_attn.q_proj.bias": "model-00002-of-00002.safetensors",
847
+ "vision_tower._vision_tower.vision_model.encoder.layers.6.self_attn.q_proj.weight": "model-00002-of-00002.safetensors",
848
+ "vision_tower._vision_tower.vision_model.encoder.layers.6.self_attn.v_proj.bias": "model-00002-of-00002.safetensors",
849
+ "vision_tower._vision_tower.vision_model.encoder.layers.6.self_attn.v_proj.weight": "model-00002-of-00002.safetensors",
850
+ "vision_tower._vision_tower.vision_model.encoder.layers.7.layer_norm1.bias": "model-00002-of-00002.safetensors",
851
+ "vision_tower._vision_tower.vision_model.encoder.layers.7.layer_norm1.weight": "model-00002-of-00002.safetensors",
852
+ "vision_tower._vision_tower.vision_model.encoder.layers.7.layer_norm2.bias": "model-00002-of-00002.safetensors",
853
+ "vision_tower._vision_tower.vision_model.encoder.layers.7.layer_norm2.weight": "model-00002-of-00002.safetensors",
854
+ "vision_tower._vision_tower.vision_model.encoder.layers.7.mlp.fc1.bias": "model-00002-of-00002.safetensors",
855
+ "vision_tower._vision_tower.vision_model.encoder.layers.7.mlp.fc1.weight": "model-00002-of-00002.safetensors",
856
+ "vision_tower._vision_tower.vision_model.encoder.layers.7.mlp.fc2.bias": "model-00002-of-00002.safetensors",
857
+ "vision_tower._vision_tower.vision_model.encoder.layers.7.mlp.fc2.weight": "model-00002-of-00002.safetensors",
858
+ "vision_tower._vision_tower.vision_model.encoder.layers.7.self_attn.k_proj.bias": "model-00002-of-00002.safetensors",
859
+ "vision_tower._vision_tower.vision_model.encoder.layers.7.self_attn.k_proj.weight": "model-00002-of-00002.safetensors",
860
+ "vision_tower._vision_tower.vision_model.encoder.layers.7.self_attn.out_proj.bias": "model-00002-of-00002.safetensors",
861
+ "vision_tower._vision_tower.vision_model.encoder.layers.7.self_attn.out_proj.weight": "model-00002-of-00002.safetensors",
862
+ "vision_tower._vision_tower.vision_model.encoder.layers.7.self_attn.q_proj.bias": "model-00002-of-00002.safetensors",
863
+ "vision_tower._vision_tower.vision_model.encoder.layers.7.self_attn.q_proj.weight": "model-00002-of-00002.safetensors",
864
+ "vision_tower._vision_tower.vision_model.encoder.layers.7.self_attn.v_proj.bias": "model-00002-of-00002.safetensors",
865
+ "vision_tower._vision_tower.vision_model.encoder.layers.7.self_attn.v_proj.weight": "model-00002-of-00002.safetensors",
866
+ "vision_tower._vision_tower.vision_model.encoder.layers.8.layer_norm1.bias": "model-00002-of-00002.safetensors",
867
+ "vision_tower._vision_tower.vision_model.encoder.layers.8.layer_norm1.weight": "model-00002-of-00002.safetensors",
868
+ "vision_tower._vision_tower.vision_model.encoder.layers.8.layer_norm2.bias": "model-00002-of-00002.safetensors",
869
+ "vision_tower._vision_tower.vision_model.encoder.layers.8.layer_norm2.weight": "model-00002-of-00002.safetensors",
870
+ "vision_tower._vision_tower.vision_model.encoder.layers.8.mlp.fc1.bias": "model-00002-of-00002.safetensors",
871
+ "vision_tower._vision_tower.vision_model.encoder.layers.8.mlp.fc1.weight": "model-00002-of-00002.safetensors",
872
+ "vision_tower._vision_tower.vision_model.encoder.layers.8.mlp.fc2.bias": "model-00002-of-00002.safetensors",
873
+ "vision_tower._vision_tower.vision_model.encoder.layers.8.mlp.fc2.weight": "model-00002-of-00002.safetensors",
874
+ "vision_tower._vision_tower.vision_model.encoder.layers.8.self_attn.k_proj.bias": "model-00002-of-00002.safetensors",
875
+ "vision_tower._vision_tower.vision_model.encoder.layers.8.self_attn.k_proj.weight": "model-00002-of-00002.safetensors",
876
+ "vision_tower._vision_tower.vision_model.encoder.layers.8.self_attn.out_proj.bias": "model-00002-of-00002.safetensors",
877
+ "vision_tower._vision_tower.vision_model.encoder.layers.8.self_attn.out_proj.weight": "model-00002-of-00002.safetensors",
878
+ "vision_tower._vision_tower.vision_model.encoder.layers.8.self_attn.q_proj.bias": "model-00002-of-00002.safetensors",
879
+ "vision_tower._vision_tower.vision_model.encoder.layers.8.self_attn.q_proj.weight": "model-00002-of-00002.safetensors",
880
+ "vision_tower._vision_tower.vision_model.encoder.layers.8.self_attn.v_proj.bias": "model-00002-of-00002.safetensors",
881
+ "vision_tower._vision_tower.vision_model.encoder.layers.8.self_attn.v_proj.weight": "model-00002-of-00002.safetensors",
882
+ "vision_tower._vision_tower.vision_model.encoder.layers.9.layer_norm1.bias": "model-00002-of-00002.safetensors",
883
+ "vision_tower._vision_tower.vision_model.encoder.layers.9.layer_norm1.weight": "model-00002-of-00002.safetensors",
884
+ "vision_tower._vision_tower.vision_model.encoder.layers.9.layer_norm2.bias": "model-00002-of-00002.safetensors",
885
+ "vision_tower._vision_tower.vision_model.encoder.layers.9.layer_norm2.weight": "model-00002-of-00002.safetensors",
886
+ "vision_tower._vision_tower.vision_model.encoder.layers.9.mlp.fc1.bias": "model-00002-of-00002.safetensors",
887
+ "vision_tower._vision_tower.vision_model.encoder.layers.9.mlp.fc1.weight": "model-00002-of-00002.safetensors",
888
+ "vision_tower._vision_tower.vision_model.encoder.layers.9.mlp.fc2.bias": "model-00002-of-00002.safetensors",
889
+ "vision_tower._vision_tower.vision_model.encoder.layers.9.mlp.fc2.weight": "model-00002-of-00002.safetensors",
890
+ "vision_tower._vision_tower.vision_model.encoder.layers.9.self_attn.k_proj.bias": "model-00002-of-00002.safetensors",
891
+ "vision_tower._vision_tower.vision_model.encoder.layers.9.self_attn.k_proj.weight": "model-00002-of-00002.safetensors",
892
+ "vision_tower._vision_tower.vision_model.encoder.layers.9.self_attn.out_proj.bias": "model-00002-of-00002.safetensors",
893
+ "vision_tower._vision_tower.vision_model.encoder.layers.9.self_attn.out_proj.weight": "model-00002-of-00002.safetensors",
894
+ "vision_tower._vision_tower.vision_model.encoder.layers.9.self_attn.q_proj.bias": "model-00002-of-00002.safetensors",
895
+ "vision_tower._vision_tower.vision_model.encoder.layers.9.self_attn.q_proj.weight": "model-00002-of-00002.safetensors",
896
+ "vision_tower._vision_tower.vision_model.encoder.layers.9.self_attn.v_proj.bias": "model-00002-of-00002.safetensors",
897
+ "vision_tower._vision_tower.vision_model.encoder.layers.9.self_attn.v_proj.weight": "model-00002-of-00002.safetensors",
898
+ "vision_tower._vision_tower.vision_model.head.attention.in_proj_bias": "model-00002-of-00002.safetensors",
899
+ "vision_tower._vision_tower.vision_model.head.attention.in_proj_weight": "model-00002-of-00002.safetensors",
900
+ "vision_tower._vision_tower.vision_model.head.attention.out_proj.bias": "model-00002-of-00002.safetensors",
901
+ "vision_tower._vision_tower.vision_model.head.attention.out_proj.weight": "model-00002-of-00002.safetensors",
902
+ "vision_tower._vision_tower.vision_model.head.layernorm.bias": "model-00002-of-00002.safetensors",
903
+ "vision_tower._vision_tower.vision_model.head.layernorm.weight": "model-00002-of-00002.safetensors",
904
+ "vision_tower._vision_tower.vision_model.head.mlp.fc1.bias": "model-00002-of-00002.safetensors",
905
+ "vision_tower._vision_tower.vision_model.head.mlp.fc1.weight": "model-00002-of-00002.safetensors",
906
+ "vision_tower._vision_tower.vision_model.head.mlp.fc2.bias": "model-00002-of-00002.safetensors",
907
+ "vision_tower._vision_tower.vision_model.head.mlp.fc2.weight": "model-00002-of-00002.safetensors",
908
+ "vision_tower._vision_tower.vision_model.head.probe": "model-00002-of-00002.safetensors",
909
+ "vision_tower._vision_tower.vision_model.post_layernorm.bias": "model-00002-of-00002.safetensors",
910
+ "vision_tower._vision_tower.vision_model.post_layernorm.weight": "model-00002-of-00002.safetensors"
911
+ }
912
+ }
checkpoint-21520/modeling_tinyllava_phi.py ADDED
@@ -0,0 +1,624 @@
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
1
+ # For licensing see accompanying LICENSE file.
2
+ # Copyright (C) 2024 TinyLLaVA. All Rights Reserved.
3
+ import time
4
+
5
+ import dataclasses
6
+ from enum import auto, Enum
7
+ from typing import List, Tuple, Optional, Union
8
+ import requests
9
+ from PIL import Image
10
+ from io import BytesIO
11
+ import base64
12
+ import re
13
+
14
+ import torch
15
+ import torch.utils.checkpoint
16
+ from torch import nn
17
+ from torch.nn import functional as F
18
+
19
+ from transformers.utils import logging
20
+ from transformers import PreTrainedModel
21
+ from transformers.modeling_outputs import CausalLMOutputWithPast
22
+ from transformers.generation.utils import GenerateOutput
23
+ from transformers import CLIPVisionModel, CLIPImageProcessor, SiglipVisionModel, SiglipImageProcessor
24
+
25
+ from .configuration import TinyLlavaConfig, IGNORE_INDEX, IMAGE_TOKEN_INDEX, DEFAULT_IMAGE_TOKEN
26
+
27
+ from transformers import AutoConfig, AutoModelForCausalLM, PhiForCausalLM
28
+
29
+
30
+
31
+ logger = logging.get_logger(__name__)
32
+
33
+ # Model Constants
34
+ IGNORE_INDEX = -100
35
+ IMAGE_TOKEN_INDEX = -200
36
+ DEFAULT_IMAGE_TOKEN = "<image>"
37
+ DEFAULT_IMAGE_PATCH_TOKEN = "<im_patch>"
38
+ DEFAULT_IM_START_TOKEN = "<im_start>"
39
+ DEFAULT_IM_END_TOKEN = "<im_end>"
40
+ IMAGE_PLACEHOLDER = "<image-placeholder>"
41
+
42
+ CONTROLLER_HEART_BEAT_EXPIRATION = 30
43
+ WORKER_HEART_BEAT_INTERVAL = 15
44
+ LOGDIR = "."
45
+
46
+
47
+ class SeparatorStyle(Enum):
48
+ """Different separator style."""
49
+ SINGLE = auto()
50
+ TWO = auto()
51
+ MPT = auto()
52
+ PLAIN = auto()
53
+ LLAMA_2 = auto()
54
+ TINY_LLAMA = auto()
55
+ QWEN_2 = auto()
56
+
57
+
58
+ @dataclasses.dataclass
59
+ class Conversation:
60
+ """A class that keeps all conversation history."""
61
+ system: str
62
+ roles: List[str]
63
+ messages: List[List[str]]
64
+ offset: int
65
+ sep_style: SeparatorStyle = SeparatorStyle.SINGLE
66
+ sep: str = "###"
67
+ sep2: str = None
68
+ version: str = "Unknown"
69
+
70
+ skip_next: bool = False
71
+
72
+ def get_prompt(self):
73
+ messages = self.messages
74
+ if len(messages) > 0 and type(messages[0][1]) is tuple:
75
+ messages = self.messages.copy()
76
+ init_role, init_msg = messages[0].copy()
77
+ init_msg = init_msg[0].replace("<image>", "").strip()
78
+ if 'mmtag' in self.version:
79
+ messages[0] = (init_role, init_msg)
80
+ messages.insert(0, (self.roles[0], "<Image><image></Image>"))
81
+ messages.insert(1, (self.roles[1], "Received."))
82
+ else:
83
+ messages[0] = (init_role, "<image>\n" + init_msg)
84
+
85
+ if self.sep_style == SeparatorStyle.TWO:
86
+ seps = [self.sep, self.sep2]
87
+ ret = self.system + seps[0]
88
+ for i, (role, message) in enumerate(messages):
89
+ if message:
90
+ if type(message) is tuple:
91
+ message, _, _ = message
92
+ ret += role + ": " + message + seps[i % 2]
93
+ else:
94
+ ret += role + ":"
95
+ else:
96
+ raise ValueError(f"Invalid style: {self.sep_style}")
97
+
98
+ return ret
99
+
100
+ def append_message(self, role, message):
101
+ self.messages.append([role, message])
102
+
103
+ def copy(self):
104
+ return Conversation(
105
+ system=self.system,
106
+ roles=self.roles,
107
+ messages=[[x, y] for x, y in self.messages],
108
+ offset=self.offset,
109
+ sep_style=self.sep_style,
110
+ sep=self.sep,
111
+ sep2=self.sep2,
112
+ version=self.version)
113
+
114
+
115
+
116
+
117
+ conv_phi_v0 = Conversation(
118
+ system="A chat between a curious user and an artificial intelligence assistant. "
119
+ "The assistant gives helpful, detailed, and polite answers to the user's questions.",
120
+ roles=("USER", "ASSISTANT"),
121
+ version="phi",
122
+ messages=(),
123
+ offset=0,
124
+ sep_style=SeparatorStyle.TWO,
125
+ sep=" ",
126
+ sep2="<|endoftext|>",
127
+ )
128
+
129
+
130
+ def load_image_from_base64(image):
131
+ return Image.open(BytesIO(base64.b64decode(image)))
132
+
133
+
134
+ def expand2square(pil_img, background_color):
135
+ width, height = pil_img.size
136
+ if width == height:
137
+ return pil_img
138
+ elif width > height:
139
+ result = Image.new(pil_img.mode, (width, width), background_color)
140
+ result.paste(pil_img, (0, (width - height) // 2))
141
+ return result
142
+ else:
143
+ result = Image.new(pil_img.mode, (height, height), background_color)
144
+ result.paste(pil_img, ((height - width) // 2, 0))
145
+ return result
146
+
147
+
148
+ def process_images(images, image_processor, model_cfg):
149
+ image_aspect_ratio = getattr(model_cfg, "image_aspect_ratio", None)
150
+ new_images = []
151
+ if image_aspect_ratio == 'pad':
152
+ for image in images:
153
+ image = expand2square(image, tuple(int(x*255) for x in image_processor.image_mean))
154
+ image = image_processor.preprocess(image, return_tensors='pt')['pixel_values'][0]
155
+ new_images.append(image)
156
+ else:
157
+ return image_processor(images, return_tensors='pt')['pixel_values']
158
+ if all(x.shape == new_images[0].shape for x in new_images):
159
+ new_images = torch.stack(new_images, dim=0)
160
+ return new_images
161
+
162
+
163
+ def tokenizer_image_token(prompt, tokenizer, image_token_index=IMAGE_TOKEN_INDEX, return_tensors=None):
164
+ prompt_chunks = [tokenizer(chunk).input_ids for chunk in prompt.split('<image>')]
165
+
166
+ def insert_separator(X, sep):
167
+ return [ele for sublist in zip(X, [sep]*len(X)) for ele in sublist][:-1]
168
+
169
+ input_ids = []
170
+ offset = 0
171
+ if len(prompt_chunks) > 0 and len(prompt_chunks[0]) > 0 and prompt_chunks[0][0] == tokenizer.bos_token_id:
172
+ offset = 1
173
+ input_ids.append(prompt_chunks[0][0])
174
+
175
+ for x in insert_separator(prompt_chunks, [image_token_index] * (offset + 1)):
176
+ input_ids.extend(x[offset:])
177
+
178
+ if return_tensors is not None:
179
+ if return_tensors == 'pt':
180
+ return torch.tensor(input_ids, dtype=torch.long)
181
+ raise ValueError(f'Unsupported tensor type: {return_tensors}')
182
+ return input_ids
183
+
184
+ def load_image(image_file):
185
+ if image_file.startswith("http") or image_file.startswith("https"):
186
+ response = requests.get(image_file)
187
+ image = Image.open(BytesIO(response.content)).convert("RGB")
188
+ else:
189
+ image = Image.open(image_file).convert("RGB")
190
+ return image
191
+
192
+ ACT_TYPE = {
193
+ 'relu': nn.ReLU,
194
+ 'gelu': nn.GELU
195
+ }
196
+
197
+ class Connector(nn.Module):
198
+ def __init__(self, config=None):
199
+ super().__init__()
200
+ mlp_gelu_match = re.match(r'^mlp(\d+)x_gelu$', config.connector_type)
201
+ act_type = config.connector_type.split('_')[-1]
202
+ mlp_depth = int(mlp_gelu_match.group(1))
203
+ modules = [nn.Linear(config.vision_hidden_size, config.hidden_size)]
204
+ for _ in range(1, mlp_depth):
205
+ modules.append(ACT_TYPE[act_type]())
206
+ modules.append(nn.Linear(config.hidden_size, config.hidden_size))
207
+
208
+ self._connector = nn.Sequential(*modules)
209
+
210
+ def forward(self, x):
211
+ return self._connector(x)
212
+
213
+ class VisionTower(nn.Module):
214
+ def __init__(self, cfg, model_name_or_path = 'clip'):
215
+ super().__init__()
216
+ if 'clip' in model_name_or_path:
217
+ self._vision_tower = CLIPVisionModel(cfg)
218
+ self._image_processor = CLIPImageProcessor.from_pretrained(cfg.model_name_or_path)
219
+ else:
220
+ self._vision_tower = SiglipVisionModel(cfg)
221
+ self._image_processor = SiglipImageProcessor.from_pretrained(cfg.model_name_or_path)
222
+
223
+ self.config = cfg
224
+
225
+ def forward(self, x, **kwargs):
226
+ image_features = self._vision_tower(x, output_hidden_states=True)
227
+ image_features = image_features.hidden_states[kwargs.get('vision_feature_layer', -2)]
228
+
229
+ if kwargs.get('vision_feature_select_strategy', 'patch') == 'patch':
230
+ image_features = image_features[:, 1:]
231
+ elif kwargs.get('vision_feature_select_strategy', 'patch') == 'cls_patch':
232
+ image_features = image_features
233
+ else:
234
+ raise ValueError(f"Unexpected select feature: {kwargs.get('vision_feature_select_strategy')}")
235
+
236
+ return image_features
237
+
238
+ @property
239
+ def vision_tower(self):
240
+ return self._vision_tower
241
+
242
+ @vision_tower.setter
243
+ def vision_tower(self, vision_tower):
244
+ self._vision_tower = vision_tower
245
+
246
+ def get_value_from_kwargs(kwargs, name):
247
+ if name in kwargs:
248
+ return kwargs.pop(name)
249
+ else:
250
+ return None
251
+
252
+
253
+ class TinyLlavaPreTrainedModel(PreTrainedModel):
254
+ config_class = TinyLlavaConfig
255
+ base_model_prefix = "model"
256
+ supports_gradient_checkpointing = True
257
+ _no_split_modules = ["LlavaVisionAttention"]
258
+ _skip_keys_device_placement = "past_key_values"
259
+ _supports_flash_attn_2 = True
260
+
261
+ def _init_weights(self, module):
262
+ std = (
263
+ self.config.initializer_range
264
+ if hasattr(self.config, "initializer_range")
265
+ else self.config.text_config.initializer_range
266
+ )
267
+
268
+ if hasattr(module, "class_embedding"):
269
+ module.class_embedding.data.normal_(mean=0.0, std=std)
270
+
271
+ if isinstance(module, (nn.Linear, nn.Conv2d)):
272
+ module.weight.data.normal_(mean=0.0, std=std)
273
+ if module.bias is not None:
274
+ module.bias.data.zero_()
275
+ elif isinstance(module, nn.Embedding):
276
+ module.weight.data.normal_(mean=0.0, std=std)
277
+ if module.padding_idx is not None:
278
+ module.weight.data[module.padding_idx].zero_()
279
+
280
+ @property
281
+ def _supports_sdpa(self):
282
+ return self.language_model._supports_sdpa
283
+
284
+
285
+ class TinyLlavaForConditionalGeneration(TinyLlavaPreTrainedModel):
286
+ def __init__(self, config: TinyLlavaConfig):
287
+
288
+ super().__init__(config)
289
+
290
+ self.language_model = PhiForCausalLM(config.text_config)
291
+ self.vision_tower = VisionTower(config.vision_config, config.vision_model_name_or_path)
292
+ self.connector = Connector(config)
293
+ self.post_init()
294
+
295
+
296
+ def get_input_embeddings(self):
297
+ return self.language_model.get_input_embeddings()
298
+
299
+ def set_input_embeddings(self, value):
300
+ self.language_model.set_input_embeddings(value)
301
+
302
+ def get_output_embeddings(self):
303
+ return self.language_model.get_output_embeddings()
304
+
305
+ def set_output_embeddings(self, new_embeddings):
306
+ self.language_model.set_output_embeddings(new_embeddings)
307
+
308
+ def set_decoder(self, decoder):
309
+ self.language_model.set_decoder(decoder)
310
+
311
+ def get_decoder(self):
312
+ return self.language_model.get_decoder()
313
+
314
+ def tie_weights(self):
315
+ return self.language_model.tie_weights()
316
+
317
+ def resize_token_embeddings(self, new_num_tokens: Optional[int] = None, pad_to_multiple_of=None) -> nn.Embedding:
318
+ model_embeds = self.language_model.resize_token_embeddings(new_num_tokens, pad_to_multiple_of)
319
+ # update vocab size
320
+ self.config.text_config.vocab_size = model_embeds.num_embeddings
321
+ self.config.vocab_size = model_embeds.num_embeddings
322
+ self.vocab_size = model_embeds.num_embeddings
323
+ return model_embeds
324
+
325
+
326
+ def forward(
327
+ self,
328
+ input_ids: torch.LongTensor = None,
329
+ attention_mask: Optional[torch.Tensor] = None,
330
+ position_ids: Optional[torch.LongTensor] = None,
331
+ past_key_values: Optional[List[torch.FloatTensor]] = None,
332
+ inputs_embeds: Optional[torch.FloatTensor] = None,
333
+ labels: Optional[torch.LongTensor] = None,
334
+ use_cache: Optional[bool] = None,
335
+ output_attentions: Optional[bool] = None,
336
+ output_hidden_states: Optional[bool] = None,
337
+ images: Optional[torch.FloatTensor] = None,
338
+ image_sizes: Optional[List[List[int]]] = None,
339
+ return_dict: Optional[bool] = None,
340
+ ) -> Union[Tuple, CausalLMOutputWithPast]:
341
+ use_cache = use_cache if use_cache is not None else self.config.use_cache
342
+ if inputs_embeds is None:
343
+ (
344
+ input_ids,
345
+ position_ids,
346
+ attention_mask,
347
+ past_key_values,
348
+ inputs_embeds,
349
+ labels
350
+ ) = self.prepare_inputs_labels_for_multimodal(
351
+ input_ids,
352
+ position_ids,
353
+ attention_mask,
354
+ past_key_values,
355
+ labels,
356
+ images,
357
+ image_sizes
358
+ )
359
+ return self.language_model.forward(
360
+ input_ids=input_ids,
361
+ attention_mask=attention_mask,
362
+ position_ids=position_ids,
363
+ past_key_values=past_key_values,
364
+ inputs_embeds=inputs_embeds,
365
+ labels=labels,
366
+ use_cache=use_cache,
367
+ output_attentions=output_attentions,
368
+ output_hidden_states=output_hidden_states,
369
+ return_dict=return_dict
370
+ )
371
+
372
+ @torch.no_grad()
373
+ def generate(
374
+ self,
375
+ inputs: Optional[torch.Tensor] = None,
376
+ images: Optional[torch.Tensor] = None,
377
+ image_sizes: Optional[torch.Tensor] = None,
378
+ **kwargs,
379
+ ) -> Union[GenerateOutput, torch.LongTensor]:
380
+ position_ids = kwargs.pop("position_ids", None)
381
+ attention_mask = kwargs.pop("attention_mask", None)
382
+ if "inputs_embeds" in kwargs:
383
+ raise NotImplementedError("`inputs_embeds` is not supported")
384
+
385
+ if images is not None:
386
+ (
387
+ inputs,
388
+ position_ids,
389
+ attention_mask,
390
+ _,
391
+ inputs_embeds,
392
+ _
393
+ ) = self.prepare_inputs_labels_for_multimodal(
394
+ inputs,
395
+ position_ids,
396
+ attention_mask,
397
+ None,
398
+ None,
399
+ images,
400
+ image_sizes=image_sizes
401
+ )
402
+ else:
403
+ inputs_embeds = self.language_model.get_input_embeddings()(inputs)
404
+
405
+ return self.language_model.generate(
406
+ position_ids=position_ids,
407
+ attention_mask=attention_mask,
408
+ inputs_embeds=inputs_embeds,
409
+ **kwargs
410
+ )
411
+
412
+ def encode_images(self, images):
413
+ kwargs = {}
414
+ kwargs['vision_feature_layer'] = self.config.vision_feature_layer
415
+ kwargs['vision_feature_select_strategy'] = self.config.vision_feature_select_strategy
416
+ images = images.to(device=self.device, dtype=self.dtype)
417
+ image_features = self.vision_tower(images, **kwargs)
418
+ image_features = self.connector(image_features)
419
+ return image_features
420
+
421
+
422
+
423
+ def prepare_inputs_for_generation(self, input_ids, past_key_values=None,
424
+ inputs_embeds=None, **kwargs):
425
+ images = kwargs.pop("images", None)
426
+ image_sizes = kwargs.pop("image_sizes", None)
427
+ inputs = self.language_model.prepare_inputs_for_generation(
428
+ input_ids, past_key_values=past_key_values, inputs_embeds=inputs_embeds, **kwargs
429
+ )
430
+ if images is not None:
431
+ inputs['images'] = images
432
+ if image_sizes is not None:
433
+ inputs['image_sizes'] = image_sizes
434
+ return inputs
435
+
436
+ def prepare_inputs_labels_for_multimodal(
437
+ self, input_ids, position_ids, attention_mask, past_key_values, labels,
438
+ images, image_sizes=None
439
+ ):
440
+ vision_tower = self.vision_tower
441
+ if vision_tower is None or images is None or input_ids.shape[1] == 1:
442
+ return input_ids, position_ids, attention_mask, past_key_values, None, labels
443
+
444
+
445
+ image_features = self.encode_images(images)
446
+
447
+ # TODO: image start / end is not implemented here to support pretraining.
448
+ if getattr(self.config, 'tune_mm_mlp_adapter', False):
449
+ raise NotImplementedError
450
+
451
+ # Let's just add dummy tensors if they do not exist,
452
+ # it is a headache to deal with None all the time.
453
+ # But it is not ideal, and if you have a better idea,
454
+ # please open an issue / submit a PR, thanks.
455
+ _labels = labels
456
+ _position_ids = position_ids
457
+ _attention_mask = attention_mask
458
+ if attention_mask is None:
459
+ attention_mask = torch.ones_like(input_ids, dtype=torch.bool)
460
+ else:
461
+ attention_mask = attention_mask.bool()
462
+ if position_ids is None:
463
+ position_ids = torch.arange(0, input_ids.shape[1], dtype=torch.long, device=input_ids.device)
464
+ if labels is None:
465
+ labels = torch.full_like(input_ids, IGNORE_INDEX)
466
+
467
+ # remove the padding using attention_mask -- FIXME
468
+ _input_ids = input_ids
469
+ input_ids = [cur_input_ids[cur_attention_mask] for cur_input_ids, cur_attention_mask in zip(input_ids, attention_mask)]
470
+ labels = [cur_labels[cur_attention_mask] for cur_labels, cur_attention_mask in zip(labels, attention_mask)]
471
+
472
+ new_input_embeds = []
473
+ new_labels = []
474
+ cur_image_idx = 0
475
+ for batch_idx, cur_input_ids in enumerate(input_ids):
476
+ num_images = (cur_input_ids == IMAGE_TOKEN_INDEX).sum()
477
+ if num_images == 0:
478
+ cur_image_features = image_features[cur_image_idx]
479
+ cur_input_embeds_1 = self.language_model.get_input_embeddings()(cur_input_ids)
480
+ cur_input_embeds = torch.cat([cur_input_embeds_1, cur_image_features[0:0]], dim=0)
481
+ new_input_embeds.append(cur_input_embeds)
482
+ new_labels.append(labels[batch_idx])
483
+ cur_image_idx += 1
484
+ continue
485
+
486
+ image_token_indices = [-1] + torch.where(cur_input_ids == IMAGE_TOKEN_INDEX)[0].tolist() + [cur_input_ids.shape[0]]
487
+ cur_input_ids_noim = []
488
+ cur_labels = labels[batch_idx]
489
+ cur_labels_noim = []
490
+ for i in range(len(image_token_indices) - 1):
491
+ cur_input_ids_noim.append(cur_input_ids[image_token_indices[i]+1:image_token_indices[i+1]])
492
+ cur_labels_noim.append(cur_labels[image_token_indices[i]+1:image_token_indices[i+1]])
493
+ split_sizes = [x.shape[0] for x in cur_labels_noim]
494
+ cur_input_embeds = self.language_model.get_input_embeddings()(torch.cat(cur_input_ids_noim))
495
+ cur_input_embeds_no_im = torch.split(cur_input_embeds, split_sizes, dim=0)
496
+ cur_new_input_embeds = []
497
+ cur_new_labels = []
498
+
499
+ for i in range(num_images + 1):
500
+ cur_new_input_embeds.append(cur_input_embeds_no_im[i])
501
+ cur_new_labels.append(cur_labels_noim[i])
502
+ if i < num_images:
503
+ cur_image_features = image_features[cur_image_idx]
504
+ cur_image_idx += 1
505
+ cur_new_input_embeds.append(cur_image_features)
506
+ cur_new_labels.append(torch.full((cur_image_features.shape[0],), IGNORE_INDEX, device=cur_labels.device, dtype=cur_labels.dtype))
507
+
508
+ cur_new_input_embeds = [x.to(self.device) for x in cur_new_input_embeds]
509
+
510
+ cur_new_input_embeds = torch.cat(cur_new_input_embeds)
511
+ cur_new_labels = torch.cat(cur_new_labels)
512
+
513
+ new_input_embeds.append(cur_new_input_embeds)
514
+ new_labels.append(cur_new_labels)
515
+
516
+ # Truncate sequences to max length as image embeddings can make the sequence longer
517
+ tokenizer_model_max_length = getattr(self.config, 'tokenizer_model_max_length', None)
518
+ if tokenizer_model_max_length is not None:
519
+ new_input_embeds = [x[:tokenizer_model_max_length] for x in new_input_embeds]
520
+ new_labels = [x[:tokenizer_model_max_length] for x in new_labels]
521
+
522
+ # Combine them
523
+ max_len = max(x.shape[0] for x in new_input_embeds)
524
+ batch_size = len(new_input_embeds)
525
+
526
+ new_input_embeds_padded = []
527
+ new_labels_padded = torch.full((batch_size, max_len), IGNORE_INDEX, dtype=new_labels[0].dtype, device=new_labels[0].device)
528
+ attention_mask = torch.zeros((batch_size, max_len), dtype=attention_mask.dtype, device=attention_mask.device)
529
+ position_ids = torch.zeros((batch_size, max_len), dtype=position_ids.dtype, device=position_ids.device)
530
+
531
+ for i, (cur_new_embed, cur_new_labels) in enumerate(zip(new_input_embeds, new_labels)):
532
+ cur_len = cur_new_embed.shape[0]
533
+ if getattr(self.config, 'tokenizer_padding_side', 'right') == "left":
534
+ new_input_embeds_padded.append(torch.cat((
535
+ torch.zeros((max_len - cur_len, cur_new_embed.shape[1]), dtype=cur_new_embed.dtype, device=cur_new_embed.device),
536
+ cur_new_embed
537
+ ), dim=0))
538
+ if cur_len > 0:
539
+ new_labels_padded[i, -cur_len:] = cur_new_labels
540
+ attention_mask[i, -cur_len:] = True
541
+ position_ids[i, -cur_len:] = torch.arange(0, cur_len, dtype=position_ids.dtype, device=position_ids.device)
542
+ else:
543
+ new_input_embeds_padded.append(torch.cat((
544
+ cur_new_embed,
545
+ torch.zeros((max_len - cur_len, cur_new_embed.shape[1]), dtype=cur_new_embed.dtype, device=cur_new_embed.device)
546
+ ), dim=0))
547
+ if cur_len > 0:
548
+ new_labels_padded[i, :cur_len] = cur_new_labels
549
+ attention_mask[i, :cur_len] = True
550
+ position_ids[i, :cur_len] = torch.arange(0, cur_len, dtype=position_ids.dtype, device=position_ids.device)
551
+
552
+ new_input_embeds = torch.stack(new_input_embeds_padded, dim=0)
553
+
554
+ if _labels is None:
555
+ new_labels = None
556
+ else:
557
+ new_labels = new_labels_padded
558
+
559
+ if _attention_mask is None:
560
+ attention_mask = None
561
+ else:
562
+ attention_mask = attention_mask.to(dtype=_attention_mask.dtype)
563
+
564
+ if _position_ids is None:
565
+ position_ids = None
566
+
567
+ return None, position_ids, attention_mask, past_key_values, new_input_embeds, new_labels
568
+
569
+ def chat(
570
+ self,
571
+ prompt: str,
572
+ tokenizer = None,
573
+ image: str = None,
574
+ max_new_tokens: int = 512,
575
+ num_beams = 1,
576
+ top_p=None,
577
+ temperature=0
578
+ ):
579
+ image_processor = self.vision_tower._image_processor
580
+
581
+ if image is not None:
582
+ prompt = DEFAULT_IMAGE_TOKEN + '\n' + prompt
583
+ conv = conv_phi_v0.copy()
584
+ conv.append_message(conv.roles[0], prompt)
585
+ conv.append_message(conv.roles[1], None)
586
+ prompt = conv.get_prompt()
587
+ if image is not None:
588
+ image = load_image(image)
589
+ image_tensor = process_images(image, image_processor, self.config).to(self.device)
590
+
591
+ input_ids = (
592
+ tokenizer_image_token(prompt, tokenizer, IMAGE_TOKEN_INDEX, return_tensors="pt")
593
+ .unsqueeze(0).to(self.device)
594
+ )
595
+ # Generate
596
+ stime = time.time()
597
+
598
+ with torch.inference_mode():
599
+ output_ids = self.generate(
600
+ input_ids,
601
+ images=image_tensor,
602
+ do_sample=True if temperature > 0 else False,
603
+ temperature=temperature,
604
+ top_p=top_p,
605
+ num_beams=num_beams,
606
+ pad_token_id=tokenizer.pad_token_id,
607
+ max_new_tokens=max_new_tokens,
608
+ use_cache=True,
609
+ # stopping_criteria=[stopping_criteria],
610
+ )
611
+
612
+ # print('inference over')
613
+ generation_time = time.time() - stime
614
+ outputs = tokenizer.batch_decode(
615
+ output_ids, skip_special_tokens=True
616
+ )[0]
617
+
618
+ outputs = outputs.strip()
619
+
620
+ return outputs, generation_time
621
+
622
+
623
+ AutoConfig.register("tinyllava", TinyLlavaConfig)
624
+ AutoModelForCausalLM.register(TinyLlavaConfig, TinyLlavaForConditionalGeneration)
checkpoint-21520/rng_state_0.pth ADDED
@@ -0,0 +1,3 @@
 
 
 
 
1
+ version https://git-lfs.github.com/spec/v1
2
+ oid sha256:51fd2d51f62075e54e1a7ef560b551524db837b3e6184fa11d84562ee93808d7
3
+ size 17655
checkpoint-21520/rng_state_1.pth ADDED
@@ -0,0 +1,3 @@
 
 
 
 
1
+ version https://git-lfs.github.com/spec/v1
2
+ oid sha256:28919776331e33e0005de492038640a1dd8222cc36ed6c2791099998646ca0f7
3
+ size 17655
checkpoint-21520/rng_state_2.pth ADDED
@@ -0,0 +1,3 @@
 
 
 
 
1
+ version https://git-lfs.github.com/spec/v1
2
+ oid sha256:9f7cc2b3d88231c6e656c0d60c946e755df510288c611cb6c8ea7b676039a8de
3
+ size 17655
checkpoint-21520/rng_state_3.pth ADDED
@@ -0,0 +1,3 @@
 
 
 
 
1
+ version https://git-lfs.github.com/spec/v1
2
+ oid sha256:fe414a689e978c23c6e1c8dbabd1c87aa6a1618d8c1ba4cffd59a2d21333309c
3
+ size 17655
checkpoint-21520/scheduler.pt ADDED
@@ -0,0 +1,3 @@
 
 
 
 
1
+ version https://git-lfs.github.com/spec/v1
2
+ oid sha256:2ba305a64c44903f7521f61b7377985052181f7864b783fc5ec4f4531a819008
3
+ size 627
checkpoint-21520/special_tokens_map.json ADDED
@@ -0,0 +1,30 @@
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
1
+ {
2
+ "bos_token": {
3
+ "content": "<|endoftext|>",
4
+ "lstrip": false,
5
+ "normalized": false,
6
+ "rstrip": false,
7
+ "single_word": false
8
+ },
9
+ "eos_token": {
10
+ "content": "<|endoftext|>",
11
+ "lstrip": false,
12
+ "normalized": false,
13
+ "rstrip": false,
14
+ "single_word": false
15
+ },
16
+ "pad_token": {
17
+ "content": "<|endoftext|>",
18
+ "lstrip": false,
19
+ "normalized": false,
20
+ "rstrip": false,
21
+ "single_word": false
22
+ },
23
+ "unk_token": {
24
+ "content": "<|endoftext|>",
25
+ "lstrip": false,
26
+ "normalized": false,
27
+ "rstrip": false,
28
+ "single_word": false
29
+ }
30
+ }
checkpoint-21520/tokenizer_config.json ADDED
@@ -0,0 +1,328 @@
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
1
+ {
2
+ "add_bos_token": false,
3
+ "add_prefix_space": false,
4
+ "added_tokens_decoder": {
5
+ "50256": {
6
+ "content": "<|endoftext|>",
7
+ "lstrip": false,
8
+ "normalized": false,
9
+ "rstrip": false,
10
+ "single_word": false,
11
+ "special": true
12
+ },
13
+ "50257": {
14
+ "content": " ",
15
+ "lstrip": false,
16
+ "normalized": true,
17
+ "rstrip": false,
18
+ "single_word": false,
19
+ "special": false
20
+ },
21
+ "50258": {
22
+ "content": " ",
23
+ "lstrip": false,
24
+ "normalized": true,
25
+ "rstrip": false,
26
+ "single_word": false,
27
+ "special": false
28
+ },
29
+ "50259": {
30
+ "content": " ",
31
+ "lstrip": false,
32
+ "normalized": true,
33
+ "rstrip": false,
34
+ "single_word": false,
35
+ "special": false
36
+ },
37
+ "50260": {
38
+ "content": " ",
39
+ "lstrip": false,
40
+ "normalized": true,
41
+ "rstrip": false,
42
+ "single_word": false,
43
+ "special": false
44
+ },
45
+ "50261": {
46
+ "content": " ",
47
+ "lstrip": false,
48
+ "normalized": true,
49
+ "rstrip": false,
50
+ "single_word": false,
51
+ "special": false
52
+ },
53
+ "50262": {
54
+ "content": " ",
55
+ "lstrip": false,
56
+ "normalized": true,
57
+ "rstrip": false,
58
+ "single_word": false,
59
+ "special": false
60
+ },
61
+ "50263": {
62
+ "content": " ",
63
+ "lstrip": false,
64
+ "normalized": true,
65
+ "rstrip": false,
66
+ "single_word": false,
67
+ "special": false
68
+ },
69
+ "50264": {
70
+ "content": " ",
71
+ "lstrip": false,
72
+ "normalized": true,
73
+ "rstrip": false,
74
+ "single_word": false,
75
+ "special": false
76
+ },
77
+ "50265": {
78
+ "content": " ",
79
+ "lstrip": false,
80
+ "normalized": true,
81
+ "rstrip": false,
82
+ "single_word": false,
83
+ "special": false
84
+ },
85
+ "50266": {
86
+ "content": " ",
87
+ "lstrip": false,
88
+ "normalized": true,
89
+ "rstrip": false,
90
+ "single_word": false,
91
+ "special": false
92
+ },
93
+ "50267": {
94
+ "content": " ",
95
+ "lstrip": false,
96
+ "normalized": true,
97
+ "rstrip": false,
98
+ "single_word": false,
99
+ "special": false
100
+ },
101
+ "50268": {
102
+ "content": " ",
103
+ "lstrip": false,
104
+ "normalized": true,
105
+ "rstrip": false,
106
+ "single_word": false,
107
+ "special": false
108
+ },
109
+ "50269": {
110
+ "content": " ",
111
+ "lstrip": false,
112
+ "normalized": true,
113
+ "rstrip": false,
114
+ "single_word": false,
115
+ "special": false
116
+ },
117
+ "50270": {
118
+ "content": " ",
119
+ "lstrip": false,
120
+ "normalized": true,
121
+ "rstrip": false,
122
+ "single_word": false,
123
+ "special": false
124
+ },
125
+ "50271": {
126
+ "content": " ",
127
+ "lstrip": false,
128
+ "normalized": true,
129
+ "rstrip": false,
130
+ "single_word": false,
131
+ "special": false
132
+ },
133
+ "50272": {
134
+ "content": " ",
135
+ "lstrip": false,
136
+ "normalized": true,
137
+ "rstrip": false,
138
+ "single_word": false,
139
+ "special": false
140
+ },
141
+ "50273": {
142
+ "content": " ",
143
+ "lstrip": false,
144
+ "normalized": true,
145
+ "rstrip": false,
146
+ "single_word": false,
147
+ "special": false
148
+ },
149
+ "50274": {
150
+ "content": " ",
151
+ "lstrip": false,
152
+ "normalized": true,
153
+ "rstrip": false,
154
+ "single_word": false,
155
+ "special": false
156
+ },
157
+ "50275": {
158
+ "content": " ",
159
+ "lstrip": false,
160
+ "normalized": true,
161
+ "rstrip": false,
162
+ "single_word": false,
163
+ "special": false
164
+ },
165
+ "50276": {
166
+ "content": " ",
167
+ "lstrip": false,
168
+ "normalized": true,
169
+ "rstrip": false,
170
+ "single_word": false,
171
+ "special": false
172
+ },
173
+ "50277": {
174
+ "content": " ",
175
+ "lstrip": false,
176
+ "normalized": true,
177
+ "rstrip": false,
178
+ "single_word": false,
179
+ "special": false
180
+ },
181
+ "50278": {
182
+ "content": " ",
183
+ "lstrip": false,
184
+ "normalized": true,
185
+ "rstrip": false,
186
+ "single_word": false,
187
+ "special": false
188
+ },
189
+ "50279": {
190
+ "content": " ",
191
+ "lstrip": false,
192
+ "normalized": true,
193
+ "rstrip": false,
194
+ "single_word": false,
195
+ "special": false
196
+ },
197
+ "50280": {
198
+ "content": " ",
199
+ "lstrip": false,
200
+ "normalized": true,
201
+ "rstrip": false,
202
+ "single_word": false,
203
+ "special": false
204
+ },
205
+ "50281": {
206
+ "content": " ",
207
+ "lstrip": false,
208
+ "normalized": true,
209
+ "rstrip": false,
210
+ "single_word": false,
211
+ "special": false
212
+ },
213
+ "50282": {
214
+ "content": " ",
215
+ "lstrip": false,
216
+ "normalized": true,
217
+ "rstrip": false,
218
+ "single_word": false,
219
+ "special": false
220
+ },
221
+ "50283": {
222
+ "content": " ",
223
+ "lstrip": false,
224
+ "normalized": true,
225
+ "rstrip": false,
226
+ "single_word": false,
227
+ "special": false
228
+ },
229
+ "50284": {
230
+ "content": " ",
231
+ "lstrip": false,
232
+ "normalized": true,
233
+ "rstrip": false,
234
+ "single_word": false,
235
+ "special": false
236
+ },
237
+ "50285": {
238
+ "content": " ",
239
+ "lstrip": false,
240
+ "normalized": true,
241
+ "rstrip": false,
242
+ "single_word": false,
243
+ "special": false
244
+ },
245
+ "50286": {
246
+ "content": " ",
247
+ "lstrip": false,
248
+ "normalized": true,
249
+ "rstrip": false,
250
+ "single_word": false,
251
+ "special": false
252
+ },
253
+ "50287": {
254
+ "content": "\t\t\t\t\t\t\t\t\t",
255
+ "lstrip": false,
256
+ "normalized": true,
257
+ "rstrip": false,
258
+ "single_word": false,
259
+ "special": false
260
+ },
261
+ "50288": {
262
+ "content": "\t\t\t\t\t\t\t\t",
263
+ "lstrip": false,
264
+ "normalized": true,
265
+ "rstrip": false,
266
+ "single_word": false,
267
+ "special": false
268
+ },
269
+ "50289": {
270
+ "content": "\t\t\t\t\t\t\t",
271
+ "lstrip": false,
272
+ "normalized": true,
273
+ "rstrip": false,
274
+ "single_word": false,
275
+ "special": false
276
+ },
277
+ "50290": {
278
+ "content": "\t\t\t\t\t\t",
279
+ "lstrip": false,
280
+ "normalized": true,
281
+ "rstrip": false,
282
+ "single_word": false,
283
+ "special": false
284
+ },
285
+ "50291": {
286
+ "content": "\t\t\t\t\t",
287
+ "lstrip": false,
288
+ "normalized": true,
289
+ "rstrip": false,
290
+ "single_word": false,
291
+ "special": false
292
+ },
293
+ "50292": {
294
+ "content": "\t\t\t\t",
295
+ "lstrip": false,
296
+ "normalized": true,
297
+ "rstrip": false,
298
+ "single_word": false,
299
+ "special": false
300
+ },
301
+ "50293": {
302
+ "content": "\t\t\t",
303
+ "lstrip": false,
304
+ "normalized": true,
305
+ "rstrip": false,
306
+ "single_word": false,
307
+ "special": false
308
+ },
309
+ "50294": {
310
+ "content": "\t\t",
311
+ "lstrip": false,
312
+ "normalized": true,
313
+ "rstrip": false,
314
+ "single_word": false,
315
+ "special": false
316
+ }
317
+ },
318
+ "bos_token": "<|endoftext|>",
319
+ "clean_up_tokenization_spaces": true,
320
+ "eos_token": "<|endoftext|>",
321
+ "errors": "replace",
322
+ "model_max_length": 3072,
323
+ "pad_token": "<|endoftext|>",
324
+ "padding_side": "right",
325
+ "return_token_type_ids": false,
326
+ "tokenizer_class": "CodeGenTokenizer",
327
+ "unk_token": "<|endoftext|>"
328
+ }
checkpoint-21520/trainer_state.json ADDED
The diff for this file is too large to render. See raw diff
 
checkpoint-21520/training_args.bin ADDED
@@ -0,0 +1,3 @@
 
 
 
 
1
+ version https://git-lfs.github.com/spec/v1
2
+ oid sha256:67d57e8cf61441f800130a44ea12b10ca884319fb350e94a1772a4f3bbf06fa2
3
+ size 7227
checkpoint-21520/vocab.json ADDED
The diff for this file is too large to render. See raw diff
 
checkpoint-21520/zero_to_fp32.py ADDED
@@ -0,0 +1,604 @@
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
1
+ #!/usr/bin/env python
2
+
3
+ # Copyright (c) Microsoft Corporation.
4
+ # SPDX-License-Identifier: Apache-2.0
5
+
6
+ # DeepSpeed Team
7
+
8
+ # This script extracts fp32 consolidated weights from a zero 1, 2 and 3 DeepSpeed checkpoints. It gets
9
+ # copied into the top level checkpoint dir, so the user can easily do the conversion at any point in
10
+ # the future. Once extracted, the weights don't require DeepSpeed and can be used in any
11
+ # application.
12
+ #
13
+ # example: python zero_to_fp32.py . pytorch_model.bin
14
+
15
+ import argparse
16
+ import torch
17
+ import glob
18
+ import math
19
+ import os
20
+ import re
21
+ from collections import OrderedDict
22
+ from dataclasses import dataclass
23
+
24
+ # while this script doesn't use deepspeed to recover data, since the checkpoints are pickled with
25
+ # DeepSpeed data structures it has to be available in the current python environment.
26
+ from deepspeed.utils import logger
27
+ from deepspeed.checkpoint.constants import (DS_VERSION, OPTIMIZER_STATE_DICT, SINGLE_PARTITION_OF_FP32_GROUPS,
28
+ FP32_FLAT_GROUPS, ZERO_STAGE, PARTITION_COUNT, PARAM_SHAPES, BUFFER_NAMES,
29
+ FROZEN_PARAM_SHAPES, FROZEN_PARAM_FRAGMENTS)
30
+
31
+
32
+ @dataclass
33
+ class zero_model_state:
34
+ buffers: dict()
35
+ param_shapes: dict()
36
+ shared_params: list
37
+ ds_version: int
38
+ frozen_param_shapes: dict()
39
+ frozen_param_fragments: dict()
40
+
41
+
42
+ debug = 0
43
+
44
+ # load to cpu
45
+ device = torch.device('cpu')
46
+
47
+
48
+ def atoi(text):
49
+ return int(text) if text.isdigit() else text
50
+
51
+
52
+ def natural_keys(text):
53
+ '''
54
+ alist.sort(key=natural_keys) sorts in human order
55
+ http://nedbatchelder.com/blog/200712/human_sorting.html
56
+ (See Toothy's implementation in the comments)
57
+ '''
58
+ return [atoi(c) for c in re.split(r'(\d+)', text)]
59
+
60
+
61
+ def get_model_state_file(checkpoint_dir, zero_stage):
62
+ if not os.path.isdir(checkpoint_dir):
63
+ raise FileNotFoundError(f"Directory '{checkpoint_dir}' doesn't exist")
64
+
65
+ # there should be only one file
66
+ if zero_stage <= 2:
67
+ file = os.path.join(checkpoint_dir, "mp_rank_00_model_states.pt")
68
+ elif zero_stage == 3:
69
+ file = os.path.join(checkpoint_dir, "zero_pp_rank_0_mp_rank_00_model_states.pt")
70
+
71
+ if not os.path.exists(file):
72
+ raise FileNotFoundError(f"can't find model states file at '{file}'")
73
+
74
+ return file
75
+
76
+
77
+ def get_checkpoint_files(checkpoint_dir, glob_pattern):
78
+ # XXX: need to test that this simple glob rule works for multi-node setup too
79
+ ckpt_files = sorted(glob.glob(os.path.join(checkpoint_dir, glob_pattern)), key=natural_keys)
80
+
81
+ if len(ckpt_files) == 0:
82
+ raise FileNotFoundError(f"can't find {glob_pattern} files in directory '{checkpoint_dir}'")
83
+
84
+ return ckpt_files
85
+
86
+
87
+ def get_optim_files(checkpoint_dir):
88
+ return get_checkpoint_files(checkpoint_dir, "*_optim_states.pt")
89
+
90
+
91
+ def get_model_state_files(checkpoint_dir):
92
+ return get_checkpoint_files(checkpoint_dir, "*_model_states.pt")
93
+
94
+
95
+ def parse_model_states(files):
96
+ zero_model_states = []
97
+ for file in files:
98
+ state_dict = torch.load(file, map_location=device)
99
+
100
+ if BUFFER_NAMES not in state_dict:
101
+ raise ValueError(f"{file} is not a model state checkpoint")
102
+ buffer_names = state_dict[BUFFER_NAMES]
103
+ if debug:
104
+ print("Found buffers:", buffer_names)
105
+
106
+ # recover just the buffers while restoring them to fp32 if they were saved in fp16
107
+ buffers = {k: v.float() for k, v in state_dict["module"].items() if k in buffer_names}
108
+ param_shapes = state_dict[PARAM_SHAPES]
109
+
110
+ # collect parameters that are included in param_shapes
111
+ param_names = []
112
+ for s in param_shapes:
113
+ for name in s.keys():
114
+ param_names.append(name)
115
+
116
+ # update with frozen parameters
117
+ frozen_param_shapes = state_dict.get(FROZEN_PARAM_SHAPES, None)
118
+ if frozen_param_shapes is not None:
119
+ if debug:
120
+ print(f"Found frozen_param_shapes: {frozen_param_shapes}")
121
+ param_names += list(frozen_param_shapes.keys())
122
+
123
+ # handle shared params
124
+ shared_params = [[k, v] for k, v in state_dict["shared_params"].items()]
125
+
126
+ ds_version = state_dict.get(DS_VERSION, None)
127
+
128
+ frozen_param_fragments = state_dict.get(FROZEN_PARAM_FRAGMENTS, None)
129
+
130
+ z_model_state = zero_model_state(buffers=buffers,
131
+ param_shapes=param_shapes,
132
+ shared_params=shared_params,
133
+ ds_version=ds_version,
134
+ frozen_param_shapes=frozen_param_shapes,
135
+ frozen_param_fragments=frozen_param_fragments)
136
+ zero_model_states.append(z_model_state)
137
+
138
+ return zero_model_states
139
+
140
+
141
+ def parse_optim_states(files, ds_checkpoint_dir):
142
+
143
+ total_files = len(files)
144
+ state_dicts = []
145
+ for f in files:
146
+ state_dict = torch.load(f, map_location=device)
147
+ # immediately discard the potentially huge 2 optimizer states as we only care for fp32 master weights
148
+ # and also handle the case where it was already removed by another helper script
149
+ state_dict["optimizer_state_dict"].pop("optimizer_state_dict", None)
150
+ state_dicts.append(state_dict)
151
+
152
+ if not ZERO_STAGE in state_dicts[0][OPTIMIZER_STATE_DICT]:
153
+ raise ValueError(f"{files[0]} is not a zero checkpoint")
154
+ zero_stage = state_dicts[0][OPTIMIZER_STATE_DICT][ZERO_STAGE]
155
+ world_size = state_dicts[0][OPTIMIZER_STATE_DICT][PARTITION_COUNT]
156
+
157
+ # For ZeRO-2 each param group can have different partition_count as data parallelism for expert
158
+ # parameters can be different from data parallelism for non-expert parameters. So we can just
159
+ # use the max of the partition_count to get the dp world_size.
160
+
161
+ if type(world_size) is list:
162
+ world_size = max(world_size)
163
+
164
+ if world_size != total_files:
165
+ raise ValueError(
166
+ f"Expected {world_size} of '*_optim_states.pt' under '{ds_checkpoint_dir}' but found {total_files} files. "
167
+ "Possibly due to an overwrite of an old checkpoint, or a checkpoint didn't get saved by one or more processes."
168
+ )
169
+
170
+ # the groups are named differently in each stage
171
+ if zero_stage <= 2:
172
+ fp32_groups_key = SINGLE_PARTITION_OF_FP32_GROUPS
173
+ elif zero_stage == 3:
174
+ fp32_groups_key = FP32_FLAT_GROUPS
175
+ else:
176
+ raise ValueError(f"unknown zero stage {zero_stage}")
177
+
178
+ if zero_stage <= 2:
179
+ fp32_flat_groups = [state_dicts[i][OPTIMIZER_STATE_DICT][fp32_groups_key] for i in range(len(state_dicts))]
180
+ elif zero_stage == 3:
181
+ # if there is more than one param group, there will be multiple flattened tensors - one
182
+ # flattened tensor per group - for simplicity merge them into a single tensor
183
+ #
184
+ # XXX: could make the script more memory efficient for when there are multiple groups - it
185
+ # will require matching the sub-lists of param_shapes for each param group flattened tensor
186
+
187
+ fp32_flat_groups = [
188
+ torch.cat(state_dicts[i][OPTIMIZER_STATE_DICT][fp32_groups_key], 0) for i in range(len(state_dicts))
189
+ ]
190
+
191
+ return zero_stage, world_size, fp32_flat_groups
192
+
193
+
194
+ def _get_fp32_state_dict_from_zero_checkpoint(ds_checkpoint_dir, exclude_frozen_parameters):
195
+ """
196
+ Returns fp32 state_dict reconstructed from ds checkpoint
197
+
198
+ Args:
199
+ - ``ds_checkpoint_dir``: path to the deepspeed checkpoint folder (where the optimizer files are)
200
+
201
+ """
202
+ print(f"Processing zero checkpoint '{ds_checkpoint_dir}'")
203
+
204
+ optim_files = get_optim_files(ds_checkpoint_dir)
205
+ zero_stage, world_size, fp32_flat_groups = parse_optim_states(optim_files, ds_checkpoint_dir)
206
+ print(f"Detected checkpoint of type zero stage {zero_stage}, world_size: {world_size}")
207
+
208
+ model_files = get_model_state_files(ds_checkpoint_dir)
209
+
210
+ zero_model_states = parse_model_states(model_files)
211
+ print(f'Parsing checkpoint created by deepspeed=={zero_model_states[0].ds_version}')
212
+
213
+ if zero_stage <= 2:
214
+ return _get_fp32_state_dict_from_zero2_checkpoint(world_size, fp32_flat_groups, zero_model_states,
215
+ exclude_frozen_parameters)
216
+ elif zero_stage == 3:
217
+ return _get_fp32_state_dict_from_zero3_checkpoint(world_size, fp32_flat_groups, zero_model_states,
218
+ exclude_frozen_parameters)
219
+
220
+
221
+ def _zero2_merge_frozen_params(state_dict, zero_model_states):
222
+ if zero_model_states[0].frozen_param_shapes is None or len(zero_model_states[0].frozen_param_shapes) == 0:
223
+ return
224
+
225
+ frozen_param_shapes = zero_model_states[0].frozen_param_shapes
226
+ frozen_param_fragments = zero_model_states[0].frozen_param_fragments
227
+
228
+ if debug:
229
+ num_elem = sum(s.numel() for s in frozen_param_shapes.values())
230
+ print(f'rank 0: {FROZEN_PARAM_SHAPES}.numel = {num_elem}')
231
+
232
+ wanted_params = len(frozen_param_shapes)
233
+ wanted_numel = sum(s.numel() for s in frozen_param_shapes.values())
234
+ avail_numel = sum([p.numel() for p in frozen_param_fragments.values()])
235
+ print(f'Frozen params: Have {avail_numel} numels to process.')
236
+ print(f'Frozen params: Need {wanted_numel} numels in {wanted_params} params')
237
+
238
+ total_params = 0
239
+ total_numel = 0
240
+ for name, shape in frozen_param_shapes.items():
241
+ total_params += 1
242
+ unpartitioned_numel = shape.numel()
243
+ total_numel += unpartitioned_numel
244
+
245
+ state_dict[name] = frozen_param_fragments[name]
246
+
247
+ if debug:
248
+ print(f"{name} full shape: {shape} unpartitioned numel {unpartitioned_numel} ")
249
+
250
+ print(f"Reconstructed Frozen fp32 state dict with {total_params} params {total_numel} elements")
251
+
252
+
253
+ def _has_callable(obj, fn):
254
+ attr = getattr(obj, fn, None)
255
+ return callable(attr)
256
+
257
+
258
+ def _zero2_merge_trainable_params(state_dict, world_size, fp32_flat_groups, zero_model_states):
259
+ param_shapes = zero_model_states[0].param_shapes
260
+
261
+ # Reconstruction protocol:
262
+ #
263
+ # XXX: document this
264
+
265
+ if debug:
266
+ for i in range(world_size):
267
+ for j in range(len(fp32_flat_groups[0])):
268
+ print(f"{FP32_FLAT_GROUPS}[{i}][{j}].shape={fp32_flat_groups[i][j].shape}")
269
+
270
+ # XXX: memory usage doubles here (zero2)
271
+ num_param_groups = len(fp32_flat_groups[0])
272
+ merged_single_partition_of_fp32_groups = []
273
+ for i in range(num_param_groups):
274
+ merged_partitions = [sd[i] for sd in fp32_flat_groups]
275
+ full_single_fp32_vector = torch.cat(merged_partitions, 0)
276
+ merged_single_partition_of_fp32_groups.append(full_single_fp32_vector)
277
+ avail_numel = sum(
278
+ [full_single_fp32_vector.numel() for full_single_fp32_vector in merged_single_partition_of_fp32_groups])
279
+
280
+ if debug:
281
+ wanted_params = sum([len(shapes) for shapes in param_shapes])
282
+ wanted_numel = sum([sum(shape.numel() for shape in shapes.values()) for shapes in param_shapes])
283
+ # not asserting if there is a mismatch due to possible padding
284
+ print(f"Have {avail_numel} numels to process.")
285
+ print(f"Need {wanted_numel} numels in {wanted_params} params.")
286
+
287
+ # params
288
+ # XXX: for huge models that can't fit into the host's RAM we will have to recode this to support
289
+ # out-of-core computing solution
290
+ total_numel = 0
291
+ total_params = 0
292
+ for shapes, full_single_fp32_vector in zip(param_shapes, merged_single_partition_of_fp32_groups):
293
+ offset = 0
294
+ avail_numel = full_single_fp32_vector.numel()
295
+ for name, shape in shapes.items():
296
+
297
+ unpartitioned_numel = shape.numel() if _has_callable(shape, 'numel') else math.prod(shape)
298
+ total_numel += unpartitioned_numel
299
+ total_params += 1
300
+
301
+ if debug:
302
+ print(f"{name} full shape: {shape} unpartitioned numel {unpartitioned_numel} ")
303
+ state_dict[name] = full_single_fp32_vector.narrow(0, offset, unpartitioned_numel).view(shape)
304
+ offset += unpartitioned_numel
305
+
306
+ # Z2 started to align to 2*world_size to improve nccl performance. Therefore both offset and
307
+ # avail_numel can differ by anywhere between 0..2*world_size. Due to two unrelated complex
308
+ # paddings performed in the code it's almost impossible to predict the exact numbers w/o the
309
+ # live optimizer object, so we are checking that the numbers are within the right range
310
+ align_to = 2 * world_size
311
+
312
+ def zero2_align(x):
313
+ return align_to * math.ceil(x / align_to)
314
+
315
+ if debug:
316
+ print(f"original offset={offset}, avail_numel={avail_numel}")
317
+
318
+ offset = zero2_align(offset)
319
+ avail_numel = zero2_align(avail_numel)
320
+
321
+ if debug:
322
+ print(f"aligned offset={offset}, avail_numel={avail_numel}")
323
+
324
+ # Sanity check
325
+ if offset != avail_numel:
326
+ raise ValueError(f"consumed {offset} numels out of {avail_numel} - something is wrong")
327
+
328
+ print(f"Reconstructed fp32 state dict with {total_params} params {total_numel} elements")
329
+
330
+
331
+ def _get_fp32_state_dict_from_zero2_checkpoint(world_size, fp32_flat_groups, zero_model_states,
332
+ exclude_frozen_parameters):
333
+ state_dict = OrderedDict()
334
+
335
+ # buffers
336
+ buffers = zero_model_states[0].buffers
337
+ state_dict.update(buffers)
338
+ if debug:
339
+ print(f"added {len(buffers)} buffers")
340
+
341
+ if not exclude_frozen_parameters:
342
+ _zero2_merge_frozen_params(state_dict, zero_model_states)
343
+
344
+ _zero2_merge_trainable_params(state_dict, world_size, fp32_flat_groups, zero_model_states)
345
+
346
+ # recover shared parameters
347
+ for pair in zero_model_states[0].shared_params:
348
+ if pair[1] in state_dict:
349
+ state_dict[pair[0]] = state_dict[pair[1]]
350
+
351
+ return state_dict
352
+
353
+
354
+ def zero3_partitioned_param_info(unpartitioned_numel, world_size):
355
+ remainder = unpartitioned_numel % world_size
356
+ padding_numel = (world_size - remainder) if remainder else 0
357
+ partitioned_numel = math.ceil(unpartitioned_numel / world_size)
358
+ return partitioned_numel, padding_numel
359
+
360
+
361
+ def _zero3_merge_frozen_params(state_dict, world_size, zero_model_states):
362
+ if zero_model_states[0].frozen_param_shapes is None or len(zero_model_states[0].frozen_param_shapes) == 0:
363
+ return
364
+
365
+ if debug:
366
+ for i in range(world_size):
367
+ num_elem = sum(s.numel() for s in zero_model_states[i].frozen_param_fragments.values())
368
+ print(f'rank {i}: {FROZEN_PARAM_SHAPES}.numel = {num_elem}')
369
+
370
+ frozen_param_shapes = zero_model_states[0].frozen_param_shapes
371
+ wanted_params = len(frozen_param_shapes)
372
+ wanted_numel = sum(s.numel() for s in frozen_param_shapes.values())
373
+ avail_numel = sum([p.numel() for p in zero_model_states[0].frozen_param_fragments.values()]) * world_size
374
+ print(f'Frozen params: Have {avail_numel} numels to process.')
375
+ print(f'Frozen params: Need {wanted_numel} numels in {wanted_params} params')
376
+
377
+ total_params = 0
378
+ total_numel = 0
379
+ for name, shape in zero_model_states[0].frozen_param_shapes.items():
380
+ total_params += 1
381
+ unpartitioned_numel = shape.numel()
382
+ total_numel += unpartitioned_numel
383
+
384
+ param_frags = tuple(model_state.frozen_param_fragments[name] for model_state in zero_model_states)
385
+ state_dict[name] = torch.cat(param_frags, 0).narrow(0, 0, unpartitioned_numel).view(shape)
386
+
387
+ partitioned_numel, partitioned_padding_numel = zero3_partitioned_param_info(unpartitioned_numel, world_size)
388
+
389
+ if debug:
390
+ print(
391
+ f"Frozen params: {total_params} {name} full shape: {shape} partition0 numel={partitioned_numel} partitioned_padding_numel={partitioned_padding_numel}"
392
+ )
393
+
394
+ print(f"Reconstructed Frozen fp32 state dict with {total_params} params {total_numel} elements")
395
+
396
+
397
+ def _zero3_merge_trainable_params(state_dict, world_size, fp32_flat_groups, zero_model_states):
398
+ param_shapes = zero_model_states[0].param_shapes
399
+ avail_numel = fp32_flat_groups[0].numel() * world_size
400
+ # Reconstruction protocol: For zero3 we need to zip the partitions together at boundary of each
401
+ # param, re-consolidating each param, while dealing with padding if any
402
+
403
+ # merge list of dicts, preserving order
404
+ param_shapes = {k: v for d in param_shapes for k, v in d.items()}
405
+
406
+ if debug:
407
+ for i in range(world_size):
408
+ print(f"{FP32_FLAT_GROUPS}[{i}].shape={fp32_flat_groups[i].shape}")
409
+
410
+ wanted_params = len(param_shapes)
411
+ wanted_numel = sum(shape.numel() for shape in param_shapes.values())
412
+ # not asserting if there is a mismatch due to possible padding
413
+ avail_numel = fp32_flat_groups[0].numel() * world_size
414
+ print(f"Trainable params: Have {avail_numel} numels to process.")
415
+ print(f"Trainable params: Need {wanted_numel} numels in {wanted_params} params.")
416
+
417
+ # params
418
+ # XXX: for huge models that can't fit into the host's RAM we will have to recode this to support
419
+ # out-of-core computing solution
420
+ offset = 0
421
+ total_numel = 0
422
+ total_params = 0
423
+ for name, shape in param_shapes.items():
424
+
425
+ unpartitioned_numel = shape.numel()
426
+ total_numel += unpartitioned_numel
427
+ total_params += 1
428
+
429
+ partitioned_numel, partitioned_padding_numel = zero3_partitioned_param_info(unpartitioned_numel, world_size)
430
+
431
+ if debug:
432
+ print(
433
+ f"Trainable params: {total_params} {name} full shape: {shape} partition0 numel={partitioned_numel} partitioned_padding_numel={partitioned_padding_numel}"
434
+ )
435
+
436
+ # XXX: memory usage doubles here
437
+ state_dict[name] = torch.cat(
438
+ tuple(fp32_flat_groups[i].narrow(0, offset, partitioned_numel) for i in range(world_size)),
439
+ 0).narrow(0, 0, unpartitioned_numel).view(shape)
440
+ offset += partitioned_numel
441
+
442
+ offset *= world_size
443
+
444
+ # Sanity check
445
+ if offset != avail_numel:
446
+ raise ValueError(f"consumed {offset} numels out of {avail_numel} - something is wrong")
447
+
448
+ print(f"Reconstructed Trainable fp32 state dict with {total_params} params {total_numel} elements")
449
+
450
+
451
+ def _get_fp32_state_dict_from_zero3_checkpoint(world_size, fp32_flat_groups, zero_model_states,
452
+ exclude_frozen_parameters):
453
+ state_dict = OrderedDict()
454
+
455
+ # buffers
456
+ buffers = zero_model_states[0].buffers
457
+ state_dict.update(buffers)
458
+ if debug:
459
+ print(f"added {len(buffers)} buffers")
460
+
461
+ if not exclude_frozen_parameters:
462
+ _zero3_merge_frozen_params(state_dict, world_size, zero_model_states)
463
+
464
+ _zero3_merge_trainable_params(state_dict, world_size, fp32_flat_groups, zero_model_states)
465
+
466
+ # recover shared parameters
467
+ for pair in zero_model_states[0].shared_params:
468
+ if pair[1] in state_dict:
469
+ state_dict[pair[0]] = state_dict[pair[1]]
470
+
471
+ return state_dict
472
+
473
+
474
+ def get_fp32_state_dict_from_zero_checkpoint(checkpoint_dir, tag=None, exclude_frozen_parameters=False):
475
+ """
476
+ Convert ZeRO 2 or 3 checkpoint into a single fp32 consolidated state_dict that can be loaded with
477
+ ``load_state_dict()`` and used for training without DeepSpeed or shared with others, for example
478
+ via a model hub.
479
+
480
+ Args:
481
+ - ``checkpoint_dir``: path to the desired checkpoint folder
482
+ - ``tag``: checkpoint tag used as a unique identifier for checkpoint. If not provided will attempt to load tag in 'latest' file. e.g., ``global_step14``
483
+ - ``exclude_frozen_parameters``: exclude frozen parameters
484
+
485
+ Returns:
486
+ - pytorch ``state_dict``
487
+
488
+ Note: this approach may not work if your application doesn't have sufficient free CPU memory and
489
+ you may need to use the offline approach using the ``zero_to_fp32.py`` script that is saved with
490
+ the checkpoint.
491
+
492
+ A typical usage might be ::
493
+
494
+ from deepspeed.utils.zero_to_fp32 import get_fp32_state_dict_from_zero_checkpoint
495
+ # do the training and checkpoint saving
496
+ state_dict = get_fp32_state_dict_from_zero_checkpoint(checkpoint_dir) # already on cpu
497
+ model = model.cpu() # move to cpu
498
+ model.load_state_dict(state_dict)
499
+ # submit to model hub or save the model to share with others
500
+
501
+ In this example the ``model`` will no longer be usable in the deepspeed context of the same
502
+ application. i.e. you will need to re-initialize the deepspeed engine, since
503
+ ``model.load_state_dict(state_dict)`` will remove all the deepspeed magic from it.
504
+
505
+ If you want it all done for you, use ``load_state_dict_from_zero_checkpoint`` instead.
506
+
507
+ """
508
+ if tag is None:
509
+ latest_path = os.path.join(checkpoint_dir, 'latest')
510
+ if os.path.isfile(latest_path):
511
+ with open(latest_path, 'r') as fd:
512
+ tag = fd.read().strip()
513
+ else:
514
+ raise ValueError(f"Unable to find 'latest' file at {latest_path}")
515
+
516
+ ds_checkpoint_dir = os.path.join(checkpoint_dir, tag)
517
+
518
+ if not os.path.isdir(ds_checkpoint_dir):
519
+ raise FileNotFoundError(f"Directory '{ds_checkpoint_dir}' doesn't exist")
520
+
521
+ return _get_fp32_state_dict_from_zero_checkpoint(ds_checkpoint_dir, exclude_frozen_parameters)
522
+
523
+
524
+ def convert_zero_checkpoint_to_fp32_state_dict(checkpoint_dir, output_file, tag=None, exclude_frozen_parameters=False):
525
+ """
526
+ Convert ZeRO 2 or 3 checkpoint into a single fp32 consolidated ``state_dict`` file that can be
527
+ loaded with ``torch.load(file)`` + ``load_state_dict()`` and used for training without DeepSpeed.
528
+
529
+ Args:
530
+ - ``checkpoint_dir``: path to the desired checkpoint folder. (one that contains the tag-folder, like ``global_step14``)
531
+ - ``output_file``: path to the pytorch fp32 state_dict output file (e.g. path/pytorch_model.bin)
532
+ - ``tag``: checkpoint tag used as a unique identifier for checkpoint. If not provided will attempt to load tag in the file named ``latest`` in the checkpoint folder, e.g., ``global_step14``
533
+ - ``exclude_frozen_parameters``: exclude frozen parameters
534
+ """
535
+
536
+ state_dict = get_fp32_state_dict_from_zero_checkpoint(checkpoint_dir, tag, exclude_frozen_parameters)
537
+ print(f"Saving fp32 state dict to {output_file}")
538
+ torch.save(state_dict, output_file)
539
+
540
+
541
+ def load_state_dict_from_zero_checkpoint(model, checkpoint_dir, tag=None):
542
+ """
543
+ 1. Put the provided model to cpu
544
+ 2. Convert ZeRO 2 or 3 checkpoint into a single fp32 consolidated ``state_dict``
545
+ 3. Load it into the provided model
546
+
547
+ Args:
548
+ - ``model``: the model object to update
549
+ - ``checkpoint_dir``: path to the desired checkpoint folder. (one that contains the tag-folder, like ``global_step14``)
550
+ - ``tag``: checkpoint tag used as a unique identifier for checkpoint. If not provided will attempt to load tag in the file named ``latest`` in the checkpoint folder, e.g., ``global_step14``
551
+
552
+ Returns:
553
+ - ``model`: modified model
554
+
555
+ Make sure you have plenty of CPU memory available before you call this function. If you don't
556
+ have enough use the ``zero_to_fp32.py`` utility to do the conversion. You will find it
557
+ conveniently placed for you in the checkpoint folder.
558
+
559
+ A typical usage might be ::
560
+
561
+ from deepspeed.utils.zero_to_fp32 import load_state_dict_from_zero_checkpoint
562
+ model = load_state_dict_from_zero_checkpoint(trainer.model, checkpoint_dir)
563
+ # submit to model hub or save the model to share with others
564
+
565
+ Note, that once this was run, the ``model`` will no longer be usable in the deepspeed context
566
+ of the same application. i.e. you will need to re-initialize the deepspeed engine, since
567
+ ``model.load_state_dict(state_dict)`` will remove all the deepspeed magic from it.
568
+
569
+ """
570
+ logger.info(f"Extracting fp32 weights")
571
+ state_dict = get_fp32_state_dict_from_zero_checkpoint(checkpoint_dir, tag)
572
+
573
+ logger.info(f"Overwriting model with fp32 weights")
574
+ model = model.cpu()
575
+ model.load_state_dict(state_dict, strict=False)
576
+
577
+ return model
578
+
579
+
580
+ if __name__ == "__main__":
581
+
582
+ parser = argparse.ArgumentParser()
583
+ parser.add_argument("checkpoint_dir",
584
+ type=str,
585
+ help="path to the desired checkpoint folder, e.g., path/checkpoint-12")
586
+ parser.add_argument(
587
+ "output_file",
588
+ type=str,
589
+ help="path to the pytorch fp32 state_dict output file (e.g. path/checkpoint-12/pytorch_model.bin)")
590
+ parser.add_argument("-t",
591
+ "--tag",
592
+ type=str,
593
+ default=None,
594
+ help="checkpoint tag used as a unique identifier for checkpoint. e.g., global_step1")
595
+ parser.add_argument("--exclude_frozen_parameters", action='store_true', help="exclude frozen parameters")
596
+ parser.add_argument("-d", "--debug", action='store_true', help="enable debug")
597
+ args = parser.parse_args()
598
+
599
+ debug = args.debug
600
+
601
+ convert_zero_checkpoint_to_fp32_state_dict(args.checkpoint_dir,
602
+ args.output_file,
603
+ tag=args.tag,
604
+ exclude_frozen_parameters=args.exclude_frozen_parameters)
config.json ADDED
@@ -0,0 +1,76 @@
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
1
+ {
2
+ "_name_or_path": "/home/work/workspace/checkpoints/tinyllava-phi2-siglip-3.1B",
3
+ "architectures": [
4
+ "TinyLlavaForConditionalGeneration"
5
+ ],
6
+ "auto_map": {
7
+ "AutoConfig": "configuration.TinyLlavaConfig",
8
+ "AutoModelForCausalLM": "modeling_tinyllava_phi.TinyLlavaForConditionalGeneration"
9
+ },
10
+ "cache_dir": null,
11
+ "connector_type": "mlp2x_gelu",
12
+ "hidden_size": 2560,
13
+ "ignore_index": -100,
14
+ "image_aspect_ratio": "square",
15
+ "image_token_index": -200,
16
+ "llm_model_name_or_path": "/home/work/workspace/checkpoints/phi-2",
17
+ "model_type": "tinyllava",
18
+ "num_queries": 128,
19
+ "num_resampler_layers": 3,
20
+ "pad_token": "<|endoftext|>",
21
+ "resampler_hidden_size": 768,
22
+ "text_config": {
23
+ "_name_or_path": "/home/work/workspace/checkpoints/phi-2",
24
+ "architectures": [
25
+ "PhiForCausalLM"
26
+ ],
27
+ "auto_map": {
28
+ "AutoConfig": "microsoft/phi-2--configuration_phi.PhiConfig",
29
+ "AutoModelForCausalLM": "microsoft/phi-2--modeling_phi.PhiForCausalLM"
30
+ },
31
+ "bos_token_id": 50256,
32
+ "embd_pdrop": 0.0,
33
+ "eos_token_id": 50256,
34
+ "hidden_act": "gelu_new",
35
+ "hidden_size": 2560,
36
+ "intermediate_size": 10240,
37
+ "layer_norm_eps": 1e-05,
38
+ "model_type": "phi",
39
+ "num_hidden_layers": 32,
40
+ "partial_rotary_factor": 0.4,
41
+ "qk_layernorm": false,
42
+ "resid_pdrop": 0.1,
43
+ "torch_dtype": "float16",
44
+ "vocab_size": 51200
45
+ },
46
+ "tokenizer_model_max_length": 3072,
47
+ "tokenizer_name_or_path": "/home/work/workspace/checkpoints/phi-2",
48
+ "tokenizer_padding_side": "right",
49
+ "tokenizer_use_fast": false,
50
+ "torch_dtype": "float16",
51
+ "transformers_version": "4.44.2",
52
+ "tune_type_connector": "full",
53
+ "tune_type_llm": "full",
54
+ "tune_type_vision_tower": "full",
55
+ "tune_vision_tower_from_layer": 0,
56
+ "use_cache": true,
57
+ "vision_config": {
58
+ "hidden_act": "gelu_pytorch_tanh",
59
+ "hidden_size": 1152,
60
+ "image_size": 384,
61
+ "intermediate_size": 4304,
62
+ "layer_norm_eps": 1e-06,
63
+ "model_name_or_path": "/home/work/workspace/checkpoints/siglip",
64
+ "model_name_or_path2": "",
65
+ "model_type": "siglip_vision_model",
66
+ "num_attention_heads": 16,
67
+ "num_hidden_layers": 27,
68
+ "patch_size": 14
69
+ },
70
+ "vision_feature_layer": -2,
71
+ "vision_feature_select_strategy": "patch",
72
+ "vision_hidden_size": 1152,
73
+ "vision_model_name_or_path": "/home/work/workspace/checkpoints/siglip",
74
+ "vision_model_name_or_path2": "",
75
+ "vocab_size": 51200
76
+ }
configuration.py ADDED
@@ -0,0 +1,116 @@
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
1
+ from transformers import PretrainedConfig
2
+ from transformers import CONFIG_MAPPING
3
+ from transformers import AutoConfig
4
+
5
+ IGNORE_INDEX = -100
6
+ IMAGE_TOKEN_INDEX = -200
7
+ DEFAULT_IMAGE_TOKEN = "<image>"
8
+
9
+
10
+ class TinyLlavaConfig(PretrainedConfig):
11
+
12
+ model_type = "tinyllava"
13
+ def __init__(
14
+ self,
15
+ llm_model_name_or_path = '',
16
+ tokenizer_name_or_path = None,
17
+ vision_model_name_or_path = '',
18
+ vision_model_name_or_path2 = '',
19
+ connector_type = None,
20
+ text_config=None,
21
+ hidden_size=2048,
22
+ vocab_size=32000,
23
+ ignore_index=-100,
24
+ image_token_index=32000,
25
+ pad_token = None,
26
+ pad_token_id = None,
27
+ tokenizer_padding_side = 'right',
28
+ tokenizer_model_max_length = 2048,
29
+ vision_config = None,
30
+ vision_hidden_size = None,
31
+ vision_feature_layer = -2,
32
+ vision_feature_select_strategy = 'patch',
33
+ image_aspect_ratio = 'square',
34
+ resampler_hidden_size = None,
35
+ num_queries = None,
36
+ num_resampler_layers = None,
37
+ use_cache = False,
38
+ cache_dir = None,
39
+ tokenizer_use_fast = False,
40
+ tune_type_llm = 'frozen',
41
+ tune_type_connector = 'frozen',
42
+ tune_type_vision_tower = 'frozen',
43
+ tune_vision_tower_from_layer = -1,
44
+
45
+ **kwargs
46
+
47
+ ):
48
+ self.llm_model_name_or_path = llm_model_name_or_path
49
+ self.tokenizer_name_or_path = tokenizer_name_or_path or self.llm_model_name_or_path
50
+ self.vision_model_name_or_path = vision_model_name_or_path
51
+ self.vision_model_name_or_path2 = vision_model_name_or_path2
52
+ self.connector_type = connector_type
53
+ self.tune_type_llm = tune_type_llm
54
+ self.tune_type_connector = tune_type_connector
55
+ self.tune_type_vision_tower = tune_type_vision_tower
56
+ self.tune_vision_tower_from_layer = tune_vision_tower_from_layer
57
+
58
+ self.ignore_index = IGNORE_INDEX
59
+ self.image_token_index = IMAGE_TOKEN_INDEX
60
+ self.pad_token = pad_token
61
+ self.pad_token_id = pad_token_id
62
+ self.tokenizer_padding_side = tokenizer_padding_side
63
+ self.tokenizer_model_max_length = tokenizer_model_max_length
64
+ self.vision_feature_layer = vision_feature_layer
65
+ self.vision_feature_select_strategy = vision_feature_select_strategy
66
+ self.image_aspect_ratio = image_aspect_ratio
67
+ self.resampler_hidden_size = resampler_hidden_size
68
+ self.num_queries = num_queries
69
+ self.num_resampler_layers = num_resampler_layers
70
+ self.use_cache = use_cache
71
+ self.cache_dir = cache_dir
72
+ self.tokenizer_use_fast = tokenizer_use_fast
73
+ self._load_text_config(text_config)
74
+ self._load_vision_config(vision_config)
75
+
76
+ super().__init__(**kwargs)
77
+
78
+
79
+ def _load_text_config(self, text_config=None):
80
+ if self.llm_model_name_or_path is None or self.llm_model_name_or_path == '':
81
+ self.text_config = CONFIG_MAPPING['llama']()
82
+
83
+ else:
84
+ self.text_config = AutoConfig.from_pretrained(self.llm_model_name_or_path, trust_remote_code=True)
85
+ if text_config is not None:
86
+ self.text_config = self.text_config.from_dict(text_config)
87
+
88
+ self.hidden_size = getattr(self.text_config, 'hidden_size', getattr(self.text_config, 'model_dim', None))
89
+ self.vocab_size = getattr(self.text_config, 'vocab_size', None)
90
+
91
+
92
+
93
+ def _load_vision_config(self, vision_config=None):
94
+ if self.vision_model_name_or_path is None or self.vision_model_name_or_path == '':
95
+ self.vision_config = CONFIG_MAPPING['clip_vision_model'](
96
+ intermediate_size=4096,
97
+ hidden_size=1024,
98
+ patch_size=14,
99
+ image_size=336,
100
+ num_hidden_layers=24,
101
+ num_attention_heads=16,
102
+ vocab_size=32000,
103
+ projection_dim=768,
104
+ )
105
+
106
+ else:
107
+ self.vision_config = AutoConfig.from_pretrained(self.vision_model_name_or_path.split(':')[-1])
108
+ self.vision_config = getattr(self.vision_config, 'vision_config', self.vision_config)
109
+ if vision_config is not None:
110
+ self.vision_config = self.vision_config.from_dict(vision_config)
111
+
112
+ self.vision_config.model_name_or_path = self.vision_model_name_or_path.split(':')[-1]
113
+ self.vision_config.model_name_or_path2 = self.vision_model_name_or_path2.split(':')[-1]
114
+ self.vision_hidden_size = getattr(self.vision_config, 'hidden_size', None)
115
+
116
+
connector/pytorch_model.bin ADDED
@@ -0,0 +1,3 @@
 
 
 
 
1
+ version https://git-lfs.github.com/spec/v1
2
+ oid sha256:5c393b9c20a8a5e5b88d1b94ec00d50e3831e4f4a350fd9c31aa8ba34085edbc
3
+ size 19017283
language_model/config.json ADDED
@@ -0,0 +1,34 @@
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
1
+ {
2
+ "_name_or_path": "/home/work/workspace/checkpoints/phi-2",
3
+ "architectures": [
4
+ "PhiForCausalLM"
5
+ ],
6
+ "attention_dropout": 0.0,
7
+ "auto_map": {
8
+ "AutoConfig": "microsoft/phi-2--configuration_phi.PhiConfig",
9
+ "AutoModelForCausalLM": "microsoft/phi-2--modeling_phi.PhiForCausalLM"
10
+ },
11
+ "bos_token_id": 50256,
12
+ "embd_pdrop": 0.0,
13
+ "eos_token_id": 50256,
14
+ "hidden_act": "gelu_new",
15
+ "hidden_size": 2560,
16
+ "initializer_range": 0.02,
17
+ "intermediate_size": 10240,
18
+ "layer_norm_eps": 1e-05,
19
+ "max_position_embeddings": 2048,
20
+ "model_type": "phi",
21
+ "num_attention_heads": 32,
22
+ "num_hidden_layers": 32,
23
+ "num_key_value_heads": 32,
24
+ "partial_rotary_factor": 0.4,
25
+ "qk_layernorm": false,
26
+ "resid_pdrop": 0.1,
27
+ "rope_scaling": null,
28
+ "rope_theta": 10000.0,
29
+ "tie_word_embeddings": false,
30
+ "torch_dtype": "float16",
31
+ "transformers_version": "4.44.2",
32
+ "use_cache": true,
33
+ "vocab_size": 51200
34
+ }
language_model/pytorch_model.bin ADDED
@@ -0,0 +1,3 @@
 
 
 
 
1
+ version https://git-lfs.github.com/spec/v1
2
+ oid sha256:97cf4cb7f9c86e211816429acc585e7ddaa0bebe8b6355373864e71342d64f7c
3
+ size 5559512417
log.txt ADDED
@@ -0,0 +1,11 @@
 
 
 
 
 
 
 
 
 
 
 
 
1
+ 2024-10-06 18:01:33,188 | INFO: Total Parameters: 0, Total Trainable Parameters: 0
2
+ 2024-10-06 18:01:33,188 | INFO: Trainable Parameters:
3
+ 2024-10-06 18:01:33,217 | WARNING: Detected kernel version 4.14.0, which is below the recommended minimum of 5.5.0; this can cause the process to hang. It is recommended to upgrade the kernel to the minimum version or higher.
4
+ 2024-10-06 18:01:33,443 | INFO: Added key: store_based_barrier_key:2 to store for rank: 3
5
+ 2024-10-06 18:01:33,446 | INFO: Added key: store_based_barrier_key:2 to store for rank: 2
6
+ 2024-10-06 18:01:33,454 | INFO: Added key: store_based_barrier_key:2 to store for rank: 1
7
+ 2024-10-06 18:01:33,559 | INFO: Added key: store_based_barrier_key:2 to store for rank: 0
8
+ 2024-10-06 18:01:33,559 | INFO: Rank 0: Completed store-based barrier for key:store_based_barrier_key:2 with 4 nodes.
9
+ 2024-10-06 18:01:33,565 | INFO: Rank 3: Completed store-based barrier for key:store_based_barrier_key:2 with 4 nodes.
10
+ 2024-10-06 18:01:33,566 | INFO: Rank 1: Completed store-based barrier for key:store_based_barrier_key:2 with 4 nodes.
11
+ 2024-10-06 18:01:33,568 | INFO: Rank 2: Completed store-based barrier for key:store_based_barrier_key:2 with 4 nodes.
merges.txt ADDED
The diff for this file is too large to render. See raw diff
 
special_tokens_map.json ADDED
@@ -0,0 +1,30 @@
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
1
+ {
2
+ "bos_token": {
3
+ "content": "<|endoftext|>",
4
+ "lstrip": false,
5
+ "normalized": false,
6
+ "rstrip": false,
7
+ "single_word": false
8
+ },
9
+ "eos_token": {
10
+ "content": "<|endoftext|>",
11
+ "lstrip": false,
12
+ "normalized": false,
13
+ "rstrip": false,
14
+ "single_word": false
15
+ },
16
+ "pad_token": {
17
+ "content": "<|endoftext|>",
18
+ "lstrip": false,
19
+ "normalized": false,
20
+ "rstrip": false,
21
+ "single_word": false
22
+ },
23
+ "unk_token": {
24
+ "content": "<|endoftext|>",
25
+ "lstrip": false,
26
+ "normalized": false,
27
+ "rstrip": false,
28
+ "single_word": false
29
+ }
30
+ }
tokenizer_config.json ADDED
@@ -0,0 +1,328 @@
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
1
+ {
2
+ "add_bos_token": false,
3
+ "add_prefix_space": false,
4
+ "added_tokens_decoder": {
5
+ "50256": {
6
+ "content": "<|endoftext|>",
7
+ "lstrip": false,
8
+ "normalized": false,
9
+ "rstrip": false,
10
+ "single_word": false,
11
+ "special": true
12
+ },
13
+ "50257": {
14
+ "content": " ",
15
+ "lstrip": false,
16
+ "normalized": true,
17
+ "rstrip": false,
18
+ "single_word": false,
19
+ "special": false
20
+ },
21
+ "50258": {
22
+ "content": " ",
23
+ "lstrip": false,
24
+ "normalized": true,
25
+ "rstrip": false,
26
+ "single_word": false,
27
+ "special": false
28
+ },
29
+ "50259": {
30
+ "content": " ",
31
+ "lstrip": false,
32
+ "normalized": true,
33
+ "rstrip": false,
34
+ "single_word": false,
35
+ "special": false
36
+ },
37
+ "50260": {
38
+ "content": " ",
39
+ "lstrip": false,
40
+ "normalized": true,
41
+ "rstrip": false,
42
+ "single_word": false,
43
+ "special": false
44
+ },
45
+ "50261": {
46
+ "content": " ",
47
+ "lstrip": false,
48
+ "normalized": true,
49
+ "rstrip": false,
50
+ "single_word": false,
51
+ "special": false
52
+ },
53
+ "50262": {
54
+ "content": " ",
55
+ "lstrip": false,
56
+ "normalized": true,
57
+ "rstrip": false,
58
+ "single_word": false,
59
+ "special": false
60
+ },
61
+ "50263": {
62
+ "content": " ",
63
+ "lstrip": false,
64
+ "normalized": true,
65
+ "rstrip": false,
66
+ "single_word": false,
67
+ "special": false
68
+ },
69
+ "50264": {
70
+ "content": " ",
71
+ "lstrip": false,
72
+ "normalized": true,
73
+ "rstrip": false,
74
+ "single_word": false,
75
+ "special": false
76
+ },
77
+ "50265": {
78
+ "content": " ",
79
+ "lstrip": false,
80
+ "normalized": true,
81
+ "rstrip": false,
82
+ "single_word": false,
83
+ "special": false
84
+ },
85
+ "50266": {
86
+ "content": " ",
87
+ "lstrip": false,
88
+ "normalized": true,
89
+ "rstrip": false,
90
+ "single_word": false,
91
+ "special": false
92
+ },
93
+ "50267": {
94
+ "content": " ",
95
+ "lstrip": false,
96
+ "normalized": true,
97
+ "rstrip": false,
98
+ "single_word": false,
99
+ "special": false
100
+ },
101
+ "50268": {
102
+ "content": " ",
103
+ "lstrip": false,
104
+ "normalized": true,
105
+ "rstrip": false,
106
+ "single_word": false,
107
+ "special": false
108
+ },
109
+ "50269": {
110
+ "content": " ",
111
+ "lstrip": false,
112
+ "normalized": true,
113
+ "rstrip": false,
114
+ "single_word": false,
115
+ "special": false
116
+ },
117
+ "50270": {
118
+ "content": " ",
119
+ "lstrip": false,
120
+ "normalized": true,
121
+ "rstrip": false,
122
+ "single_word": false,
123
+ "special": false
124
+ },
125
+ "50271": {
126
+ "content": " ",
127
+ "lstrip": false,
128
+ "normalized": true,
129
+ "rstrip": false,
130
+ "single_word": false,
131
+ "special": false
132
+ },
133
+ "50272": {
134
+ "content": " ",
135
+ "lstrip": false,
136
+ "normalized": true,
137
+ "rstrip": false,
138
+ "single_word": false,
139
+ "special": false
140
+ },
141
+ "50273": {
142
+ "content": " ",
143
+ "lstrip": false,
144
+ "normalized": true,
145
+ "rstrip": false,
146
+ "single_word": false,
147
+ "special": false
148
+ },
149
+ "50274": {
150
+ "content": " ",
151
+ "lstrip": false,
152
+ "normalized": true,
153
+ "rstrip": false,
154
+ "single_word": false,
155
+ "special": false
156
+ },
157
+ "50275": {
158
+ "content": " ",
159
+ "lstrip": false,
160
+ "normalized": true,
161
+ "rstrip": false,
162
+ "single_word": false,
163
+ "special": false
164
+ },
165
+ "50276": {
166
+ "content": " ",
167
+ "lstrip": false,
168
+ "normalized": true,
169
+ "rstrip": false,
170
+ "single_word": false,
171
+ "special": false
172
+ },
173
+ "50277": {
174
+ "content": " ",
175
+ "lstrip": false,
176
+ "normalized": true,
177
+ "rstrip": false,
178
+ "single_word": false,
179
+ "special": false
180
+ },
181
+ "50278": {
182
+ "content": " ",
183
+ "lstrip": false,
184
+ "normalized": true,
185
+ "rstrip": false,
186
+ "single_word": false,
187
+ "special": false
188
+ },
189
+ "50279": {
190
+ "content": " ",
191
+ "lstrip": false,
192
+ "normalized": true,
193
+ "rstrip": false,
194
+ "single_word": false,
195
+ "special": false
196
+ },
197
+ "50280": {
198
+ "content": " ",
199
+ "lstrip": false,
200
+ "normalized": true,
201
+ "rstrip": false,
202
+ "single_word": false,
203
+ "special": false
204
+ },
205
+ "50281": {
206
+ "content": " ",
207
+ "lstrip": false,
208
+ "normalized": true,
209
+ "rstrip": false,
210
+ "single_word": false,
211
+ "special": false
212
+ },
213
+ "50282": {
214
+ "content": " ",
215
+ "lstrip": false,
216
+ "normalized": true,
217
+ "rstrip": false,
218
+ "single_word": false,
219
+ "special": false
220
+ },
221
+ "50283": {
222
+ "content": " ",
223
+ "lstrip": false,
224
+ "normalized": true,
225
+ "rstrip": false,
226
+ "single_word": false,
227
+ "special": false
228
+ },
229
+ "50284": {
230
+ "content": " ",
231
+ "lstrip": false,
232
+ "normalized": true,
233
+ "rstrip": false,
234
+ "single_word": false,
235
+ "special": false
236
+ },
237
+ "50285": {
238
+ "content": " ",
239
+ "lstrip": false,
240
+ "normalized": true,
241
+ "rstrip": false,
242
+ "single_word": false,
243
+ "special": false
244
+ },
245
+ "50286": {
246
+ "content": " ",
247
+ "lstrip": false,
248
+ "normalized": true,
249
+ "rstrip": false,
250
+ "single_word": false,
251
+ "special": false
252
+ },
253
+ "50287": {
254
+ "content": "\t\t\t\t\t\t\t\t\t",
255
+ "lstrip": false,
256
+ "normalized": true,
257
+ "rstrip": false,
258
+ "single_word": false,
259
+ "special": false
260
+ },
261
+ "50288": {
262
+ "content": "\t\t\t\t\t\t\t\t",
263
+ "lstrip": false,
264
+ "normalized": true,
265
+ "rstrip": false,
266
+ "single_word": false,
267
+ "special": false
268
+ },
269
+ "50289": {
270
+ "content": "\t\t\t\t\t\t\t",
271
+ "lstrip": false,
272
+ "normalized": true,
273
+ "rstrip": false,
274
+ "single_word": false,
275
+ "special": false
276
+ },
277
+ "50290": {
278
+ "content": "\t\t\t\t\t\t",
279
+ "lstrip": false,
280
+ "normalized": true,
281
+ "rstrip": false,
282
+ "single_word": false,
283
+ "special": false
284
+ },
285
+ "50291": {
286
+ "content": "\t\t\t\t\t",
287
+ "lstrip": false,
288
+ "normalized": true,
289
+ "rstrip": false,
290
+ "single_word": false,
291
+ "special": false
292
+ },
293
+ "50292": {
294
+ "content": "\t\t\t\t",
295
+ "lstrip": false,
296
+ "normalized": true,
297
+ "rstrip": false,
298
+ "single_word": false,
299
+ "special": false
300
+ },
301
+ "50293": {
302
+ "content": "\t\t\t",
303
+ "lstrip": false,
304
+ "normalized": true,
305
+ "rstrip": false,
306
+ "single_word": false,
307
+ "special": false
308
+ },
309
+ "50294": {
310
+ "content": "\t\t",
311
+ "lstrip": false,
312
+ "normalized": true,
313
+ "rstrip": false,
314
+ "single_word": false,
315
+ "special": false
316
+ }
317
+ },
318
+ "bos_token": "<|endoftext|>",
319
+ "clean_up_tokenization_spaces": true,
320
+ "eos_token": "<|endoftext|>",
321
+ "errors": "replace",
322
+ "model_max_length": 3072,
323
+ "pad_token": "<|endoftext|>",
324
+ "padding_side": "right",
325
+ "return_token_type_ids": false,
326
+ "tokenizer_class": "CodeGenTokenizer",
327
+ "unk_token": "<|endoftext|>"
328
+ }
trainer_state.json ADDED
The diff for this file is too large to render. See raw diff
 
vision_tower/config.json ADDED
@@ -0,0 +1,16 @@
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
1
+ {
2
+ "attention_dropout": 0.0,
3
+ "hidden_act": "gelu_pytorch_tanh",
4
+ "hidden_size": 1152,
5
+ "image_size": 384,
6
+ "intermediate_size": 4304,
7
+ "layer_norm_eps": 1e-06,
8
+ "model_name_or_path": "/home/work/workspace/checkpoints/siglip",
9
+ "model_name_or_path2": "",
10
+ "model_type": "siglip_vision_model",
11
+ "num_attention_heads": 16,
12
+ "num_channels": 3,
13
+ "num_hidden_layers": 27,
14
+ "patch_size": 14,
15
+ "transformers_version": "4.44.2"
16
+ }
vision_tower/pytorch_model.bin ADDED
@@ -0,0 +1,3 @@
 
 
 
 
1
+ version https://git-lfs.github.com/spec/v1
2
+ oid sha256:6b7778d8cdb7dffb024ddd5413e912218a8fd280effd545b8655a938378be94e
3
+ size 856598605
vocab.json ADDED
The diff for this file is too large to render. See raw diff