runninglsy commited on
Commit
3ff2bcd
1 Parent(s): 45c9265

upload model

Browse files
.gitattributes CHANGED
@@ -33,3 +33,4 @@ saved_model/**/* filter=lfs diff=lfs merge=lfs -text
33
  *.zip filter=lfs diff=lfs merge=lfs -text
34
  *.zst filter=lfs diff=lfs merge=lfs -text
35
  *tfevents* filter=lfs diff=lfs merge=lfs -text
 
 
33
  *.zip filter=lfs diff=lfs merge=lfs -text
34
  *.zst filter=lfs diff=lfs merge=lfs -text
35
  *tfevents* filter=lfs diff=lfs merge=lfs -text
36
+ tokenizer.json filter=lfs diff=lfs merge=lfs -text
README.md CHANGED
@@ -1,3 +1,89 @@
1
- ---
2
- license: apache-2.0
3
- ---
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
1
+ ---
2
+ license: apache-2.0
3
+ datasets:
4
+ - AIDC-AI/Ovis-dataset
5
+ library_name: transformers
6
+ tags:
7
+ - MLLM
8
+ pipeline_tag: image-text-to-text
9
+ ---
10
+
11
+ ## Introduction
12
+ Ovis is a novel Multimodal Large Language Model (MLLM) architecture, designed to structurally align visual and textual embeddings. For a comprehensive introduction, please refer to [Ovis paper](https://arxiv.org/abs/2405.20797) and [Ovis GitHub](https://github.com/AIDC-AI/Ovis).
13
+
14
+ <div align="center">
15
+ <img src="https://cdn-uploads.huggingface.co/production/uploads/658a8a837959448ef5500ce5/TIlymOb86R6_Mez3bpmcB.png" width="100%" />
16
+ </div>
17
+
18
+ ## Model
19
+ Built upon Ovis1.5, Ovis1.6 further enhances high-resolution image processing, is trained on a larger, more diverse, and higher-quality dataset, and refines the training process with DPO training following instruction-tuning.
20
+
21
+ | Ovis MLLMs | ViT | LLM | Model Weights |
22
+ |:------------------|:-----------:|:------------------:|:---------------------------------------------------------------:|
23
+ | Ovis1.6-Gemma2-9B | Siglip-400M | Gemma2-9B-It | [Huggingface](https://huggingface.co/AIDC-AI/Ovis1.6-Gemma2-9B) |
24
+
25
+ ## Performance
26
+ With just **10B** parameters, Ovis1.6-Gemma2-9B leads the [OpenCompass](https://github.com/open-compass/VLMEvalKit) benchmark among open-source MLLMs within **30B** parameters.
27
+
28
+ ## Usage
29
+ Below is a code snippet to run Ovis with multimodal inputs. For additional usage instructions, including inference wrapper and Gradio UI, please refer to [Ovis GitHub](https://github.com/AIDC-AI/Ovis?tab=readme-ov-file#inference).
30
+ ```bash
31
+ pip install torch==2.2.0 transformers==4.44.2 numpy==1.24.3
32
+ ```
33
+ ```python
34
+ import torch
35
+ from PIL import Image
36
+ from transformers import AutoModelForCausalLM
37
+
38
+ # load model
39
+ model = AutoModelForCausalLM.from_pretrained("AIDC-AI/Ovis1.6-Gemma2-9B",
40
+ torch_dtype=torch.bfloat16,
41
+ multimodal_max_length=8192,
42
+ trust_remote_code=True).cuda()
43
+ text_tokenizer = model.get_text_tokenizer()
44
+ visual_tokenizer = model.get_visual_tokenizer()
45
+
46
+ # enter image path and prompt
47
+ image_path = input("Enter image path: ")
48
+ image = Image.open(image_path)
49
+ text = input("Enter prompt: ")
50
+ query = f'<image>\n{text}'
51
+
52
+ # format conversation
53
+ prompt, input_ids, pixel_values = model.preprocess_inputs(query, [image])
54
+ attention_mask = torch.ne(input_ids, text_tokenizer.pad_token_id)
55
+ input_ids = input_ids.unsqueeze(0).to(device=model.device)
56
+ attention_mask = attention_mask.unsqueeze(0).to(device=model.device)
57
+ pixel_values = [pixel_values.to(dtype=visual_tokenizer.dtype, device=visual_tokenizer.device)]
58
+
59
+ # generate output
60
+ with torch.inference_mode():
61
+ gen_kwargs = dict(
62
+ max_new_tokens=1024,
63
+ do_sample=False,
64
+ top_p=None,
65
+ top_k=None,
66
+ temperature=None,
67
+ repetition_penalty=None,
68
+ eos_token_id=model.generation_config.eos_token_id,
69
+ pad_token_id=text_tokenizer.pad_token_id,
70
+ use_cache=True
71
+ )
72
+ output_ids = model.generate(input_ids, pixel_values=pixel_values, attention_mask=attention_mask, **gen_kwargs)[0]
73
+ output = text_tokenizer.decode(output_ids, skip_special_tokens=True)
74
+ print(f'Output:\n{output}')
75
+ ```
76
+
77
+ ## Citation
78
+ If you find Ovis useful, please cite the paper
79
+ ```
80
+ @article{lu2024ovis,
81
+ title={Ovis: Structural Embedding Alignment for Multimodal Large Language Model},
82
+ author={Shiyin Lu and Yang Li and Qing-Guo Chen and Zhao Xu and Weihua Luo and Kaifu Zhang and Han-Jia Ye},
83
+ year={2024},
84
+ journal={arXiv:2405.20797}
85
+ }
86
+ ```
87
+
88
+ ## License
89
+ The project is licensed under the Apache 2.0 License and is restricted to uses that comply with the license agreements of Gemma2 and Siglip.
config.json ADDED
@@ -0,0 +1,248 @@
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
1
+ {
2
+ "architectures": [
3
+ "Ovis"
4
+ ],
5
+ "auto_map": {
6
+ "AutoConfig": "configuration_ovis.OvisConfig",
7
+ "AutoModelForCausalLM": "modeling_ovis.Ovis"
8
+ },
9
+ "conversation_formatter_class": "GemmaConversationFormatter",
10
+ "disable_tie_weight": false,
11
+ "hidden_size": 3584,
12
+ "llm_attn_implementation": "eager",
13
+ "llm_config": {
14
+ "_name_or_path": "google/gemma-2-9b-it",
15
+ "add_cross_attention": false,
16
+ "architectures": [
17
+ "Gemma2ForCausalLM"
18
+ ],
19
+ "attention_bias": false,
20
+ "attention_dropout": 0.0,
21
+ "attn_logit_softcapping": 50.0,
22
+ "bad_words_ids": null,
23
+ "begin_suppress_tokens": null,
24
+ "bos_token_id": 2,
25
+ "cache_implementation": "hybrid",
26
+ "chunk_size_feed_forward": 0,
27
+ "cross_attention_hidden_size": null,
28
+ "decoder_start_token_id": null,
29
+ "diversity_penalty": 0.0,
30
+ "do_sample": false,
31
+ "early_stopping": false,
32
+ "encoder_no_repeat_ngram_size": 0,
33
+ "eos_token_id": 1,
34
+ "exponential_decay_length_penalty": null,
35
+ "final_logit_softcapping": 30.0,
36
+ "finetuning_task": null,
37
+ "forced_bos_token_id": null,
38
+ "forced_eos_token_id": null,
39
+ "head_dim": 256,
40
+ "hidden_act": "gelu_pytorch_tanh",
41
+ "hidden_activation": "gelu_pytorch_tanh",
42
+ "hidden_size": 3584,
43
+ "id2label": {
44
+ "0": "LABEL_0",
45
+ "1": "LABEL_1"
46
+ },
47
+ "initializer_range": 0.02,
48
+ "intermediate_size": 14336,
49
+ "is_decoder": false,
50
+ "is_encoder_decoder": false,
51
+ "label2id": {
52
+ "LABEL_0": 0,
53
+ "LABEL_1": 1
54
+ },
55
+ "length_penalty": 1.0,
56
+ "max_length": 20,
57
+ "max_position_embeddings": 8192,
58
+ "min_length": 0,
59
+ "model_type": "gemma2",
60
+ "no_repeat_ngram_size": 0,
61
+ "num_attention_heads": 16,
62
+ "num_beam_groups": 1,
63
+ "num_beams": 1,
64
+ "num_hidden_layers": 42,
65
+ "num_key_value_heads": 8,
66
+ "num_return_sequences": 1,
67
+ "output_attentions": false,
68
+ "output_hidden_states": false,
69
+ "output_scores": false,
70
+ "pad_token_id": 0,
71
+ "prefix": null,
72
+ "problem_type": null,
73
+ "pruned_heads": {},
74
+ "query_pre_attn_scalar": 256,
75
+ "remove_invalid_values": false,
76
+ "repetition_penalty": 1.0,
77
+ "return_dict": true,
78
+ "return_dict_in_generate": false,
79
+ "rms_norm_eps": 1e-06,
80
+ "rope_theta": 10000.0,
81
+ "sep_token_id": null,
82
+ "sliding_window": 4096,
83
+ "sliding_window_size": 4096,
84
+ "suppress_tokens": null,
85
+ "task_specific_params": null,
86
+ "temperature": 1.0,
87
+ "tf_legacy_loss": false,
88
+ "tie_encoder_decoder": false,
89
+ "tie_word_embeddings": true,
90
+ "tokenizer_class": null,
91
+ "top_k": 50,
92
+ "top_p": 1.0,
93
+ "torch_dtype": "bfloat16",
94
+ "torchscript": false,
95
+ "typical_p": 1.0,
96
+ "use_bfloat16": false,
97
+ "use_cache": true,
98
+ "vocab_size": 256000
99
+ },
100
+ "model_type": "ovis",
101
+ "multimodal_max_length": 8192,
102
+ "torch_dtype": "bfloat16",
103
+ "transformers_version": "4.44.2",
104
+ "use_cache": true,
105
+ "visual_tokenizer_config": {
106
+ "_name_or_path": "",
107
+ "add_cross_attention": false,
108
+ "architectures": null,
109
+ "backbone_config": {
110
+ "_name_or_path": "google/siglip-so400m-patch14-384",
111
+ "add_cross_attention": false,
112
+ "architectures": null,
113
+ "attention_dropout": 0.0,
114
+ "bad_words_ids": null,
115
+ "begin_suppress_tokens": null,
116
+ "bos_token_id": null,
117
+ "chunk_size_feed_forward": 0,
118
+ "cross_attention_hidden_size": null,
119
+ "decoder_start_token_id": null,
120
+ "diversity_penalty": 0.0,
121
+ "do_sample": false,
122
+ "early_stopping": false,
123
+ "encoder_no_repeat_ngram_size": 0,
124
+ "eos_token_id": null,
125
+ "exponential_decay_length_penalty": null,
126
+ "finetuning_task": null,
127
+ "forced_bos_token_id": null,
128
+ "forced_eos_token_id": null,
129
+ "hidden_act": "gelu_pytorch_tanh",
130
+ "hidden_size": 1152,
131
+ "id2label": {
132
+ "0": "LABEL_0",
133
+ "1": "LABEL_1"
134
+ },
135
+ "image_size": 384,
136
+ "intermediate_size": 4304,
137
+ "is_decoder": false,
138
+ "is_encoder_decoder": false,
139
+ "label2id": {
140
+ "LABEL_0": 0,
141
+ "LABEL_1": 1
142
+ },
143
+ "layer_norm_eps": 1e-06,
144
+ "length_penalty": 1.0,
145
+ "max_length": 20,
146
+ "min_length": 0,
147
+ "model_type": "siglip_vision_model",
148
+ "no_repeat_ngram_size": 0,
149
+ "num_attention_heads": 16,
150
+ "num_beam_groups": 1,
151
+ "num_beams": 1,
152
+ "num_channels": 3,
153
+ "num_hidden_layers": 27,
154
+ "num_return_sequences": 1,
155
+ "output_attentions": false,
156
+ "output_hidden_states": false,
157
+ "output_scores": false,
158
+ "pad_token_id": null,
159
+ "patch_size": 14,
160
+ "prefix": null,
161
+ "problem_type": null,
162
+ "pruned_heads": {},
163
+ "remove_invalid_values": false,
164
+ "repetition_penalty": 1.0,
165
+ "return_dict": true,
166
+ "return_dict_in_generate": false,
167
+ "sep_token_id": null,
168
+ "suppress_tokens": null,
169
+ "task_specific_params": null,
170
+ "temperature": 1.0,
171
+ "tf_legacy_loss": false,
172
+ "tie_encoder_decoder": false,
173
+ "tie_word_embeddings": true,
174
+ "tokenizer_class": null,
175
+ "top_k": 50,
176
+ "top_p": 1.0,
177
+ "torch_dtype": null,
178
+ "torchscript": false,
179
+ "typical_p": 1.0,
180
+ "use_bfloat16": false
181
+ },
182
+ "backbone_kwargs": {},
183
+ "bad_words_ids": null,
184
+ "begin_suppress_tokens": null,
185
+ "bos_token_id": null,
186
+ "chunk_size_feed_forward": 0,
187
+ "cross_attention_hidden_size": null,
188
+ "decoder_start_token_id": null,
189
+ "depths": null,
190
+ "diversity_penalty": 0.0,
191
+ "do_sample": false,
192
+ "drop_cls_token": false,
193
+ "early_stopping": false,
194
+ "encoder_no_repeat_ngram_size": 0,
195
+ "eos_token_id": null,
196
+ "exponential_decay_length_penalty": null,
197
+ "finetuning_task": null,
198
+ "forced_bos_token_id": null,
199
+ "forced_eos_token_id": null,
200
+ "hidden_stride": 2,
201
+ "id2label": {
202
+ "0": "LABEL_0",
203
+ "1": "LABEL_1"
204
+ },
205
+ "is_decoder": false,
206
+ "is_encoder_decoder": false,
207
+ "label2id": {
208
+ "LABEL_0": 0,
209
+ "LABEL_1": 1
210
+ },
211
+ "length_penalty": 1.0,
212
+ "max_length": 20,
213
+ "min_length": 0,
214
+ "model_type": "siglip_visual_tokenizer",
215
+ "no_repeat_ngram_size": 0,
216
+ "num_beam_groups": 1,
217
+ "num_beams": 1,
218
+ "num_return_sequences": 1,
219
+ "output_attentions": false,
220
+ "output_hidden_states": false,
221
+ "output_scores": false,
222
+ "pad_token_id": null,
223
+ "prefix": null,
224
+ "problem_type": null,
225
+ "pruned_heads": {},
226
+ "remove_invalid_values": false,
227
+ "repetition_penalty": 1.0,
228
+ "return_dict": true,
229
+ "return_dict_in_generate": false,
230
+ "sep_token_id": null,
231
+ "suppress_tokens": null,
232
+ "task_specific_params": null,
233
+ "tau": 1.0,
234
+ "temperature": 1.0,
235
+ "tf_legacy_loss": false,
236
+ "tie_encoder_decoder": false,
237
+ "tie_word_embeddings": true,
238
+ "tokenize_function": "softmax",
239
+ "tokenizer_class": null,
240
+ "top_k": 50,
241
+ "top_p": 1.0,
242
+ "torch_dtype": null,
243
+ "torchscript": false,
244
+ "typical_p": 1.0,
245
+ "use_bfloat16": false,
246
+ "vocab_size": 65536
247
+ }
248
+ }
configuration_ovis.py ADDED
@@ -0,0 +1,201 @@
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
1
+ from abc import ABC, abstractmethod
2
+ from typing import List, Dict, Union, Optional
3
+
4
+ from transformers import PretrainedConfig, AutoConfig
5
+
6
+ IGNORE_ID = -100
7
+ IMAGE_TOKEN_ID = -200
8
+ IMAGE_TOKEN = "<image>"
9
+ IMAGE_ATOM_ID = -300
10
+ IMAGE_INDICATOR_IDS = [-301, -302, -303, -304, -305]
11
+
12
+
13
+ # ----------------------------------------------------------------------
14
+ # Visual Tokenizer Configuration
15
+ # ----------------------------------------------------------------------
16
+ class BaseVisualTokenizerConfig(PretrainedConfig):
17
+ def __init__(
18
+ self,
19
+ vocab_size=16384,
20
+ tokenize_function="softmax",
21
+ tau=1.0,
22
+ depths=None,
23
+ drop_cls_token=False,
24
+ backbone_config: Optional[Union[PretrainedConfig, dict]] = None,
25
+ hidden_stride: int = 1,
26
+ **kwargs
27
+ ):
28
+ super().__init__(**kwargs)
29
+ self.vocab_size = vocab_size
30
+ self.tokenize_function = tokenize_function
31
+ self.tau = tau
32
+ if isinstance(depths, str):
33
+ depths = [int(x) for x in depths.split('|')]
34
+ self.depths = depths
35
+ self.backbone_kwargs = {}
36
+ self.drop_cls_token = drop_cls_token
37
+ if backbone_config is not None:
38
+ assert isinstance(backbone_config, (PretrainedConfig, dict)), \
39
+ f"expect `backbone_config` to be instance of PretrainedConfig or dict, but got {type(backbone_config)} type"
40
+ if not isinstance(backbone_config, PretrainedConfig):
41
+ model_type = backbone_config['model_type']
42
+ backbone_config.pop('model_type')
43
+ backbone_config = AutoConfig.for_model(model_type, **backbone_config)
44
+ self.backbone_config = backbone_config
45
+ self.hidden_stride = hidden_stride
46
+
47
+
48
+ class SiglipVisualTokenizerConfig(BaseVisualTokenizerConfig):
49
+ model_type = "siglip_visual_tokenizer"
50
+
51
+ def __init__(self, **kwargs):
52
+ super().__init__(**kwargs)
53
+ if self.drop_cls_token:
54
+ self.drop_cls_token = False
55
+ if self.depths:
56
+ assert len(self.depths) == 1
57
+ self.backbone_kwargs['num_hidden_layers'] = self.depths[0]
58
+
59
+
60
+ AutoConfig.register("siglip_visual_tokenizer", SiglipVisualTokenizerConfig)
61
+
62
+
63
+ # ----------------------------------------------------------------------
64
+ # Ovis Configuration
65
+ # ----------------------------------------------------------------------
66
+ class OvisConfig(PretrainedConfig):
67
+ model_type = "ovis"
68
+
69
+ def __init__(
70
+ self,
71
+ llm_config: Optional[Union[PretrainedConfig, dict]] = None,
72
+ visual_tokenizer_config: Optional[Union[PretrainedConfig, dict]] = None,
73
+ multimodal_max_length=8192,
74
+ hidden_size=None,
75
+ conversation_formatter_class=None,
76
+ llm_attn_implementation=None,
77
+ disable_tie_weight=False,
78
+ **kwargs
79
+ ):
80
+ super().__init__(**kwargs)
81
+ if llm_config is not None:
82
+ assert isinstance(llm_config, (PretrainedConfig, dict)), \
83
+ f"expect `llm_config` to be instance of PretrainedConfig or dict, but got {type(llm_config)} type"
84
+ if not isinstance(llm_config, PretrainedConfig):
85
+ model_type = llm_config['model_type']
86
+ llm_config.pop('model_type')
87
+ llm_config = AutoConfig.for_model(model_type, **llm_config)
88
+ self.llm_config = llm_config
89
+ if visual_tokenizer_config is not None:
90
+ assert isinstance(visual_tokenizer_config, (PretrainedConfig, dict)), \
91
+ f"expect `visual_tokenizer_config` to be instance of PretrainedConfig or dict, but got {type(visual_tokenizer_config)} type"
92
+ if not isinstance(visual_tokenizer_config, PretrainedConfig):
93
+ model_type = visual_tokenizer_config['model_type']
94
+ visual_tokenizer_config.pop('model_type')
95
+ visual_tokenizer_config = AutoConfig.for_model(model_type, **visual_tokenizer_config)
96
+ self.visual_tokenizer_config = visual_tokenizer_config
97
+ self.multimodal_max_length = multimodal_max_length
98
+ self.hidden_size = hidden_size
99
+ self.conversation_formatter_class = conversation_formatter_class
100
+ self.llm_attn_implementation = llm_attn_implementation
101
+ self.disable_tie_weight = disable_tie_weight
102
+
103
+
104
+ # ----------------------------------------------------------------------
105
+ # Conversation Formatter
106
+ # ----------------------------------------------------------------------
107
+ class ConversationFormatter(ABC):
108
+ support_tokenizer_types = None
109
+
110
+ def __init__(self, tokenizer):
111
+ tokenizer_type = type(tokenizer).__name__
112
+ assert tokenizer_type in self.support_tokenizer_types, \
113
+ f'Invalid tokenizer type, expected one from `{self.support_tokenizer_types}`, but got `{tokenizer_type}`'
114
+ self.tokenizer = tokenizer
115
+ self.image_token = IMAGE_TOKEN
116
+ self.image_token_id = IMAGE_TOKEN_ID
117
+ self.ignore_id = IGNORE_ID
118
+
119
+ def _tokenize_with_image_symbol(self, text):
120
+ text_chunks = [self.tokenizer(chunk, add_special_tokens=False).input_ids for chunk in
121
+ text.split(self.image_token)]
122
+ token_ids = []
123
+ num_chuck = len(text_chunks)
124
+ for i, chunk in enumerate(text_chunks):
125
+ token_ids.extend(chunk)
126
+ if i < num_chuck - 1:
127
+ token_ids.append(self.image_token_id)
128
+ return token_ids
129
+
130
+ @abstractmethod
131
+ def format(self, conversations: List[Dict], generation_preface=None):
132
+ pass
133
+
134
+ @abstractmethod
135
+ def format_query(self, query, generation_preface=""):
136
+ pass
137
+
138
+
139
+ class GemmaConversationFormatter(ConversationFormatter):
140
+ support_tokenizer_types = ['GemmaTokenizer', 'GemmaTokenizerFast']
141
+
142
+ def __init__(self, tokenizer):
143
+ super().__init__(tokenizer)
144
+ # Gemma does not support system prompt
145
+ self.from2role = {
146
+ "human": "<start_of_turn>user\n",
147
+ "gpt": "<start_of_turn>model\n",
148
+ }
149
+ self.gpt_token_num = None
150
+ self.im_end = "<end_of_turn>\n"
151
+ self.bos_token = "<bos>"
152
+ self.bos_token_ids = None
153
+
154
+ def format(self, conversations: List[Dict], generation_preface=None):
155
+ if self.gpt_token_num is None:
156
+ self.gpt_token_num = len(self.tokenizer(self.from2role["gpt"], add_special_tokens=False).input_ids)
157
+
158
+ if self.bos_token_ids is None:
159
+ self.bos_token_ids = self.tokenizer(self.bos_token, add_special_tokens=False).input_ids
160
+
161
+ if conversations[0]["from"] == "system":
162
+ raise ValueError("Gemma does not support system prompt")
163
+
164
+ if generation_preface is not None:
165
+ conversations.append({
166
+ "from": "gpt",
167
+ "value": generation_preface
168
+ })
169
+
170
+ prompt = "" + self.bos_token
171
+ input_ids = [] + self.bos_token_ids
172
+ labels = [] + [IGNORE_ID] * len(input_ids)
173
+ num_conversation = len(conversations)
174
+ for i, conversation in enumerate(conversations):
175
+ frm = conversation["from"]
176
+ role = self.from2role[frm]
177
+ message = conversation["value"].strip()
178
+ text = role + message
179
+ if i < num_conversation - 1 or generation_preface is None:
180
+ text += self.im_end
181
+ prompt += text
182
+ token_ids = self._tokenize_with_image_symbol(text)
183
+ input_ids.extend(token_ids)
184
+ label_ids = [self.ignore_id] * len(token_ids)
185
+ if frm == "gpt":
186
+ # learning `\n` following `im_end` is meaningless, so the last `\n` token is ignored in label
187
+ label_ids[self.gpt_token_num:-1] = token_ids[self.gpt_token_num:-1]
188
+ labels.extend(label_ids)
189
+
190
+ assert self._tokenize_with_image_symbol(prompt) == input_ids
191
+ assert len(input_ids) == len(labels)
192
+
193
+ return prompt, input_ids, labels
194
+
195
+ def format_query(self, query, generation_preface=""):
196
+ prompt, input_ids, _ = self.format([{
197
+ "from": "human",
198
+ "value": query
199
+ }], generation_preface=generation_preface)
200
+
201
+ return prompt, input_ids
generation_config.json ADDED
@@ -0,0 +1,11 @@
 
 
 
 
 
 
 
 
 
 
 
 
1
+ {
2
+ "_from_model_config": true,
3
+ "bos_token_id": 2,
4
+ "cache_implementation": "hybrid",
5
+ "eos_token_id": [
6
+ 1,
7
+ 107
8
+ ],
9
+ "pad_token_id": 0,
10
+ "transformers_version": "4.44.2"
11
+ }
model-00001-of-00005.safetensors ADDED
@@ -0,0 +1,3 @@
 
 
 
 
1
+ version https://git-lfs.github.com/spec/v1
2
+ oid sha256:4be2fe245c4c3c52ce9cf1447c39a283d49c4cb902ddf7709fcec4dfae0275ce
3
+ size 4903352248
model-00002-of-00005.safetensors ADDED
@@ -0,0 +1,3 @@
 
 
 
 
1
+ version https://git-lfs.github.com/spec/v1
2
+ oid sha256:ad3948fdc3904281c0d102b981594339724944a8058881354f24e94544483faa
3
+ size 4947571432
model-00003-of-00005.safetensors ADDED
@@ -0,0 +1,3 @@
 
 
 
 
1
+ version https://git-lfs.github.com/spec/v1
2
+ oid sha256:79352679bd5af40bf245c43750378403df96a1fa50106d0eb398da163f997dc3
3
+ size 4962222000
model-00004-of-00005.safetensors ADDED
@@ -0,0 +1,3 @@
 
 
 
 
1
+ version https://git-lfs.github.com/spec/v1
2
+ oid sha256:389198229ab08c096a303436eee2ecaab074b4fe14c7ff527822124bc128c4d9
3
+ size 4526841384
model-00005-of-00005.safetensors ADDED
@@ -0,0 +1,3 @@
 
 
 
 
1
+ version https://git-lfs.github.com/spec/v1
2
+ oid sha256:4589319220d8f6ebdc34abdc5492446aeef6c8f21c5f83ea798f4470138dab12
3
+ size 1073958308
model.safetensors.index.json ADDED
@@ -0,0 +1,923 @@
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
1
+ {
2
+ "metadata": {
3
+ "total_size": 20413821036
4
+ },
5
+ "weight_map": {
6
+ "llm.model.embed_tokens.weight": "model-00001-of-00005.safetensors",
7
+ "llm.model.layers.0.input_layernorm.weight": "model-00001-of-00005.safetensors",
8
+ "llm.model.layers.0.mlp.down_proj.weight": "model-00001-of-00005.safetensors",
9
+ "llm.model.layers.0.mlp.gate_proj.weight": "model-00001-of-00005.safetensors",
10
+ "llm.model.layers.0.mlp.up_proj.weight": "model-00001-of-00005.safetensors",
11
+ "llm.model.layers.0.post_attention_layernorm.weight": "model-00001-of-00005.safetensors",
12
+ "llm.model.layers.0.post_feedforward_layernorm.weight": "model-00001-of-00005.safetensors",
13
+ "llm.model.layers.0.pre_feedforward_layernorm.weight": "model-00001-of-00005.safetensors",
14
+ "llm.model.layers.0.self_attn.k_proj.weight": "model-00001-of-00005.safetensors",
15
+ "llm.model.layers.0.self_attn.o_proj.weight": "model-00001-of-00005.safetensors",
16
+ "llm.model.layers.0.self_attn.q_proj.weight": "model-00001-of-00005.safetensors",
17
+ "llm.model.layers.0.self_attn.v_proj.weight": "model-00001-of-00005.safetensors",
18
+ "llm.model.layers.1.input_layernorm.weight": "model-00001-of-00005.safetensors",
19
+ "llm.model.layers.1.mlp.down_proj.weight": "model-00001-of-00005.safetensors",
20
+ "llm.model.layers.1.mlp.gate_proj.weight": "model-00001-of-00005.safetensors",
21
+ "llm.model.layers.1.mlp.up_proj.weight": "model-00001-of-00005.safetensors",
22
+ "llm.model.layers.1.post_attention_layernorm.weight": "model-00001-of-00005.safetensors",
23
+ "llm.model.layers.1.post_feedforward_layernorm.weight": "model-00001-of-00005.safetensors",
24
+ "llm.model.layers.1.pre_feedforward_layernorm.weight": "model-00001-of-00005.safetensors",
25
+ "llm.model.layers.1.self_attn.k_proj.weight": "model-00001-of-00005.safetensors",
26
+ "llm.model.layers.1.self_attn.o_proj.weight": "model-00001-of-00005.safetensors",
27
+ "llm.model.layers.1.self_attn.q_proj.weight": "model-00001-of-00005.safetensors",
28
+ "llm.model.layers.1.self_attn.v_proj.weight": "model-00001-of-00005.safetensors",
29
+ "llm.model.layers.10.input_layernorm.weight": "model-00002-of-00005.safetensors",
30
+ "llm.model.layers.10.mlp.down_proj.weight": "model-00002-of-00005.safetensors",
31
+ "llm.model.layers.10.mlp.gate_proj.weight": "model-00002-of-00005.safetensors",
32
+ "llm.model.layers.10.mlp.up_proj.weight": "model-00002-of-00005.safetensors",
33
+ "llm.model.layers.10.post_attention_layernorm.weight": "model-00002-of-00005.safetensors",
34
+ "llm.model.layers.10.post_feedforward_layernorm.weight": "model-00002-of-00005.safetensors",
35
+ "llm.model.layers.10.pre_feedforward_layernorm.weight": "model-00002-of-00005.safetensors",
36
+ "llm.model.layers.10.self_attn.k_proj.weight": "model-00002-of-00005.safetensors",
37
+ "llm.model.layers.10.self_attn.o_proj.weight": "model-00002-of-00005.safetensors",
38
+ "llm.model.layers.10.self_attn.q_proj.weight": "model-00002-of-00005.safetensors",
39
+ "llm.model.layers.10.self_attn.v_proj.weight": "model-00002-of-00005.safetensors",
40
+ "llm.model.layers.11.input_layernorm.weight": "model-00002-of-00005.safetensors",
41
+ "llm.model.layers.11.mlp.down_proj.weight": "model-00002-of-00005.safetensors",
42
+ "llm.model.layers.11.mlp.gate_proj.weight": "model-00002-of-00005.safetensors",
43
+ "llm.model.layers.11.mlp.up_proj.weight": "model-00002-of-00005.safetensors",
44
+ "llm.model.layers.11.post_attention_layernorm.weight": "model-00002-of-00005.safetensors",
45
+ "llm.model.layers.11.post_feedforward_layernorm.weight": "model-00002-of-00005.safetensors",
46
+ "llm.model.layers.11.pre_feedforward_layernorm.weight": "model-00002-of-00005.safetensors",
47
+ "llm.model.layers.11.self_attn.k_proj.weight": "model-00002-of-00005.safetensors",
48
+ "llm.model.layers.11.self_attn.o_proj.weight": "model-00002-of-00005.safetensors",
49
+ "llm.model.layers.11.self_attn.q_proj.weight": "model-00002-of-00005.safetensors",
50
+ "llm.model.layers.11.self_attn.v_proj.weight": "model-00002-of-00005.safetensors",
51
+ "llm.model.layers.12.input_layernorm.weight": "model-00002-of-00005.safetensors",
52
+ "llm.model.layers.12.mlp.down_proj.weight": "model-00002-of-00005.safetensors",
53
+ "llm.model.layers.12.mlp.gate_proj.weight": "model-00002-of-00005.safetensors",
54
+ "llm.model.layers.12.mlp.up_proj.weight": "model-00002-of-00005.safetensors",
55
+ "llm.model.layers.12.post_attention_layernorm.weight": "model-00002-of-00005.safetensors",
56
+ "llm.model.layers.12.post_feedforward_layernorm.weight": "model-00002-of-00005.safetensors",
57
+ "llm.model.layers.12.pre_feedforward_layernorm.weight": "model-00002-of-00005.safetensors",
58
+ "llm.model.layers.12.self_attn.k_proj.weight": "model-00002-of-00005.safetensors",
59
+ "llm.model.layers.12.self_attn.o_proj.weight": "model-00002-of-00005.safetensors",
60
+ "llm.model.layers.12.self_attn.q_proj.weight": "model-00002-of-00005.safetensors",
61
+ "llm.model.layers.12.self_attn.v_proj.weight": "model-00002-of-00005.safetensors",
62
+ "llm.model.layers.13.input_layernorm.weight": "model-00002-of-00005.safetensors",
63
+ "llm.model.layers.13.mlp.down_proj.weight": "model-00002-of-00005.safetensors",
64
+ "llm.model.layers.13.mlp.gate_proj.weight": "model-00002-of-00005.safetensors",
65
+ "llm.model.layers.13.mlp.up_proj.weight": "model-00002-of-00005.safetensors",
66
+ "llm.model.layers.13.post_attention_layernorm.weight": "model-00002-of-00005.safetensors",
67
+ "llm.model.layers.13.post_feedforward_layernorm.weight": "model-00002-of-00005.safetensors",
68
+ "llm.model.layers.13.pre_feedforward_layernorm.weight": "model-00002-of-00005.safetensors",
69
+ "llm.model.layers.13.self_attn.k_proj.weight": "model-00002-of-00005.safetensors",
70
+ "llm.model.layers.13.self_attn.o_proj.weight": "model-00002-of-00005.safetensors",
71
+ "llm.model.layers.13.self_attn.q_proj.weight": "model-00002-of-00005.safetensors",
72
+ "llm.model.layers.13.self_attn.v_proj.weight": "model-00002-of-00005.safetensors",
73
+ "llm.model.layers.14.input_layernorm.weight": "model-00002-of-00005.safetensors",
74
+ "llm.model.layers.14.mlp.down_proj.weight": "model-00002-of-00005.safetensors",
75
+ "llm.model.layers.14.mlp.gate_proj.weight": "model-00002-of-00005.safetensors",
76
+ "llm.model.layers.14.mlp.up_proj.weight": "model-00002-of-00005.safetensors",
77
+ "llm.model.layers.14.post_attention_layernorm.weight": "model-00002-of-00005.safetensors",
78
+ "llm.model.layers.14.post_feedforward_layernorm.weight": "model-00002-of-00005.safetensors",
79
+ "llm.model.layers.14.pre_feedforward_layernorm.weight": "model-00002-of-00005.safetensors",
80
+ "llm.model.layers.14.self_attn.k_proj.weight": "model-00002-of-00005.safetensors",
81
+ "llm.model.layers.14.self_attn.o_proj.weight": "model-00002-of-00005.safetensors",
82
+ "llm.model.layers.14.self_attn.q_proj.weight": "model-00002-of-00005.safetensors",
83
+ "llm.model.layers.14.self_attn.v_proj.weight": "model-00002-of-00005.safetensors",
84
+ "llm.model.layers.15.input_layernorm.weight": "model-00002-of-00005.safetensors",
85
+ "llm.model.layers.15.mlp.down_proj.weight": "model-00002-of-00005.safetensors",
86
+ "llm.model.layers.15.mlp.gate_proj.weight": "model-00002-of-00005.safetensors",
87
+ "llm.model.layers.15.mlp.up_proj.weight": "model-00002-of-00005.safetensors",
88
+ "llm.model.layers.15.post_attention_layernorm.weight": "model-00002-of-00005.safetensors",
89
+ "llm.model.layers.15.post_feedforward_layernorm.weight": "model-00002-of-00005.safetensors",
90
+ "llm.model.layers.15.pre_feedforward_layernorm.weight": "model-00002-of-00005.safetensors",
91
+ "llm.model.layers.15.self_attn.k_proj.weight": "model-00002-of-00005.safetensors",
92
+ "llm.model.layers.15.self_attn.o_proj.weight": "model-00002-of-00005.safetensors",
93
+ "llm.model.layers.15.self_attn.q_proj.weight": "model-00002-of-00005.safetensors",
94
+ "llm.model.layers.15.self_attn.v_proj.weight": "model-00002-of-00005.safetensors",
95
+ "llm.model.layers.16.input_layernorm.weight": "model-00002-of-00005.safetensors",
96
+ "llm.model.layers.16.mlp.down_proj.weight": "model-00002-of-00005.safetensors",
97
+ "llm.model.layers.16.mlp.gate_proj.weight": "model-00002-of-00005.safetensors",
98
+ "llm.model.layers.16.mlp.up_proj.weight": "model-00002-of-00005.safetensors",
99
+ "llm.model.layers.16.post_attention_layernorm.weight": "model-00002-of-00005.safetensors",
100
+ "llm.model.layers.16.post_feedforward_layernorm.weight": "model-00002-of-00005.safetensors",
101
+ "llm.model.layers.16.pre_feedforward_layernorm.weight": "model-00002-of-00005.safetensors",
102
+ "llm.model.layers.16.self_attn.k_proj.weight": "model-00002-of-00005.safetensors",
103
+ "llm.model.layers.16.self_attn.o_proj.weight": "model-00002-of-00005.safetensors",
104
+ "llm.model.layers.16.self_attn.q_proj.weight": "model-00002-of-00005.safetensors",
105
+ "llm.model.layers.16.self_attn.v_proj.weight": "model-00002-of-00005.safetensors",
106
+ "llm.model.layers.17.input_layernorm.weight": "model-00002-of-00005.safetensors",
107
+ "llm.model.layers.17.mlp.down_proj.weight": "model-00002-of-00005.safetensors",
108
+ "llm.model.layers.17.mlp.gate_proj.weight": "model-00002-of-00005.safetensors",
109
+ "llm.model.layers.17.mlp.up_proj.weight": "model-00002-of-00005.safetensors",
110
+ "llm.model.layers.17.post_attention_layernorm.weight": "model-00002-of-00005.safetensors",
111
+ "llm.model.layers.17.post_feedforward_layernorm.weight": "model-00002-of-00005.safetensors",
112
+ "llm.model.layers.17.pre_feedforward_layernorm.weight": "model-00002-of-00005.safetensors",
113
+ "llm.model.layers.17.self_attn.k_proj.weight": "model-00002-of-00005.safetensors",
114
+ "llm.model.layers.17.self_attn.o_proj.weight": "model-00002-of-00005.safetensors",
115
+ "llm.model.layers.17.self_attn.q_proj.weight": "model-00002-of-00005.safetensors",
116
+ "llm.model.layers.17.self_attn.v_proj.weight": "model-00002-of-00005.safetensors",
117
+ "llm.model.layers.18.input_layernorm.weight": "model-00002-of-00005.safetensors",
118
+ "llm.model.layers.18.mlp.down_proj.weight": "model-00002-of-00005.safetensors",
119
+ "llm.model.layers.18.mlp.gate_proj.weight": "model-00002-of-00005.safetensors",
120
+ "llm.model.layers.18.mlp.up_proj.weight": "model-00002-of-00005.safetensors",
121
+ "llm.model.layers.18.post_attention_layernorm.weight": "model-00002-of-00005.safetensors",
122
+ "llm.model.layers.18.post_feedforward_layernorm.weight": "model-00002-of-00005.safetensors",
123
+ "llm.model.layers.18.pre_feedforward_layernorm.weight": "model-00002-of-00005.safetensors",
124
+ "llm.model.layers.18.self_attn.k_proj.weight": "model-00002-of-00005.safetensors",
125
+ "llm.model.layers.18.self_attn.o_proj.weight": "model-00002-of-00005.safetensors",
126
+ "llm.model.layers.18.self_attn.q_proj.weight": "model-00002-of-00005.safetensors",
127
+ "llm.model.layers.18.self_attn.v_proj.weight": "model-00002-of-00005.safetensors",
128
+ "llm.model.layers.19.input_layernorm.weight": "model-00002-of-00005.safetensors",
129
+ "llm.model.layers.19.mlp.down_proj.weight": "model-00002-of-00005.safetensors",
130
+ "llm.model.layers.19.mlp.gate_proj.weight": "model-00002-of-00005.safetensors",
131
+ "llm.model.layers.19.mlp.up_proj.weight": "model-00002-of-00005.safetensors",
132
+ "llm.model.layers.19.post_attention_layernorm.weight": "model-00002-of-00005.safetensors",
133
+ "llm.model.layers.19.post_feedforward_layernorm.weight": "model-00002-of-00005.safetensors",
134
+ "llm.model.layers.19.pre_feedforward_layernorm.weight": "model-00002-of-00005.safetensors",
135
+ "llm.model.layers.19.self_attn.k_proj.weight": "model-00002-of-00005.safetensors",
136
+ "llm.model.layers.19.self_attn.o_proj.weight": "model-00002-of-00005.safetensors",
137
+ "llm.model.layers.19.self_attn.q_proj.weight": "model-00002-of-00005.safetensors",
138
+ "llm.model.layers.19.self_attn.v_proj.weight": "model-00002-of-00005.safetensors",
139
+ "llm.model.layers.2.input_layernorm.weight": "model-00001-of-00005.safetensors",
140
+ "llm.model.layers.2.mlp.down_proj.weight": "model-00001-of-00005.safetensors",
141
+ "llm.model.layers.2.mlp.gate_proj.weight": "model-00001-of-00005.safetensors",
142
+ "llm.model.layers.2.mlp.up_proj.weight": "model-00001-of-00005.safetensors",
143
+ "llm.model.layers.2.post_attention_layernorm.weight": "model-00001-of-00005.safetensors",
144
+ "llm.model.layers.2.post_feedforward_layernorm.weight": "model-00001-of-00005.safetensors",
145
+ "llm.model.layers.2.pre_feedforward_layernorm.weight": "model-00001-of-00005.safetensors",
146
+ "llm.model.layers.2.self_attn.k_proj.weight": "model-00001-of-00005.safetensors",
147
+ "llm.model.layers.2.self_attn.o_proj.weight": "model-00001-of-00005.safetensors",
148
+ "llm.model.layers.2.self_attn.q_proj.weight": "model-00001-of-00005.safetensors",
149
+ "llm.model.layers.2.self_attn.v_proj.weight": "model-00001-of-00005.safetensors",
150
+ "llm.model.layers.20.input_layernorm.weight": "model-00003-of-00005.safetensors",
151
+ "llm.model.layers.20.mlp.down_proj.weight": "model-00003-of-00005.safetensors",
152
+ "llm.model.layers.20.mlp.gate_proj.weight": "model-00003-of-00005.safetensors",
153
+ "llm.model.layers.20.mlp.up_proj.weight": "model-00003-of-00005.safetensors",
154
+ "llm.model.layers.20.post_attention_layernorm.weight": "model-00003-of-00005.safetensors",
155
+ "llm.model.layers.20.post_feedforward_layernorm.weight": "model-00003-of-00005.safetensors",
156
+ "llm.model.layers.20.pre_feedforward_layernorm.weight": "model-00003-of-00005.safetensors",
157
+ "llm.model.layers.20.self_attn.k_proj.weight": "model-00002-of-00005.safetensors",
158
+ "llm.model.layers.20.self_attn.o_proj.weight": "model-00002-of-00005.safetensors",
159
+ "llm.model.layers.20.self_attn.q_proj.weight": "model-00002-of-00005.safetensors",
160
+ "llm.model.layers.20.self_attn.v_proj.weight": "model-00002-of-00005.safetensors",
161
+ "llm.model.layers.21.input_layernorm.weight": "model-00003-of-00005.safetensors",
162
+ "llm.model.layers.21.mlp.down_proj.weight": "model-00003-of-00005.safetensors",
163
+ "llm.model.layers.21.mlp.gate_proj.weight": "model-00003-of-00005.safetensors",
164
+ "llm.model.layers.21.mlp.up_proj.weight": "model-00003-of-00005.safetensors",
165
+ "llm.model.layers.21.post_attention_layernorm.weight": "model-00003-of-00005.safetensors",
166
+ "llm.model.layers.21.post_feedforward_layernorm.weight": "model-00003-of-00005.safetensors",
167
+ "llm.model.layers.21.pre_feedforward_layernorm.weight": "model-00003-of-00005.safetensors",
168
+ "llm.model.layers.21.self_attn.k_proj.weight": "model-00003-of-00005.safetensors",
169
+ "llm.model.layers.21.self_attn.o_proj.weight": "model-00003-of-00005.safetensors",
170
+ "llm.model.layers.21.self_attn.q_proj.weight": "model-00003-of-00005.safetensors",
171
+ "llm.model.layers.21.self_attn.v_proj.weight": "model-00003-of-00005.safetensors",
172
+ "llm.model.layers.22.input_layernorm.weight": "model-00003-of-00005.safetensors",
173
+ "llm.model.layers.22.mlp.down_proj.weight": "model-00003-of-00005.safetensors",
174
+ "llm.model.layers.22.mlp.gate_proj.weight": "model-00003-of-00005.safetensors",
175
+ "llm.model.layers.22.mlp.up_proj.weight": "model-00003-of-00005.safetensors",
176
+ "llm.model.layers.22.post_attention_layernorm.weight": "model-00003-of-00005.safetensors",
177
+ "llm.model.layers.22.post_feedforward_layernorm.weight": "model-00003-of-00005.safetensors",
178
+ "llm.model.layers.22.pre_feedforward_layernorm.weight": "model-00003-of-00005.safetensors",
179
+ "llm.model.layers.22.self_attn.k_proj.weight": "model-00003-of-00005.safetensors",
180
+ "llm.model.layers.22.self_attn.o_proj.weight": "model-00003-of-00005.safetensors",
181
+ "llm.model.layers.22.self_attn.q_proj.weight": "model-00003-of-00005.safetensors",
182
+ "llm.model.layers.22.self_attn.v_proj.weight": "model-00003-of-00005.safetensors",
183
+ "llm.model.layers.23.input_layernorm.weight": "model-00003-of-00005.safetensors",
184
+ "llm.model.layers.23.mlp.down_proj.weight": "model-00003-of-00005.safetensors",
185
+ "llm.model.layers.23.mlp.gate_proj.weight": "model-00003-of-00005.safetensors",
186
+ "llm.model.layers.23.mlp.up_proj.weight": "model-00003-of-00005.safetensors",
187
+ "llm.model.layers.23.post_attention_layernorm.weight": "model-00003-of-00005.safetensors",
188
+ "llm.model.layers.23.post_feedforward_layernorm.weight": "model-00003-of-00005.safetensors",
189
+ "llm.model.layers.23.pre_feedforward_layernorm.weight": "model-00003-of-00005.safetensors",
190
+ "llm.model.layers.23.self_attn.k_proj.weight": "model-00003-of-00005.safetensors",
191
+ "llm.model.layers.23.self_attn.o_proj.weight": "model-00003-of-00005.safetensors",
192
+ "llm.model.layers.23.self_attn.q_proj.weight": "model-00003-of-00005.safetensors",
193
+ "llm.model.layers.23.self_attn.v_proj.weight": "model-00003-of-00005.safetensors",
194
+ "llm.model.layers.24.input_layernorm.weight": "model-00003-of-00005.safetensors",
195
+ "llm.model.layers.24.mlp.down_proj.weight": "model-00003-of-00005.safetensors",
196
+ "llm.model.layers.24.mlp.gate_proj.weight": "model-00003-of-00005.safetensors",
197
+ "llm.model.layers.24.mlp.up_proj.weight": "model-00003-of-00005.safetensors",
198
+ "llm.model.layers.24.post_attention_layernorm.weight": "model-00003-of-00005.safetensors",
199
+ "llm.model.layers.24.post_feedforward_layernorm.weight": "model-00003-of-00005.safetensors",
200
+ "llm.model.layers.24.pre_feedforward_layernorm.weight": "model-00003-of-00005.safetensors",
201
+ "llm.model.layers.24.self_attn.k_proj.weight": "model-00003-of-00005.safetensors",
202
+ "llm.model.layers.24.self_attn.o_proj.weight": "model-00003-of-00005.safetensors",
203
+ "llm.model.layers.24.self_attn.q_proj.weight": "model-00003-of-00005.safetensors",
204
+ "llm.model.layers.24.self_attn.v_proj.weight": "model-00003-of-00005.safetensors",
205
+ "llm.model.layers.25.input_layernorm.weight": "model-00003-of-00005.safetensors",
206
+ "llm.model.layers.25.mlp.down_proj.weight": "model-00003-of-00005.safetensors",
207
+ "llm.model.layers.25.mlp.gate_proj.weight": "model-00003-of-00005.safetensors",
208
+ "llm.model.layers.25.mlp.up_proj.weight": "model-00003-of-00005.safetensors",
209
+ "llm.model.layers.25.post_attention_layernorm.weight": "model-00003-of-00005.safetensors",
210
+ "llm.model.layers.25.post_feedforward_layernorm.weight": "model-00003-of-00005.safetensors",
211
+ "llm.model.layers.25.pre_feedforward_layernorm.weight": "model-00003-of-00005.safetensors",
212
+ "llm.model.layers.25.self_attn.k_proj.weight": "model-00003-of-00005.safetensors",
213
+ "llm.model.layers.25.self_attn.o_proj.weight": "model-00003-of-00005.safetensors",
214
+ "llm.model.layers.25.self_attn.q_proj.weight": "model-00003-of-00005.safetensors",
215
+ "llm.model.layers.25.self_attn.v_proj.weight": "model-00003-of-00005.safetensors",
216
+ "llm.model.layers.26.input_layernorm.weight": "model-00003-of-00005.safetensors",
217
+ "llm.model.layers.26.mlp.down_proj.weight": "model-00003-of-00005.safetensors",
218
+ "llm.model.layers.26.mlp.gate_proj.weight": "model-00003-of-00005.safetensors",
219
+ "llm.model.layers.26.mlp.up_proj.weight": "model-00003-of-00005.safetensors",
220
+ "llm.model.layers.26.post_attention_layernorm.weight": "model-00003-of-00005.safetensors",
221
+ "llm.model.layers.26.post_feedforward_layernorm.weight": "model-00003-of-00005.safetensors",
222
+ "llm.model.layers.26.pre_feedforward_layernorm.weight": "model-00003-of-00005.safetensors",
223
+ "llm.model.layers.26.self_attn.k_proj.weight": "model-00003-of-00005.safetensors",
224
+ "llm.model.layers.26.self_attn.o_proj.weight": "model-00003-of-00005.safetensors",
225
+ "llm.model.layers.26.self_attn.q_proj.weight": "model-00003-of-00005.safetensors",
226
+ "llm.model.layers.26.self_attn.v_proj.weight": "model-00003-of-00005.safetensors",
227
+ "llm.model.layers.27.input_layernorm.weight": "model-00003-of-00005.safetensors",
228
+ "llm.model.layers.27.mlp.down_proj.weight": "model-00003-of-00005.safetensors",
229
+ "llm.model.layers.27.mlp.gate_proj.weight": "model-00003-of-00005.safetensors",
230
+ "llm.model.layers.27.mlp.up_proj.weight": "model-00003-of-00005.safetensors",
231
+ "llm.model.layers.27.post_attention_layernorm.weight": "model-00003-of-00005.safetensors",
232
+ "llm.model.layers.27.post_feedforward_layernorm.weight": "model-00003-of-00005.safetensors",
233
+ "llm.model.layers.27.pre_feedforward_layernorm.weight": "model-00003-of-00005.safetensors",
234
+ "llm.model.layers.27.self_attn.k_proj.weight": "model-00003-of-00005.safetensors",
235
+ "llm.model.layers.27.self_attn.o_proj.weight": "model-00003-of-00005.safetensors",
236
+ "llm.model.layers.27.self_attn.q_proj.weight": "model-00003-of-00005.safetensors",
237
+ "llm.model.layers.27.self_attn.v_proj.weight": "model-00003-of-00005.safetensors",
238
+ "llm.model.layers.28.input_layernorm.weight": "model-00003-of-00005.safetensors",
239
+ "llm.model.layers.28.mlp.down_proj.weight": "model-00003-of-00005.safetensors",
240
+ "llm.model.layers.28.mlp.gate_proj.weight": "model-00003-of-00005.safetensors",
241
+ "llm.model.layers.28.mlp.up_proj.weight": "model-00003-of-00005.safetensors",
242
+ "llm.model.layers.28.post_attention_layernorm.weight": "model-00003-of-00005.safetensors",
243
+ "llm.model.layers.28.post_feedforward_layernorm.weight": "model-00003-of-00005.safetensors",
244
+ "llm.model.layers.28.pre_feedforward_layernorm.weight": "model-00003-of-00005.safetensors",
245
+ "llm.model.layers.28.self_attn.k_proj.weight": "model-00003-of-00005.safetensors",
246
+ "llm.model.layers.28.self_attn.o_proj.weight": "model-00003-of-00005.safetensors",
247
+ "llm.model.layers.28.self_attn.q_proj.weight": "model-00003-of-00005.safetensors",
248
+ "llm.model.layers.28.self_attn.v_proj.weight": "model-00003-of-00005.safetensors",
249
+ "llm.model.layers.29.input_layernorm.weight": "model-00003-of-00005.safetensors",
250
+ "llm.model.layers.29.mlp.down_proj.weight": "model-00003-of-00005.safetensors",
251
+ "llm.model.layers.29.mlp.gate_proj.weight": "model-00003-of-00005.safetensors",
252
+ "llm.model.layers.29.mlp.up_proj.weight": "model-00003-of-00005.safetensors",
253
+ "llm.model.layers.29.post_attention_layernorm.weight": "model-00003-of-00005.safetensors",
254
+ "llm.model.layers.29.post_feedforward_layernorm.weight": "model-00003-of-00005.safetensors",
255
+ "llm.model.layers.29.pre_feedforward_layernorm.weight": "model-00003-of-00005.safetensors",
256
+ "llm.model.layers.29.self_attn.k_proj.weight": "model-00003-of-00005.safetensors",
257
+ "llm.model.layers.29.self_attn.o_proj.weight": "model-00003-of-00005.safetensors",
258
+ "llm.model.layers.29.self_attn.q_proj.weight": "model-00003-of-00005.safetensors",
259
+ "llm.model.layers.29.self_attn.v_proj.weight": "model-00003-of-00005.safetensors",
260
+ "llm.model.layers.3.input_layernorm.weight": "model-00001-of-00005.safetensors",
261
+ "llm.model.layers.3.mlp.down_proj.weight": "model-00001-of-00005.safetensors",
262
+ "llm.model.layers.3.mlp.gate_proj.weight": "model-00001-of-00005.safetensors",
263
+ "llm.model.layers.3.mlp.up_proj.weight": "model-00001-of-00005.safetensors",
264
+ "llm.model.layers.3.post_attention_layernorm.weight": "model-00001-of-00005.safetensors",
265
+ "llm.model.layers.3.post_feedforward_layernorm.weight": "model-00001-of-00005.safetensors",
266
+ "llm.model.layers.3.pre_feedforward_layernorm.weight": "model-00001-of-00005.safetensors",
267
+ "llm.model.layers.3.self_attn.k_proj.weight": "model-00001-of-00005.safetensors",
268
+ "llm.model.layers.3.self_attn.o_proj.weight": "model-00001-of-00005.safetensors",
269
+ "llm.model.layers.3.self_attn.q_proj.weight": "model-00001-of-00005.safetensors",
270
+ "llm.model.layers.3.self_attn.v_proj.weight": "model-00001-of-00005.safetensors",
271
+ "llm.model.layers.30.input_layernorm.weight": "model-00003-of-00005.safetensors",
272
+ "llm.model.layers.30.mlp.down_proj.weight": "model-00003-of-00005.safetensors",
273
+ "llm.model.layers.30.mlp.gate_proj.weight": "model-00003-of-00005.safetensors",
274
+ "llm.model.layers.30.mlp.up_proj.weight": "model-00003-of-00005.safetensors",
275
+ "llm.model.layers.30.post_attention_layernorm.weight": "model-00003-of-00005.safetensors",
276
+ "llm.model.layers.30.post_feedforward_layernorm.weight": "model-00003-of-00005.safetensors",
277
+ "llm.model.layers.30.pre_feedforward_layernorm.weight": "model-00003-of-00005.safetensors",
278
+ "llm.model.layers.30.self_attn.k_proj.weight": "model-00003-of-00005.safetensors",
279
+ "llm.model.layers.30.self_attn.o_proj.weight": "model-00003-of-00005.safetensors",
280
+ "llm.model.layers.30.self_attn.q_proj.weight": "model-00003-of-00005.safetensors",
281
+ "llm.model.layers.30.self_attn.v_proj.weight": "model-00003-of-00005.safetensors",
282
+ "llm.model.layers.31.input_layernorm.weight": "model-00003-of-00005.safetensors",
283
+ "llm.model.layers.31.mlp.down_proj.weight": "model-00003-of-00005.safetensors",
284
+ "llm.model.layers.31.mlp.gate_proj.weight": "model-00003-of-00005.safetensors",
285
+ "llm.model.layers.31.mlp.up_proj.weight": "model-00003-of-00005.safetensors",
286
+ "llm.model.layers.31.post_attention_layernorm.weight": "model-00003-of-00005.safetensors",
287
+ "llm.model.layers.31.post_feedforward_layernorm.weight": "model-00003-of-00005.safetensors",
288
+ "llm.model.layers.31.pre_feedforward_layernorm.weight": "model-00003-of-00005.safetensors",
289
+ "llm.model.layers.31.self_attn.k_proj.weight": "model-00003-of-00005.safetensors",
290
+ "llm.model.layers.31.self_attn.o_proj.weight": "model-00003-of-00005.safetensors",
291
+ "llm.model.layers.31.self_attn.q_proj.weight": "model-00003-of-00005.safetensors",
292
+ "llm.model.layers.31.self_attn.v_proj.weight": "model-00003-of-00005.safetensors",
293
+ "llm.model.layers.32.input_layernorm.weight": "model-00004-of-00005.safetensors",
294
+ "llm.model.layers.32.mlp.down_proj.weight": "model-00004-of-00005.safetensors",
295
+ "llm.model.layers.32.mlp.gate_proj.weight": "model-00003-of-00005.safetensors",
296
+ "llm.model.layers.32.mlp.up_proj.weight": "model-00003-of-00005.safetensors",
297
+ "llm.model.layers.32.post_attention_layernorm.weight": "model-00004-of-00005.safetensors",
298
+ "llm.model.layers.32.post_feedforward_layernorm.weight": "model-00004-of-00005.safetensors",
299
+ "llm.model.layers.32.pre_feedforward_layernorm.weight": "model-00004-of-00005.safetensors",
300
+ "llm.model.layers.32.self_attn.k_proj.weight": "model-00003-of-00005.safetensors",
301
+ "llm.model.layers.32.self_attn.o_proj.weight": "model-00003-of-00005.safetensors",
302
+ "llm.model.layers.32.self_attn.q_proj.weight": "model-00003-of-00005.safetensors",
303
+ "llm.model.layers.32.self_attn.v_proj.weight": "model-00003-of-00005.safetensors",
304
+ "llm.model.layers.33.input_layernorm.weight": "model-00004-of-00005.safetensors",
305
+ "llm.model.layers.33.mlp.down_proj.weight": "model-00004-of-00005.safetensors",
306
+ "llm.model.layers.33.mlp.gate_proj.weight": "model-00004-of-00005.safetensors",
307
+ "llm.model.layers.33.mlp.up_proj.weight": "model-00004-of-00005.safetensors",
308
+ "llm.model.layers.33.post_attention_layernorm.weight": "model-00004-of-00005.safetensors",
309
+ "llm.model.layers.33.post_feedforward_layernorm.weight": "model-00004-of-00005.safetensors",
310
+ "llm.model.layers.33.pre_feedforward_layernorm.weight": "model-00004-of-00005.safetensors",
311
+ "llm.model.layers.33.self_attn.k_proj.weight": "model-00004-of-00005.safetensors",
312
+ "llm.model.layers.33.self_attn.o_proj.weight": "model-00004-of-00005.safetensors",
313
+ "llm.model.layers.33.self_attn.q_proj.weight": "model-00004-of-00005.safetensors",
314
+ "llm.model.layers.33.self_attn.v_proj.weight": "model-00004-of-00005.safetensors",
315
+ "llm.model.layers.34.input_layernorm.weight": "model-00004-of-00005.safetensors",
316
+ "llm.model.layers.34.mlp.down_proj.weight": "model-00004-of-00005.safetensors",
317
+ "llm.model.layers.34.mlp.gate_proj.weight": "model-00004-of-00005.safetensors",
318
+ "llm.model.layers.34.mlp.up_proj.weight": "model-00004-of-00005.safetensors",
319
+ "llm.model.layers.34.post_attention_layernorm.weight": "model-00004-of-00005.safetensors",
320
+ "llm.model.layers.34.post_feedforward_layernorm.weight": "model-00004-of-00005.safetensors",
321
+ "llm.model.layers.34.pre_feedforward_layernorm.weight": "model-00004-of-00005.safetensors",
322
+ "llm.model.layers.34.self_attn.k_proj.weight": "model-00004-of-00005.safetensors",
323
+ "llm.model.layers.34.self_attn.o_proj.weight": "model-00004-of-00005.safetensors",
324
+ "llm.model.layers.34.self_attn.q_proj.weight": "model-00004-of-00005.safetensors",
325
+ "llm.model.layers.34.self_attn.v_proj.weight": "model-00004-of-00005.safetensors",
326
+ "llm.model.layers.35.input_layernorm.weight": "model-00004-of-00005.safetensors",
327
+ "llm.model.layers.35.mlp.down_proj.weight": "model-00004-of-00005.safetensors",
328
+ "llm.model.layers.35.mlp.gate_proj.weight": "model-00004-of-00005.safetensors",
329
+ "llm.model.layers.35.mlp.up_proj.weight": "model-00004-of-00005.safetensors",
330
+ "llm.model.layers.35.post_attention_layernorm.weight": "model-00004-of-00005.safetensors",
331
+ "llm.model.layers.35.post_feedforward_layernorm.weight": "model-00004-of-00005.safetensors",
332
+ "llm.model.layers.35.pre_feedforward_layernorm.weight": "model-00004-of-00005.safetensors",
333
+ "llm.model.layers.35.self_attn.k_proj.weight": "model-00004-of-00005.safetensors",
334
+ "llm.model.layers.35.self_attn.o_proj.weight": "model-00004-of-00005.safetensors",
335
+ "llm.model.layers.35.self_attn.q_proj.weight": "model-00004-of-00005.safetensors",
336
+ "llm.model.layers.35.self_attn.v_proj.weight": "model-00004-of-00005.safetensors",
337
+ "llm.model.layers.36.input_layernorm.weight": "model-00004-of-00005.safetensors",
338
+ "llm.model.layers.36.mlp.down_proj.weight": "model-00004-of-00005.safetensors",
339
+ "llm.model.layers.36.mlp.gate_proj.weight": "model-00004-of-00005.safetensors",
340
+ "llm.model.layers.36.mlp.up_proj.weight": "model-00004-of-00005.safetensors",
341
+ "llm.model.layers.36.post_attention_layernorm.weight": "model-00004-of-00005.safetensors",
342
+ "llm.model.layers.36.post_feedforward_layernorm.weight": "model-00004-of-00005.safetensors",
343
+ "llm.model.layers.36.pre_feedforward_layernorm.weight": "model-00004-of-00005.safetensors",
344
+ "llm.model.layers.36.self_attn.k_proj.weight": "model-00004-of-00005.safetensors",
345
+ "llm.model.layers.36.self_attn.o_proj.weight": "model-00004-of-00005.safetensors",
346
+ "llm.model.layers.36.self_attn.q_proj.weight": "model-00004-of-00005.safetensors",
347
+ "llm.model.layers.36.self_attn.v_proj.weight": "model-00004-of-00005.safetensors",
348
+ "llm.model.layers.37.input_layernorm.weight": "model-00004-of-00005.safetensors",
349
+ "llm.model.layers.37.mlp.down_proj.weight": "model-00004-of-00005.safetensors",
350
+ "llm.model.layers.37.mlp.gate_proj.weight": "model-00004-of-00005.safetensors",
351
+ "llm.model.layers.37.mlp.up_proj.weight": "model-00004-of-00005.safetensors",
352
+ "llm.model.layers.37.post_attention_layernorm.weight": "model-00004-of-00005.safetensors",
353
+ "llm.model.layers.37.post_feedforward_layernorm.weight": "model-00004-of-00005.safetensors",
354
+ "llm.model.layers.37.pre_feedforward_layernorm.weight": "model-00004-of-00005.safetensors",
355
+ "llm.model.layers.37.self_attn.k_proj.weight": "model-00004-of-00005.safetensors",
356
+ "llm.model.layers.37.self_attn.o_proj.weight": "model-00004-of-00005.safetensors",
357
+ "llm.model.layers.37.self_attn.q_proj.weight": "model-00004-of-00005.safetensors",
358
+ "llm.model.layers.37.self_attn.v_proj.weight": "model-00004-of-00005.safetensors",
359
+ "llm.model.layers.38.input_layernorm.weight": "model-00004-of-00005.safetensors",
360
+ "llm.model.layers.38.mlp.down_proj.weight": "model-00004-of-00005.safetensors",
361
+ "llm.model.layers.38.mlp.gate_proj.weight": "model-00004-of-00005.safetensors",
362
+ "llm.model.layers.38.mlp.up_proj.weight": "model-00004-of-00005.safetensors",
363
+ "llm.model.layers.38.post_attention_layernorm.weight": "model-00004-of-00005.safetensors",
364
+ "llm.model.layers.38.post_feedforward_layernorm.weight": "model-00004-of-00005.safetensors",
365
+ "llm.model.layers.38.pre_feedforward_layernorm.weight": "model-00004-of-00005.safetensors",
366
+ "llm.model.layers.38.self_attn.k_proj.weight": "model-00004-of-00005.safetensors",
367
+ "llm.model.layers.38.self_attn.o_proj.weight": "model-00004-of-00005.safetensors",
368
+ "llm.model.layers.38.self_attn.q_proj.weight": "model-00004-of-00005.safetensors",
369
+ "llm.model.layers.38.self_attn.v_proj.weight": "model-00004-of-00005.safetensors",
370
+ "llm.model.layers.39.input_layernorm.weight": "model-00004-of-00005.safetensors",
371
+ "llm.model.layers.39.mlp.down_proj.weight": "model-00004-of-00005.safetensors",
372
+ "llm.model.layers.39.mlp.gate_proj.weight": "model-00004-of-00005.safetensors",
373
+ "llm.model.layers.39.mlp.up_proj.weight": "model-00004-of-00005.safetensors",
374
+ "llm.model.layers.39.post_attention_layernorm.weight": "model-00004-of-00005.safetensors",
375
+ "llm.model.layers.39.post_feedforward_layernorm.weight": "model-00004-of-00005.safetensors",
376
+ "llm.model.layers.39.pre_feedforward_layernorm.weight": "model-00004-of-00005.safetensors",
377
+ "llm.model.layers.39.self_attn.k_proj.weight": "model-00004-of-00005.safetensors",
378
+ "llm.model.layers.39.self_attn.o_proj.weight": "model-00004-of-00005.safetensors",
379
+ "llm.model.layers.39.self_attn.q_proj.weight": "model-00004-of-00005.safetensors",
380
+ "llm.model.layers.39.self_attn.v_proj.weight": "model-00004-of-00005.safetensors",
381
+ "llm.model.layers.4.input_layernorm.weight": "model-00001-of-00005.safetensors",
382
+ "llm.model.layers.4.mlp.down_proj.weight": "model-00001-of-00005.safetensors",
383
+ "llm.model.layers.4.mlp.gate_proj.weight": "model-00001-of-00005.safetensors",
384
+ "llm.model.layers.4.mlp.up_proj.weight": "model-00001-of-00005.safetensors",
385
+ "llm.model.layers.4.post_attention_layernorm.weight": "model-00001-of-00005.safetensors",
386
+ "llm.model.layers.4.post_feedforward_layernorm.weight": "model-00001-of-00005.safetensors",
387
+ "llm.model.layers.4.pre_feedforward_layernorm.weight": "model-00001-of-00005.safetensors",
388
+ "llm.model.layers.4.self_attn.k_proj.weight": "model-00001-of-00005.safetensors",
389
+ "llm.model.layers.4.self_attn.o_proj.weight": "model-00001-of-00005.safetensors",
390
+ "llm.model.layers.4.self_attn.q_proj.weight": "model-00001-of-00005.safetensors",
391
+ "llm.model.layers.4.self_attn.v_proj.weight": "model-00001-of-00005.safetensors",
392
+ "llm.model.layers.40.input_layernorm.weight": "model-00004-of-00005.safetensors",
393
+ "llm.model.layers.40.mlp.down_proj.weight": "model-00004-of-00005.safetensors",
394
+ "llm.model.layers.40.mlp.gate_proj.weight": "model-00004-of-00005.safetensors",
395
+ "llm.model.layers.40.mlp.up_proj.weight": "model-00004-of-00005.safetensors",
396
+ "llm.model.layers.40.post_attention_layernorm.weight": "model-00004-of-00005.safetensors",
397
+ "llm.model.layers.40.post_feedforward_layernorm.weight": "model-00004-of-00005.safetensors",
398
+ "llm.model.layers.40.pre_feedforward_layernorm.weight": "model-00004-of-00005.safetensors",
399
+ "llm.model.layers.40.self_attn.k_proj.weight": "model-00004-of-00005.safetensors",
400
+ "llm.model.layers.40.self_attn.o_proj.weight": "model-00004-of-00005.safetensors",
401
+ "llm.model.layers.40.self_attn.q_proj.weight": "model-00004-of-00005.safetensors",
402
+ "llm.model.layers.40.self_attn.v_proj.weight": "model-00004-of-00005.safetensors",
403
+ "llm.model.layers.41.input_layernorm.weight": "model-00004-of-00005.safetensors",
404
+ "llm.model.layers.41.mlp.down_proj.weight": "model-00004-of-00005.safetensors",
405
+ "llm.model.layers.41.mlp.gate_proj.weight": "model-00004-of-00005.safetensors",
406
+ "llm.model.layers.41.mlp.up_proj.weight": "model-00004-of-00005.safetensors",
407
+ "llm.model.layers.41.post_attention_layernorm.weight": "model-00004-of-00005.safetensors",
408
+ "llm.model.layers.41.post_feedforward_layernorm.weight": "model-00004-of-00005.safetensors",
409
+ "llm.model.layers.41.pre_feedforward_layernorm.weight": "model-00004-of-00005.safetensors",
410
+ "llm.model.layers.41.self_attn.k_proj.weight": "model-00004-of-00005.safetensors",
411
+ "llm.model.layers.41.self_attn.o_proj.weight": "model-00004-of-00005.safetensors",
412
+ "llm.model.layers.41.self_attn.q_proj.weight": "model-00004-of-00005.safetensors",
413
+ "llm.model.layers.41.self_attn.v_proj.weight": "model-00004-of-00005.safetensors",
414
+ "llm.model.layers.5.input_layernorm.weight": "model-00001-of-00005.safetensors",
415
+ "llm.model.layers.5.mlp.down_proj.weight": "model-00001-of-00005.safetensors",
416
+ "llm.model.layers.5.mlp.gate_proj.weight": "model-00001-of-00005.safetensors",
417
+ "llm.model.layers.5.mlp.up_proj.weight": "model-00001-of-00005.safetensors",
418
+ "llm.model.layers.5.post_attention_layernorm.weight": "model-00001-of-00005.safetensors",
419
+ "llm.model.layers.5.post_feedforward_layernorm.weight": "model-00001-of-00005.safetensors",
420
+ "llm.model.layers.5.pre_feedforward_layernorm.weight": "model-00001-of-00005.safetensors",
421
+ "llm.model.layers.5.self_attn.k_proj.weight": "model-00001-of-00005.safetensors",
422
+ "llm.model.layers.5.self_attn.o_proj.weight": "model-00001-of-00005.safetensors",
423
+ "llm.model.layers.5.self_attn.q_proj.weight": "model-00001-of-00005.safetensors",
424
+ "llm.model.layers.5.self_attn.v_proj.weight": "model-00001-of-00005.safetensors",
425
+ "llm.model.layers.6.input_layernorm.weight": "model-00001-of-00005.safetensors",
426
+ "llm.model.layers.6.mlp.down_proj.weight": "model-00001-of-00005.safetensors",
427
+ "llm.model.layers.6.mlp.gate_proj.weight": "model-00001-of-00005.safetensors",
428
+ "llm.model.layers.6.mlp.up_proj.weight": "model-00001-of-00005.safetensors",
429
+ "llm.model.layers.6.post_attention_layernorm.weight": "model-00001-of-00005.safetensors",
430
+ "llm.model.layers.6.post_feedforward_layernorm.weight": "model-00001-of-00005.safetensors",
431
+ "llm.model.layers.6.pre_feedforward_layernorm.weight": "model-00001-of-00005.safetensors",
432
+ "llm.model.layers.6.self_attn.k_proj.weight": "model-00001-of-00005.safetensors",
433
+ "llm.model.layers.6.self_attn.o_proj.weight": "model-00001-of-00005.safetensors",
434
+ "llm.model.layers.6.self_attn.q_proj.weight": "model-00001-of-00005.safetensors",
435
+ "llm.model.layers.6.self_attn.v_proj.weight": "model-00001-of-00005.safetensors",
436
+ "llm.model.layers.7.input_layernorm.weight": "model-00002-of-00005.safetensors",
437
+ "llm.model.layers.7.mlp.down_proj.weight": "model-00002-of-00005.safetensors",
438
+ "llm.model.layers.7.mlp.gate_proj.weight": "model-00001-of-00005.safetensors",
439
+ "llm.model.layers.7.mlp.up_proj.weight": "model-00001-of-00005.safetensors",
440
+ "llm.model.layers.7.post_attention_layernorm.weight": "model-00002-of-00005.safetensors",
441
+ "llm.model.layers.7.post_feedforward_layernorm.weight": "model-00002-of-00005.safetensors",
442
+ "llm.model.layers.7.pre_feedforward_layernorm.weight": "model-00002-of-00005.safetensors",
443
+ "llm.model.layers.7.self_attn.k_proj.weight": "model-00001-of-00005.safetensors",
444
+ "llm.model.layers.7.self_attn.o_proj.weight": "model-00001-of-00005.safetensors",
445
+ "llm.model.layers.7.self_attn.q_proj.weight": "model-00001-of-00005.safetensors",
446
+ "llm.model.layers.7.self_attn.v_proj.weight": "model-00001-of-00005.safetensors",
447
+ "llm.model.layers.8.input_layernorm.weight": "model-00002-of-00005.safetensors",
448
+ "llm.model.layers.8.mlp.down_proj.weight": "model-00002-of-00005.safetensors",
449
+ "llm.model.layers.8.mlp.gate_proj.weight": "model-00002-of-00005.safetensors",
450
+ "llm.model.layers.8.mlp.up_proj.weight": "model-00002-of-00005.safetensors",
451
+ "llm.model.layers.8.post_attention_layernorm.weight": "model-00002-of-00005.safetensors",
452
+ "llm.model.layers.8.post_feedforward_layernorm.weight": "model-00002-of-00005.safetensors",
453
+ "llm.model.layers.8.pre_feedforward_layernorm.weight": "model-00002-of-00005.safetensors",
454
+ "llm.model.layers.8.self_attn.k_proj.weight": "model-00002-of-00005.safetensors",
455
+ "llm.model.layers.8.self_attn.o_proj.weight": "model-00002-of-00005.safetensors",
456
+ "llm.model.layers.8.self_attn.q_proj.weight": "model-00002-of-00005.safetensors",
457
+ "llm.model.layers.8.self_attn.v_proj.weight": "model-00002-of-00005.safetensors",
458
+ "llm.model.layers.9.input_layernorm.weight": "model-00002-of-00005.safetensors",
459
+ "llm.model.layers.9.mlp.down_proj.weight": "model-00002-of-00005.safetensors",
460
+ "llm.model.layers.9.mlp.gate_proj.weight": "model-00002-of-00005.safetensors",
461
+ "llm.model.layers.9.mlp.up_proj.weight": "model-00002-of-00005.safetensors",
462
+ "llm.model.layers.9.post_attention_layernorm.weight": "model-00002-of-00005.safetensors",
463
+ "llm.model.layers.9.post_feedforward_layernorm.weight": "model-00002-of-00005.safetensors",
464
+ "llm.model.layers.9.pre_feedforward_layernorm.weight": "model-00002-of-00005.safetensors",
465
+ "llm.model.layers.9.self_attn.k_proj.weight": "model-00002-of-00005.safetensors",
466
+ "llm.model.layers.9.self_attn.o_proj.weight": "model-00002-of-00005.safetensors",
467
+ "llm.model.layers.9.self_attn.q_proj.weight": "model-00002-of-00005.safetensors",
468
+ "llm.model.layers.9.self_attn.v_proj.weight": "model-00002-of-00005.safetensors",
469
+ "llm.model.norm.weight": "model-00004-of-00005.safetensors",
470
+ "visual_tokenizer.backbone.vision_model.embeddings.patch_embedding.bias": "model-00004-of-00005.safetensors",
471
+ "visual_tokenizer.backbone.vision_model.embeddings.patch_embedding.weight": "model-00004-of-00005.safetensors",
472
+ "visual_tokenizer.backbone.vision_model.embeddings.position_embedding.weight": "model-00004-of-00005.safetensors",
473
+ "visual_tokenizer.backbone.vision_model.encoder.layers.0.layer_norm1.bias": "model-00004-of-00005.safetensors",
474
+ "visual_tokenizer.backbone.vision_model.encoder.layers.0.layer_norm1.weight": "model-00004-of-00005.safetensors",
475
+ "visual_tokenizer.backbone.vision_model.encoder.layers.0.layer_norm2.bias": "model-00004-of-00005.safetensors",
476
+ "visual_tokenizer.backbone.vision_model.encoder.layers.0.layer_norm2.weight": "model-00004-of-00005.safetensors",
477
+ "visual_tokenizer.backbone.vision_model.encoder.layers.0.mlp.fc1.bias": "model-00004-of-00005.safetensors",
478
+ "visual_tokenizer.backbone.vision_model.encoder.layers.0.mlp.fc1.weight": "model-00004-of-00005.safetensors",
479
+ "visual_tokenizer.backbone.vision_model.encoder.layers.0.mlp.fc2.bias": "model-00004-of-00005.safetensors",
480
+ "visual_tokenizer.backbone.vision_model.encoder.layers.0.mlp.fc2.weight": "model-00004-of-00005.safetensors",
481
+ "visual_tokenizer.backbone.vision_model.encoder.layers.0.self_attn.k_proj.bias": "model-00004-of-00005.safetensors",
482
+ "visual_tokenizer.backbone.vision_model.encoder.layers.0.self_attn.k_proj.weight": "model-00004-of-00005.safetensors",
483
+ "visual_tokenizer.backbone.vision_model.encoder.layers.0.self_attn.out_proj.bias": "model-00004-of-00005.safetensors",
484
+ "visual_tokenizer.backbone.vision_model.encoder.layers.0.self_attn.out_proj.weight": "model-00004-of-00005.safetensors",
485
+ "visual_tokenizer.backbone.vision_model.encoder.layers.0.self_attn.q_proj.bias": "model-00004-of-00005.safetensors",
486
+ "visual_tokenizer.backbone.vision_model.encoder.layers.0.self_attn.q_proj.weight": "model-00004-of-00005.safetensors",
487
+ "visual_tokenizer.backbone.vision_model.encoder.layers.0.self_attn.v_proj.bias": "model-00004-of-00005.safetensors",
488
+ "visual_tokenizer.backbone.vision_model.encoder.layers.0.self_attn.v_proj.weight": "model-00004-of-00005.safetensors",
489
+ "visual_tokenizer.backbone.vision_model.encoder.layers.1.layer_norm1.bias": "model-00004-of-00005.safetensors",
490
+ "visual_tokenizer.backbone.vision_model.encoder.layers.1.layer_norm1.weight": "model-00004-of-00005.safetensors",
491
+ "visual_tokenizer.backbone.vision_model.encoder.layers.1.layer_norm2.bias": "model-00004-of-00005.safetensors",
492
+ "visual_tokenizer.backbone.vision_model.encoder.layers.1.layer_norm2.weight": "model-00004-of-00005.safetensors",
493
+ "visual_tokenizer.backbone.vision_model.encoder.layers.1.mlp.fc1.bias": "model-00004-of-00005.safetensors",
494
+ "visual_tokenizer.backbone.vision_model.encoder.layers.1.mlp.fc1.weight": "model-00004-of-00005.safetensors",
495
+ "visual_tokenizer.backbone.vision_model.encoder.layers.1.mlp.fc2.bias": "model-00004-of-00005.safetensors",
496
+ "visual_tokenizer.backbone.vision_model.encoder.layers.1.mlp.fc2.weight": "model-00004-of-00005.safetensors",
497
+ "visual_tokenizer.backbone.vision_model.encoder.layers.1.self_attn.k_proj.bias": "model-00004-of-00005.safetensors",
498
+ "visual_tokenizer.backbone.vision_model.encoder.layers.1.self_attn.k_proj.weight": "model-00004-of-00005.safetensors",
499
+ "visual_tokenizer.backbone.vision_model.encoder.layers.1.self_attn.out_proj.bias": "model-00004-of-00005.safetensors",
500
+ "visual_tokenizer.backbone.vision_model.encoder.layers.1.self_attn.out_proj.weight": "model-00004-of-00005.safetensors",
501
+ "visual_tokenizer.backbone.vision_model.encoder.layers.1.self_attn.q_proj.bias": "model-00004-of-00005.safetensors",
502
+ "visual_tokenizer.backbone.vision_model.encoder.layers.1.self_attn.q_proj.weight": "model-00004-of-00005.safetensors",
503
+ "visual_tokenizer.backbone.vision_model.encoder.layers.1.self_attn.v_proj.bias": "model-00004-of-00005.safetensors",
504
+ "visual_tokenizer.backbone.vision_model.encoder.layers.1.self_attn.v_proj.weight": "model-00004-of-00005.safetensors",
505
+ "visual_tokenizer.backbone.vision_model.encoder.layers.10.layer_norm1.bias": "model-00004-of-00005.safetensors",
506
+ "visual_tokenizer.backbone.vision_model.encoder.layers.10.layer_norm1.weight": "model-00004-of-00005.safetensors",
507
+ "visual_tokenizer.backbone.vision_model.encoder.layers.10.layer_norm2.bias": "model-00004-of-00005.safetensors",
508
+ "visual_tokenizer.backbone.vision_model.encoder.layers.10.layer_norm2.weight": "model-00004-of-00005.safetensors",
509
+ "visual_tokenizer.backbone.vision_model.encoder.layers.10.mlp.fc1.bias": "model-00004-of-00005.safetensors",
510
+ "visual_tokenizer.backbone.vision_model.encoder.layers.10.mlp.fc1.weight": "model-00004-of-00005.safetensors",
511
+ "visual_tokenizer.backbone.vision_model.encoder.layers.10.mlp.fc2.bias": "model-00004-of-00005.safetensors",
512
+ "visual_tokenizer.backbone.vision_model.encoder.layers.10.mlp.fc2.weight": "model-00004-of-00005.safetensors",
513
+ "visual_tokenizer.backbone.vision_model.encoder.layers.10.self_attn.k_proj.bias": "model-00004-of-00005.safetensors",
514
+ "visual_tokenizer.backbone.vision_model.encoder.layers.10.self_attn.k_proj.weight": "model-00004-of-00005.safetensors",
515
+ "visual_tokenizer.backbone.vision_model.encoder.layers.10.self_attn.out_proj.bias": "model-00004-of-00005.safetensors",
516
+ "visual_tokenizer.backbone.vision_model.encoder.layers.10.self_attn.out_proj.weight": "model-00004-of-00005.safetensors",
517
+ "visual_tokenizer.backbone.vision_model.encoder.layers.10.self_attn.q_proj.bias": "model-00004-of-00005.safetensors",
518
+ "visual_tokenizer.backbone.vision_model.encoder.layers.10.self_attn.q_proj.weight": "model-00004-of-00005.safetensors",
519
+ "visual_tokenizer.backbone.vision_model.encoder.layers.10.self_attn.v_proj.bias": "model-00004-of-00005.safetensors",
520
+ "visual_tokenizer.backbone.vision_model.encoder.layers.10.self_attn.v_proj.weight": "model-00004-of-00005.safetensors",
521
+ "visual_tokenizer.backbone.vision_model.encoder.layers.11.layer_norm1.bias": "model-00004-of-00005.safetensors",
522
+ "visual_tokenizer.backbone.vision_model.encoder.layers.11.layer_norm1.weight": "model-00004-of-00005.safetensors",
523
+ "visual_tokenizer.backbone.vision_model.encoder.layers.11.layer_norm2.bias": "model-00004-of-00005.safetensors",
524
+ "visual_tokenizer.backbone.vision_model.encoder.layers.11.layer_norm2.weight": "model-00004-of-00005.safetensors",
525
+ "visual_tokenizer.backbone.vision_model.encoder.layers.11.mlp.fc1.bias": "model-00004-of-00005.safetensors",
526
+ "visual_tokenizer.backbone.vision_model.encoder.layers.11.mlp.fc1.weight": "model-00004-of-00005.safetensors",
527
+ "visual_tokenizer.backbone.vision_model.encoder.layers.11.mlp.fc2.bias": "model-00004-of-00005.safetensors",
528
+ "visual_tokenizer.backbone.vision_model.encoder.layers.11.mlp.fc2.weight": "model-00004-of-00005.safetensors",
529
+ "visual_tokenizer.backbone.vision_model.encoder.layers.11.self_attn.k_proj.bias": "model-00004-of-00005.safetensors",
530
+ "visual_tokenizer.backbone.vision_model.encoder.layers.11.self_attn.k_proj.weight": "model-00004-of-00005.safetensors",
531
+ "visual_tokenizer.backbone.vision_model.encoder.layers.11.self_attn.out_proj.bias": "model-00004-of-00005.safetensors",
532
+ "visual_tokenizer.backbone.vision_model.encoder.layers.11.self_attn.out_proj.weight": "model-00004-of-00005.safetensors",
533
+ "visual_tokenizer.backbone.vision_model.encoder.layers.11.self_attn.q_proj.bias": "model-00004-of-00005.safetensors",
534
+ "visual_tokenizer.backbone.vision_model.encoder.layers.11.self_attn.q_proj.weight": "model-00004-of-00005.safetensors",
535
+ "visual_tokenizer.backbone.vision_model.encoder.layers.11.self_attn.v_proj.bias": "model-00004-of-00005.safetensors",
536
+ "visual_tokenizer.backbone.vision_model.encoder.layers.11.self_attn.v_proj.weight": "model-00004-of-00005.safetensors",
537
+ "visual_tokenizer.backbone.vision_model.encoder.layers.12.layer_norm1.bias": "model-00004-of-00005.safetensors",
538
+ "visual_tokenizer.backbone.vision_model.encoder.layers.12.layer_norm1.weight": "model-00004-of-00005.safetensors",
539
+ "visual_tokenizer.backbone.vision_model.encoder.layers.12.layer_norm2.bias": "model-00004-of-00005.safetensors",
540
+ "visual_tokenizer.backbone.vision_model.encoder.layers.12.layer_norm2.weight": "model-00004-of-00005.safetensors",
541
+ "visual_tokenizer.backbone.vision_model.encoder.layers.12.mlp.fc1.bias": "model-00004-of-00005.safetensors",
542
+ "visual_tokenizer.backbone.vision_model.encoder.layers.12.mlp.fc1.weight": "model-00004-of-00005.safetensors",
543
+ "visual_tokenizer.backbone.vision_model.encoder.layers.12.mlp.fc2.bias": "model-00004-of-00005.safetensors",
544
+ "visual_tokenizer.backbone.vision_model.encoder.layers.12.mlp.fc2.weight": "model-00004-of-00005.safetensors",
545
+ "visual_tokenizer.backbone.vision_model.encoder.layers.12.self_attn.k_proj.bias": "model-00004-of-00005.safetensors",
546
+ "visual_tokenizer.backbone.vision_model.encoder.layers.12.self_attn.k_proj.weight": "model-00004-of-00005.safetensors",
547
+ "visual_tokenizer.backbone.vision_model.encoder.layers.12.self_attn.out_proj.bias": "model-00004-of-00005.safetensors",
548
+ "visual_tokenizer.backbone.vision_model.encoder.layers.12.self_attn.out_proj.weight": "model-00004-of-00005.safetensors",
549
+ "visual_tokenizer.backbone.vision_model.encoder.layers.12.self_attn.q_proj.bias": "model-00004-of-00005.safetensors",
550
+ "visual_tokenizer.backbone.vision_model.encoder.layers.12.self_attn.q_proj.weight": "model-00004-of-00005.safetensors",
551
+ "visual_tokenizer.backbone.vision_model.encoder.layers.12.self_attn.v_proj.bias": "model-00004-of-00005.safetensors",
552
+ "visual_tokenizer.backbone.vision_model.encoder.layers.12.self_attn.v_proj.weight": "model-00004-of-00005.safetensors",
553
+ "visual_tokenizer.backbone.vision_model.encoder.layers.13.layer_norm1.bias": "model-00004-of-00005.safetensors",
554
+ "visual_tokenizer.backbone.vision_model.encoder.layers.13.layer_norm1.weight": "model-00004-of-00005.safetensors",
555
+ "visual_tokenizer.backbone.vision_model.encoder.layers.13.layer_norm2.bias": "model-00004-of-00005.safetensors",
556
+ "visual_tokenizer.backbone.vision_model.encoder.layers.13.layer_norm2.weight": "model-00004-of-00005.safetensors",
557
+ "visual_tokenizer.backbone.vision_model.encoder.layers.13.mlp.fc1.bias": "model-00004-of-00005.safetensors",
558
+ "visual_tokenizer.backbone.vision_model.encoder.layers.13.mlp.fc1.weight": "model-00004-of-00005.safetensors",
559
+ "visual_tokenizer.backbone.vision_model.encoder.layers.13.mlp.fc2.bias": "model-00004-of-00005.safetensors",
560
+ "visual_tokenizer.backbone.vision_model.encoder.layers.13.mlp.fc2.weight": "model-00004-of-00005.safetensors",
561
+ "visual_tokenizer.backbone.vision_model.encoder.layers.13.self_attn.k_proj.bias": "model-00004-of-00005.safetensors",
562
+ "visual_tokenizer.backbone.vision_model.encoder.layers.13.self_attn.k_proj.weight": "model-00004-of-00005.safetensors",
563
+ "visual_tokenizer.backbone.vision_model.encoder.layers.13.self_attn.out_proj.bias": "model-00004-of-00005.safetensors",
564
+ "visual_tokenizer.backbone.vision_model.encoder.layers.13.self_attn.out_proj.weight": "model-00004-of-00005.safetensors",
565
+ "visual_tokenizer.backbone.vision_model.encoder.layers.13.self_attn.q_proj.bias": "model-00004-of-00005.safetensors",
566
+ "visual_tokenizer.backbone.vision_model.encoder.layers.13.self_attn.q_proj.weight": "model-00004-of-00005.safetensors",
567
+ "visual_tokenizer.backbone.vision_model.encoder.layers.13.self_attn.v_proj.bias": "model-00004-of-00005.safetensors",
568
+ "visual_tokenizer.backbone.vision_model.encoder.layers.13.self_attn.v_proj.weight": "model-00004-of-00005.safetensors",
569
+ "visual_tokenizer.backbone.vision_model.encoder.layers.14.layer_norm1.bias": "model-00004-of-00005.safetensors",
570
+ "visual_tokenizer.backbone.vision_model.encoder.layers.14.layer_norm1.weight": "model-00004-of-00005.safetensors",
571
+ "visual_tokenizer.backbone.vision_model.encoder.layers.14.layer_norm2.bias": "model-00004-of-00005.safetensors",
572
+ "visual_tokenizer.backbone.vision_model.encoder.layers.14.layer_norm2.weight": "model-00004-of-00005.safetensors",
573
+ "visual_tokenizer.backbone.vision_model.encoder.layers.14.mlp.fc1.bias": "model-00004-of-00005.safetensors",
574
+ "visual_tokenizer.backbone.vision_model.encoder.layers.14.mlp.fc1.weight": "model-00004-of-00005.safetensors",
575
+ "visual_tokenizer.backbone.vision_model.encoder.layers.14.mlp.fc2.bias": "model-00004-of-00005.safetensors",
576
+ "visual_tokenizer.backbone.vision_model.encoder.layers.14.mlp.fc2.weight": "model-00004-of-00005.safetensors",
577
+ "visual_tokenizer.backbone.vision_model.encoder.layers.14.self_attn.k_proj.bias": "model-00004-of-00005.safetensors",
578
+ "visual_tokenizer.backbone.vision_model.encoder.layers.14.self_attn.k_proj.weight": "model-00004-of-00005.safetensors",
579
+ "visual_tokenizer.backbone.vision_model.encoder.layers.14.self_attn.out_proj.bias": "model-00004-of-00005.safetensors",
580
+ "visual_tokenizer.backbone.vision_model.encoder.layers.14.self_attn.out_proj.weight": "model-00004-of-00005.safetensors",
581
+ "visual_tokenizer.backbone.vision_model.encoder.layers.14.self_attn.q_proj.bias": "model-00004-of-00005.safetensors",
582
+ "visual_tokenizer.backbone.vision_model.encoder.layers.14.self_attn.q_proj.weight": "model-00004-of-00005.safetensors",
583
+ "visual_tokenizer.backbone.vision_model.encoder.layers.14.self_attn.v_proj.bias": "model-00004-of-00005.safetensors",
584
+ "visual_tokenizer.backbone.vision_model.encoder.layers.14.self_attn.v_proj.weight": "model-00004-of-00005.safetensors",
585
+ "visual_tokenizer.backbone.vision_model.encoder.layers.15.layer_norm1.bias": "model-00004-of-00005.safetensors",
586
+ "visual_tokenizer.backbone.vision_model.encoder.layers.15.layer_norm1.weight": "model-00004-of-00005.safetensors",
587
+ "visual_tokenizer.backbone.vision_model.encoder.layers.15.layer_norm2.bias": "model-00004-of-00005.safetensors",
588
+ "visual_tokenizer.backbone.vision_model.encoder.layers.15.layer_norm2.weight": "model-00004-of-00005.safetensors",
589
+ "visual_tokenizer.backbone.vision_model.encoder.layers.15.mlp.fc1.bias": "model-00004-of-00005.safetensors",
590
+ "visual_tokenizer.backbone.vision_model.encoder.layers.15.mlp.fc1.weight": "model-00004-of-00005.safetensors",
591
+ "visual_tokenizer.backbone.vision_model.encoder.layers.15.mlp.fc2.bias": "model-00004-of-00005.safetensors",
592
+ "visual_tokenizer.backbone.vision_model.encoder.layers.15.mlp.fc2.weight": "model-00004-of-00005.safetensors",
593
+ "visual_tokenizer.backbone.vision_model.encoder.layers.15.self_attn.k_proj.bias": "model-00004-of-00005.safetensors",
594
+ "visual_tokenizer.backbone.vision_model.encoder.layers.15.self_attn.k_proj.weight": "model-00004-of-00005.safetensors",
595
+ "visual_tokenizer.backbone.vision_model.encoder.layers.15.self_attn.out_proj.bias": "model-00004-of-00005.safetensors",
596
+ "visual_tokenizer.backbone.vision_model.encoder.layers.15.self_attn.out_proj.weight": "model-00004-of-00005.safetensors",
597
+ "visual_tokenizer.backbone.vision_model.encoder.layers.15.self_attn.q_proj.bias": "model-00004-of-00005.safetensors",
598
+ "visual_tokenizer.backbone.vision_model.encoder.layers.15.self_attn.q_proj.weight": "model-00004-of-00005.safetensors",
599
+ "visual_tokenizer.backbone.vision_model.encoder.layers.15.self_attn.v_proj.bias": "model-00004-of-00005.safetensors",
600
+ "visual_tokenizer.backbone.vision_model.encoder.layers.15.self_attn.v_proj.weight": "model-00004-of-00005.safetensors",
601
+ "visual_tokenizer.backbone.vision_model.encoder.layers.16.layer_norm1.bias": "model-00004-of-00005.safetensors",
602
+ "visual_tokenizer.backbone.vision_model.encoder.layers.16.layer_norm1.weight": "model-00004-of-00005.safetensors",
603
+ "visual_tokenizer.backbone.vision_model.encoder.layers.16.layer_norm2.bias": "model-00004-of-00005.safetensors",
604
+ "visual_tokenizer.backbone.vision_model.encoder.layers.16.layer_norm2.weight": "model-00004-of-00005.safetensors",
605
+ "visual_tokenizer.backbone.vision_model.encoder.layers.16.mlp.fc1.bias": "model-00004-of-00005.safetensors",
606
+ "visual_tokenizer.backbone.vision_model.encoder.layers.16.mlp.fc1.weight": "model-00004-of-00005.safetensors",
607
+ "visual_tokenizer.backbone.vision_model.encoder.layers.16.mlp.fc2.bias": "model-00004-of-00005.safetensors",
608
+ "visual_tokenizer.backbone.vision_model.encoder.layers.16.mlp.fc2.weight": "model-00004-of-00005.safetensors",
609
+ "visual_tokenizer.backbone.vision_model.encoder.layers.16.self_attn.k_proj.bias": "model-00004-of-00005.safetensors",
610
+ "visual_tokenizer.backbone.vision_model.encoder.layers.16.self_attn.k_proj.weight": "model-00004-of-00005.safetensors",
611
+ "visual_tokenizer.backbone.vision_model.encoder.layers.16.self_attn.out_proj.bias": "model-00004-of-00005.safetensors",
612
+ "visual_tokenizer.backbone.vision_model.encoder.layers.16.self_attn.out_proj.weight": "model-00004-of-00005.safetensors",
613
+ "visual_tokenizer.backbone.vision_model.encoder.layers.16.self_attn.q_proj.bias": "model-00004-of-00005.safetensors",
614
+ "visual_tokenizer.backbone.vision_model.encoder.layers.16.self_attn.q_proj.weight": "model-00004-of-00005.safetensors",
615
+ "visual_tokenizer.backbone.vision_model.encoder.layers.16.self_attn.v_proj.bias": "model-00004-of-00005.safetensors",
616
+ "visual_tokenizer.backbone.vision_model.encoder.layers.16.self_attn.v_proj.weight": "model-00004-of-00005.safetensors",
617
+ "visual_tokenizer.backbone.vision_model.encoder.layers.17.layer_norm1.bias": "model-00004-of-00005.safetensors",
618
+ "visual_tokenizer.backbone.vision_model.encoder.layers.17.layer_norm1.weight": "model-00004-of-00005.safetensors",
619
+ "visual_tokenizer.backbone.vision_model.encoder.layers.17.layer_norm2.bias": "model-00004-of-00005.safetensors",
620
+ "visual_tokenizer.backbone.vision_model.encoder.layers.17.layer_norm2.weight": "model-00004-of-00005.safetensors",
621
+ "visual_tokenizer.backbone.vision_model.encoder.layers.17.mlp.fc1.bias": "model-00004-of-00005.safetensors",
622
+ "visual_tokenizer.backbone.vision_model.encoder.layers.17.mlp.fc1.weight": "model-00004-of-00005.safetensors",
623
+ "visual_tokenizer.backbone.vision_model.encoder.layers.17.mlp.fc2.bias": "model-00004-of-00005.safetensors",
624
+ "visual_tokenizer.backbone.vision_model.encoder.layers.17.mlp.fc2.weight": "model-00004-of-00005.safetensors",
625
+ "visual_tokenizer.backbone.vision_model.encoder.layers.17.self_attn.k_proj.bias": "model-00004-of-00005.safetensors",
626
+ "visual_tokenizer.backbone.vision_model.encoder.layers.17.self_attn.k_proj.weight": "model-00004-of-00005.safetensors",
627
+ "visual_tokenizer.backbone.vision_model.encoder.layers.17.self_attn.out_proj.bias": "model-00004-of-00005.safetensors",
628
+ "visual_tokenizer.backbone.vision_model.encoder.layers.17.self_attn.out_proj.weight": "model-00004-of-00005.safetensors",
629
+ "visual_tokenizer.backbone.vision_model.encoder.layers.17.self_attn.q_proj.bias": "model-00004-of-00005.safetensors",
630
+ "visual_tokenizer.backbone.vision_model.encoder.layers.17.self_attn.q_proj.weight": "model-00004-of-00005.safetensors",
631
+ "visual_tokenizer.backbone.vision_model.encoder.layers.17.self_attn.v_proj.bias": "model-00004-of-00005.safetensors",
632
+ "visual_tokenizer.backbone.vision_model.encoder.layers.17.self_attn.v_proj.weight": "model-00004-of-00005.safetensors",
633
+ "visual_tokenizer.backbone.vision_model.encoder.layers.18.layer_norm1.bias": "model-00004-of-00005.safetensors",
634
+ "visual_tokenizer.backbone.vision_model.encoder.layers.18.layer_norm1.weight": "model-00004-of-00005.safetensors",
635
+ "visual_tokenizer.backbone.vision_model.encoder.layers.18.layer_norm2.bias": "model-00004-of-00005.safetensors",
636
+ "visual_tokenizer.backbone.vision_model.encoder.layers.18.layer_norm2.weight": "model-00004-of-00005.safetensors",
637
+ "visual_tokenizer.backbone.vision_model.encoder.layers.18.mlp.fc1.bias": "model-00004-of-00005.safetensors",
638
+ "visual_tokenizer.backbone.vision_model.encoder.layers.18.mlp.fc1.weight": "model-00004-of-00005.safetensors",
639
+ "visual_tokenizer.backbone.vision_model.encoder.layers.18.mlp.fc2.bias": "model-00004-of-00005.safetensors",
640
+ "visual_tokenizer.backbone.vision_model.encoder.layers.18.mlp.fc2.weight": "model-00004-of-00005.safetensors",
641
+ "visual_tokenizer.backbone.vision_model.encoder.layers.18.self_attn.k_proj.bias": "model-00004-of-00005.safetensors",
642
+ "visual_tokenizer.backbone.vision_model.encoder.layers.18.self_attn.k_proj.weight": "model-00004-of-00005.safetensors",
643
+ "visual_tokenizer.backbone.vision_model.encoder.layers.18.self_attn.out_proj.bias": "model-00004-of-00005.safetensors",
644
+ "visual_tokenizer.backbone.vision_model.encoder.layers.18.self_attn.out_proj.weight": "model-00004-of-00005.safetensors",
645
+ "visual_tokenizer.backbone.vision_model.encoder.layers.18.self_attn.q_proj.bias": "model-00004-of-00005.safetensors",
646
+ "visual_tokenizer.backbone.vision_model.encoder.layers.18.self_attn.q_proj.weight": "model-00004-of-00005.safetensors",
647
+ "visual_tokenizer.backbone.vision_model.encoder.layers.18.self_attn.v_proj.bias": "model-00004-of-00005.safetensors",
648
+ "visual_tokenizer.backbone.vision_model.encoder.layers.18.self_attn.v_proj.weight": "model-00004-of-00005.safetensors",
649
+ "visual_tokenizer.backbone.vision_model.encoder.layers.19.layer_norm1.bias": "model-00004-of-00005.safetensors",
650
+ "visual_tokenizer.backbone.vision_model.encoder.layers.19.layer_norm1.weight": "model-00004-of-00005.safetensors",
651
+ "visual_tokenizer.backbone.vision_model.encoder.layers.19.layer_norm2.bias": "model-00004-of-00005.safetensors",
652
+ "visual_tokenizer.backbone.vision_model.encoder.layers.19.layer_norm2.weight": "model-00004-of-00005.safetensors",
653
+ "visual_tokenizer.backbone.vision_model.encoder.layers.19.mlp.fc1.bias": "model-00004-of-00005.safetensors",
654
+ "visual_tokenizer.backbone.vision_model.encoder.layers.19.mlp.fc1.weight": "model-00004-of-00005.safetensors",
655
+ "visual_tokenizer.backbone.vision_model.encoder.layers.19.mlp.fc2.bias": "model-00004-of-00005.safetensors",
656
+ "visual_tokenizer.backbone.vision_model.encoder.layers.19.mlp.fc2.weight": "model-00004-of-00005.safetensors",
657
+ "visual_tokenizer.backbone.vision_model.encoder.layers.19.self_attn.k_proj.bias": "model-00004-of-00005.safetensors",
658
+ "visual_tokenizer.backbone.vision_model.encoder.layers.19.self_attn.k_proj.weight": "model-00004-of-00005.safetensors",
659
+ "visual_tokenizer.backbone.vision_model.encoder.layers.19.self_attn.out_proj.bias": "model-00004-of-00005.safetensors",
660
+ "visual_tokenizer.backbone.vision_model.encoder.layers.19.self_attn.out_proj.weight": "model-00004-of-00005.safetensors",
661
+ "visual_tokenizer.backbone.vision_model.encoder.layers.19.self_attn.q_proj.bias": "model-00004-of-00005.safetensors",
662
+ "visual_tokenizer.backbone.vision_model.encoder.layers.19.self_attn.q_proj.weight": "model-00004-of-00005.safetensors",
663
+ "visual_tokenizer.backbone.vision_model.encoder.layers.19.self_attn.v_proj.bias": "model-00004-of-00005.safetensors",
664
+ "visual_tokenizer.backbone.vision_model.encoder.layers.19.self_attn.v_proj.weight": "model-00004-of-00005.safetensors",
665
+ "visual_tokenizer.backbone.vision_model.encoder.layers.2.layer_norm1.bias": "model-00004-of-00005.safetensors",
666
+ "visual_tokenizer.backbone.vision_model.encoder.layers.2.layer_norm1.weight": "model-00004-of-00005.safetensors",
667
+ "visual_tokenizer.backbone.vision_model.encoder.layers.2.layer_norm2.bias": "model-00004-of-00005.safetensors",
668
+ "visual_tokenizer.backbone.vision_model.encoder.layers.2.layer_norm2.weight": "model-00004-of-00005.safetensors",
669
+ "visual_tokenizer.backbone.vision_model.encoder.layers.2.mlp.fc1.bias": "model-00004-of-00005.safetensors",
670
+ "visual_tokenizer.backbone.vision_model.encoder.layers.2.mlp.fc1.weight": "model-00004-of-00005.safetensors",
671
+ "visual_tokenizer.backbone.vision_model.encoder.layers.2.mlp.fc2.bias": "model-00004-of-00005.safetensors",
672
+ "visual_tokenizer.backbone.vision_model.encoder.layers.2.mlp.fc2.weight": "model-00004-of-00005.safetensors",
673
+ "visual_tokenizer.backbone.vision_model.encoder.layers.2.self_attn.k_proj.bias": "model-00004-of-00005.safetensors",
674
+ "visual_tokenizer.backbone.vision_model.encoder.layers.2.self_attn.k_proj.weight": "model-00004-of-00005.safetensors",
675
+ "visual_tokenizer.backbone.vision_model.encoder.layers.2.self_attn.out_proj.bias": "model-00004-of-00005.safetensors",
676
+ "visual_tokenizer.backbone.vision_model.encoder.layers.2.self_attn.out_proj.weight": "model-00004-of-00005.safetensors",
677
+ "visual_tokenizer.backbone.vision_model.encoder.layers.2.self_attn.q_proj.bias": "model-00004-of-00005.safetensors",
678
+ "visual_tokenizer.backbone.vision_model.encoder.layers.2.self_attn.q_proj.weight": "model-00004-of-00005.safetensors",
679
+ "visual_tokenizer.backbone.vision_model.encoder.layers.2.self_attn.v_proj.bias": "model-00004-of-00005.safetensors",
680
+ "visual_tokenizer.backbone.vision_model.encoder.layers.2.self_attn.v_proj.weight": "model-00004-of-00005.safetensors",
681
+ "visual_tokenizer.backbone.vision_model.encoder.layers.20.layer_norm1.bias": "model-00004-of-00005.safetensors",
682
+ "visual_tokenizer.backbone.vision_model.encoder.layers.20.layer_norm1.weight": "model-00004-of-00005.safetensors",
683
+ "visual_tokenizer.backbone.vision_model.encoder.layers.20.layer_norm2.bias": "model-00004-of-00005.safetensors",
684
+ "visual_tokenizer.backbone.vision_model.encoder.layers.20.layer_norm2.weight": "model-00004-of-00005.safetensors",
685
+ "visual_tokenizer.backbone.vision_model.encoder.layers.20.mlp.fc1.bias": "model-00004-of-00005.safetensors",
686
+ "visual_tokenizer.backbone.vision_model.encoder.layers.20.mlp.fc1.weight": "model-00004-of-00005.safetensors",
687
+ "visual_tokenizer.backbone.vision_model.encoder.layers.20.mlp.fc2.bias": "model-00004-of-00005.safetensors",
688
+ "visual_tokenizer.backbone.vision_model.encoder.layers.20.mlp.fc2.weight": "model-00004-of-00005.safetensors",
689
+ "visual_tokenizer.backbone.vision_model.encoder.layers.20.self_attn.k_proj.bias": "model-00004-of-00005.safetensors",
690
+ "visual_tokenizer.backbone.vision_model.encoder.layers.20.self_attn.k_proj.weight": "model-00004-of-00005.safetensors",
691
+ "visual_tokenizer.backbone.vision_model.encoder.layers.20.self_attn.out_proj.bias": "model-00004-of-00005.safetensors",
692
+ "visual_tokenizer.backbone.vision_model.encoder.layers.20.self_attn.out_proj.weight": "model-00004-of-00005.safetensors",
693
+ "visual_tokenizer.backbone.vision_model.encoder.layers.20.self_attn.q_proj.bias": "model-00004-of-00005.safetensors",
694
+ "visual_tokenizer.backbone.vision_model.encoder.layers.20.self_attn.q_proj.weight": "model-00004-of-00005.safetensors",
695
+ "visual_tokenizer.backbone.vision_model.encoder.layers.20.self_attn.v_proj.bias": "model-00004-of-00005.safetensors",
696
+ "visual_tokenizer.backbone.vision_model.encoder.layers.20.self_attn.v_proj.weight": "model-00004-of-00005.safetensors",
697
+ "visual_tokenizer.backbone.vision_model.encoder.layers.21.layer_norm1.bias": "model-00004-of-00005.safetensors",
698
+ "visual_tokenizer.backbone.vision_model.encoder.layers.21.layer_norm1.weight": "model-00004-of-00005.safetensors",
699
+ "visual_tokenizer.backbone.vision_model.encoder.layers.21.layer_norm2.bias": "model-00004-of-00005.safetensors",
700
+ "visual_tokenizer.backbone.vision_model.encoder.layers.21.layer_norm2.weight": "model-00004-of-00005.safetensors",
701
+ "visual_tokenizer.backbone.vision_model.encoder.layers.21.mlp.fc1.bias": "model-00004-of-00005.safetensors",
702
+ "visual_tokenizer.backbone.vision_model.encoder.layers.21.mlp.fc1.weight": "model-00004-of-00005.safetensors",
703
+ "visual_tokenizer.backbone.vision_model.encoder.layers.21.mlp.fc2.bias": "model-00004-of-00005.safetensors",
704
+ "visual_tokenizer.backbone.vision_model.encoder.layers.21.mlp.fc2.weight": "model-00004-of-00005.safetensors",
705
+ "visual_tokenizer.backbone.vision_model.encoder.layers.21.self_attn.k_proj.bias": "model-00004-of-00005.safetensors",
706
+ "visual_tokenizer.backbone.vision_model.encoder.layers.21.self_attn.k_proj.weight": "model-00004-of-00005.safetensors",
707
+ "visual_tokenizer.backbone.vision_model.encoder.layers.21.self_attn.out_proj.bias": "model-00004-of-00005.safetensors",
708
+ "visual_tokenizer.backbone.vision_model.encoder.layers.21.self_attn.out_proj.weight": "model-00004-of-00005.safetensors",
709
+ "visual_tokenizer.backbone.vision_model.encoder.layers.21.self_attn.q_proj.bias": "model-00004-of-00005.safetensors",
710
+ "visual_tokenizer.backbone.vision_model.encoder.layers.21.self_attn.q_proj.weight": "model-00004-of-00005.safetensors",
711
+ "visual_tokenizer.backbone.vision_model.encoder.layers.21.self_attn.v_proj.bias": "model-00004-of-00005.safetensors",
712
+ "visual_tokenizer.backbone.vision_model.encoder.layers.21.self_attn.v_proj.weight": "model-00004-of-00005.safetensors",
713
+ "visual_tokenizer.backbone.vision_model.encoder.layers.22.layer_norm1.bias": "model-00004-of-00005.safetensors",
714
+ "visual_tokenizer.backbone.vision_model.encoder.layers.22.layer_norm1.weight": "model-00004-of-00005.safetensors",
715
+ "visual_tokenizer.backbone.vision_model.encoder.layers.22.layer_norm2.bias": "model-00004-of-00005.safetensors",
716
+ "visual_tokenizer.backbone.vision_model.encoder.layers.22.layer_norm2.weight": "model-00004-of-00005.safetensors",
717
+ "visual_tokenizer.backbone.vision_model.encoder.layers.22.mlp.fc1.bias": "model-00004-of-00005.safetensors",
718
+ "visual_tokenizer.backbone.vision_model.encoder.layers.22.mlp.fc1.weight": "model-00004-of-00005.safetensors",
719
+ "visual_tokenizer.backbone.vision_model.encoder.layers.22.mlp.fc2.bias": "model-00004-of-00005.safetensors",
720
+ "visual_tokenizer.backbone.vision_model.encoder.layers.22.mlp.fc2.weight": "model-00004-of-00005.safetensors",
721
+ "visual_tokenizer.backbone.vision_model.encoder.layers.22.self_attn.k_proj.bias": "model-00004-of-00005.safetensors",
722
+ "visual_tokenizer.backbone.vision_model.encoder.layers.22.self_attn.k_proj.weight": "model-00004-of-00005.safetensors",
723
+ "visual_tokenizer.backbone.vision_model.encoder.layers.22.self_attn.out_proj.bias": "model-00004-of-00005.safetensors",
724
+ "visual_tokenizer.backbone.vision_model.encoder.layers.22.self_attn.out_proj.weight": "model-00004-of-00005.safetensors",
725
+ "visual_tokenizer.backbone.vision_model.encoder.layers.22.self_attn.q_proj.bias": "model-00004-of-00005.safetensors",
726
+ "visual_tokenizer.backbone.vision_model.encoder.layers.22.self_attn.q_proj.weight": "model-00004-of-00005.safetensors",
727
+ "visual_tokenizer.backbone.vision_model.encoder.layers.22.self_attn.v_proj.bias": "model-00004-of-00005.safetensors",
728
+ "visual_tokenizer.backbone.vision_model.encoder.layers.22.self_attn.v_proj.weight": "model-00004-of-00005.safetensors",
729
+ "visual_tokenizer.backbone.vision_model.encoder.layers.23.layer_norm1.bias": "model-00004-of-00005.safetensors",
730
+ "visual_tokenizer.backbone.vision_model.encoder.layers.23.layer_norm1.weight": "model-00004-of-00005.safetensors",
731
+ "visual_tokenizer.backbone.vision_model.encoder.layers.23.layer_norm2.bias": "model-00004-of-00005.safetensors",
732
+ "visual_tokenizer.backbone.vision_model.encoder.layers.23.layer_norm2.weight": "model-00004-of-00005.safetensors",
733
+ "visual_tokenizer.backbone.vision_model.encoder.layers.23.mlp.fc1.bias": "model-00004-of-00005.safetensors",
734
+ "visual_tokenizer.backbone.vision_model.encoder.layers.23.mlp.fc1.weight": "model-00004-of-00005.safetensors",
735
+ "visual_tokenizer.backbone.vision_model.encoder.layers.23.mlp.fc2.bias": "model-00004-of-00005.safetensors",
736
+ "visual_tokenizer.backbone.vision_model.encoder.layers.23.mlp.fc2.weight": "model-00004-of-00005.safetensors",
737
+ "visual_tokenizer.backbone.vision_model.encoder.layers.23.self_attn.k_proj.bias": "model-00004-of-00005.safetensors",
738
+ "visual_tokenizer.backbone.vision_model.encoder.layers.23.self_attn.k_proj.weight": "model-00004-of-00005.safetensors",
739
+ "visual_tokenizer.backbone.vision_model.encoder.layers.23.self_attn.out_proj.bias": "model-00004-of-00005.safetensors",
740
+ "visual_tokenizer.backbone.vision_model.encoder.layers.23.self_attn.out_proj.weight": "model-00004-of-00005.safetensors",
741
+ "visual_tokenizer.backbone.vision_model.encoder.layers.23.self_attn.q_proj.bias": "model-00004-of-00005.safetensors",
742
+ "visual_tokenizer.backbone.vision_model.encoder.layers.23.self_attn.q_proj.weight": "model-00004-of-00005.safetensors",
743
+ "visual_tokenizer.backbone.vision_model.encoder.layers.23.self_attn.v_proj.bias": "model-00004-of-00005.safetensors",
744
+ "visual_tokenizer.backbone.vision_model.encoder.layers.23.self_attn.v_proj.weight": "model-00004-of-00005.safetensors",
745
+ "visual_tokenizer.backbone.vision_model.encoder.layers.24.layer_norm1.bias": "model-00004-of-00005.safetensors",
746
+ "visual_tokenizer.backbone.vision_model.encoder.layers.24.layer_norm1.weight": "model-00004-of-00005.safetensors",
747
+ "visual_tokenizer.backbone.vision_model.encoder.layers.24.layer_norm2.bias": "model-00004-of-00005.safetensors",
748
+ "visual_tokenizer.backbone.vision_model.encoder.layers.24.layer_norm2.weight": "model-00004-of-00005.safetensors",
749
+ "visual_tokenizer.backbone.vision_model.encoder.layers.24.mlp.fc1.bias": "model-00004-of-00005.safetensors",
750
+ "visual_tokenizer.backbone.vision_model.encoder.layers.24.mlp.fc1.weight": "model-00004-of-00005.safetensors",
751
+ "visual_tokenizer.backbone.vision_model.encoder.layers.24.mlp.fc2.bias": "model-00004-of-00005.safetensors",
752
+ "visual_tokenizer.backbone.vision_model.encoder.layers.24.mlp.fc2.weight": "model-00004-of-00005.safetensors",
753
+ "visual_tokenizer.backbone.vision_model.encoder.layers.24.self_attn.k_proj.bias": "model-00004-of-00005.safetensors",
754
+ "visual_tokenizer.backbone.vision_model.encoder.layers.24.self_attn.k_proj.weight": "model-00004-of-00005.safetensors",
755
+ "visual_tokenizer.backbone.vision_model.encoder.layers.24.self_attn.out_proj.bias": "model-00004-of-00005.safetensors",
756
+ "visual_tokenizer.backbone.vision_model.encoder.layers.24.self_attn.out_proj.weight": "model-00004-of-00005.safetensors",
757
+ "visual_tokenizer.backbone.vision_model.encoder.layers.24.self_attn.q_proj.bias": "model-00004-of-00005.safetensors",
758
+ "visual_tokenizer.backbone.vision_model.encoder.layers.24.self_attn.q_proj.weight": "model-00004-of-00005.safetensors",
759
+ "visual_tokenizer.backbone.vision_model.encoder.layers.24.self_attn.v_proj.bias": "model-00004-of-00005.safetensors",
760
+ "visual_tokenizer.backbone.vision_model.encoder.layers.24.self_attn.v_proj.weight": "model-00004-of-00005.safetensors",
761
+ "visual_tokenizer.backbone.vision_model.encoder.layers.25.layer_norm1.bias": "model-00004-of-00005.safetensors",
762
+ "visual_tokenizer.backbone.vision_model.encoder.layers.25.layer_norm1.weight": "model-00004-of-00005.safetensors",
763
+ "visual_tokenizer.backbone.vision_model.encoder.layers.25.layer_norm2.bias": "model-00004-of-00005.safetensors",
764
+ "visual_tokenizer.backbone.vision_model.encoder.layers.25.layer_norm2.weight": "model-00004-of-00005.safetensors",
765
+ "visual_tokenizer.backbone.vision_model.encoder.layers.25.mlp.fc1.bias": "model-00004-of-00005.safetensors",
766
+ "visual_tokenizer.backbone.vision_model.encoder.layers.25.mlp.fc1.weight": "model-00004-of-00005.safetensors",
767
+ "visual_tokenizer.backbone.vision_model.encoder.layers.25.mlp.fc2.bias": "model-00004-of-00005.safetensors",
768
+ "visual_tokenizer.backbone.vision_model.encoder.layers.25.mlp.fc2.weight": "model-00004-of-00005.safetensors",
769
+ "visual_tokenizer.backbone.vision_model.encoder.layers.25.self_attn.k_proj.bias": "model-00004-of-00005.safetensors",
770
+ "visual_tokenizer.backbone.vision_model.encoder.layers.25.self_attn.k_proj.weight": "model-00004-of-00005.safetensors",
771
+ "visual_tokenizer.backbone.vision_model.encoder.layers.25.self_attn.out_proj.bias": "model-00004-of-00005.safetensors",
772
+ "visual_tokenizer.backbone.vision_model.encoder.layers.25.self_attn.out_proj.weight": "model-00004-of-00005.safetensors",
773
+ "visual_tokenizer.backbone.vision_model.encoder.layers.25.self_attn.q_proj.bias": "model-00004-of-00005.safetensors",
774
+ "visual_tokenizer.backbone.vision_model.encoder.layers.25.self_attn.q_proj.weight": "model-00004-of-00005.safetensors",
775
+ "visual_tokenizer.backbone.vision_model.encoder.layers.25.self_attn.v_proj.bias": "model-00004-of-00005.safetensors",
776
+ "visual_tokenizer.backbone.vision_model.encoder.layers.25.self_attn.v_proj.weight": "model-00004-of-00005.safetensors",
777
+ "visual_tokenizer.backbone.vision_model.encoder.layers.26.layer_norm1.bias": "model-00004-of-00005.safetensors",
778
+ "visual_tokenizer.backbone.vision_model.encoder.layers.26.layer_norm1.weight": "model-00004-of-00005.safetensors",
779
+ "visual_tokenizer.backbone.vision_model.encoder.layers.26.layer_norm2.bias": "model-00004-of-00005.safetensors",
780
+ "visual_tokenizer.backbone.vision_model.encoder.layers.26.layer_norm2.weight": "model-00004-of-00005.safetensors",
781
+ "visual_tokenizer.backbone.vision_model.encoder.layers.26.mlp.fc1.bias": "model-00004-of-00005.safetensors",
782
+ "visual_tokenizer.backbone.vision_model.encoder.layers.26.mlp.fc1.weight": "model-00004-of-00005.safetensors",
783
+ "visual_tokenizer.backbone.vision_model.encoder.layers.26.mlp.fc2.bias": "model-00004-of-00005.safetensors",
784
+ "visual_tokenizer.backbone.vision_model.encoder.layers.26.mlp.fc2.weight": "model-00004-of-00005.safetensors",
785
+ "visual_tokenizer.backbone.vision_model.encoder.layers.26.self_attn.k_proj.bias": "model-00004-of-00005.safetensors",
786
+ "visual_tokenizer.backbone.vision_model.encoder.layers.26.self_attn.k_proj.weight": "model-00004-of-00005.safetensors",
787
+ "visual_tokenizer.backbone.vision_model.encoder.layers.26.self_attn.out_proj.bias": "model-00004-of-00005.safetensors",
788
+ "visual_tokenizer.backbone.vision_model.encoder.layers.26.self_attn.out_proj.weight": "model-00004-of-00005.safetensors",
789
+ "visual_tokenizer.backbone.vision_model.encoder.layers.26.self_attn.q_proj.bias": "model-00004-of-00005.safetensors",
790
+ "visual_tokenizer.backbone.vision_model.encoder.layers.26.self_attn.q_proj.weight": "model-00004-of-00005.safetensors",
791
+ "visual_tokenizer.backbone.vision_model.encoder.layers.26.self_attn.v_proj.bias": "model-00004-of-00005.safetensors",
792
+ "visual_tokenizer.backbone.vision_model.encoder.layers.26.self_attn.v_proj.weight": "model-00004-of-00005.safetensors",
793
+ "visual_tokenizer.backbone.vision_model.encoder.layers.3.layer_norm1.bias": "model-00004-of-00005.safetensors",
794
+ "visual_tokenizer.backbone.vision_model.encoder.layers.3.layer_norm1.weight": "model-00004-of-00005.safetensors",
795
+ "visual_tokenizer.backbone.vision_model.encoder.layers.3.layer_norm2.bias": "model-00004-of-00005.safetensors",
796
+ "visual_tokenizer.backbone.vision_model.encoder.layers.3.layer_norm2.weight": "model-00004-of-00005.safetensors",
797
+ "visual_tokenizer.backbone.vision_model.encoder.layers.3.mlp.fc1.bias": "model-00004-of-00005.safetensors",
798
+ "visual_tokenizer.backbone.vision_model.encoder.layers.3.mlp.fc1.weight": "model-00004-of-00005.safetensors",
799
+ "visual_tokenizer.backbone.vision_model.encoder.layers.3.mlp.fc2.bias": "model-00004-of-00005.safetensors",
800
+ "visual_tokenizer.backbone.vision_model.encoder.layers.3.mlp.fc2.weight": "model-00004-of-00005.safetensors",
801
+ "visual_tokenizer.backbone.vision_model.encoder.layers.3.self_attn.k_proj.bias": "model-00004-of-00005.safetensors",
802
+ "visual_tokenizer.backbone.vision_model.encoder.layers.3.self_attn.k_proj.weight": "model-00004-of-00005.safetensors",
803
+ "visual_tokenizer.backbone.vision_model.encoder.layers.3.self_attn.out_proj.bias": "model-00004-of-00005.safetensors",
804
+ "visual_tokenizer.backbone.vision_model.encoder.layers.3.self_attn.out_proj.weight": "model-00004-of-00005.safetensors",
805
+ "visual_tokenizer.backbone.vision_model.encoder.layers.3.self_attn.q_proj.bias": "model-00004-of-00005.safetensors",
806
+ "visual_tokenizer.backbone.vision_model.encoder.layers.3.self_attn.q_proj.weight": "model-00004-of-00005.safetensors",
807
+ "visual_tokenizer.backbone.vision_model.encoder.layers.3.self_attn.v_proj.bias": "model-00004-of-00005.safetensors",
808
+ "visual_tokenizer.backbone.vision_model.encoder.layers.3.self_attn.v_proj.weight": "model-00004-of-00005.safetensors",
809
+ "visual_tokenizer.backbone.vision_model.encoder.layers.4.layer_norm1.bias": "model-00004-of-00005.safetensors",
810
+ "visual_tokenizer.backbone.vision_model.encoder.layers.4.layer_norm1.weight": "model-00004-of-00005.safetensors",
811
+ "visual_tokenizer.backbone.vision_model.encoder.layers.4.layer_norm2.bias": "model-00004-of-00005.safetensors",
812
+ "visual_tokenizer.backbone.vision_model.encoder.layers.4.layer_norm2.weight": "model-00004-of-00005.safetensors",
813
+ "visual_tokenizer.backbone.vision_model.encoder.layers.4.mlp.fc1.bias": "model-00004-of-00005.safetensors",
814
+ "visual_tokenizer.backbone.vision_model.encoder.layers.4.mlp.fc1.weight": "model-00004-of-00005.safetensors",
815
+ "visual_tokenizer.backbone.vision_model.encoder.layers.4.mlp.fc2.bias": "model-00004-of-00005.safetensors",
816
+ "visual_tokenizer.backbone.vision_model.encoder.layers.4.mlp.fc2.weight": "model-00004-of-00005.safetensors",
817
+ "visual_tokenizer.backbone.vision_model.encoder.layers.4.self_attn.k_proj.bias": "model-00004-of-00005.safetensors",
818
+ "visual_tokenizer.backbone.vision_model.encoder.layers.4.self_attn.k_proj.weight": "model-00004-of-00005.safetensors",
819
+ "visual_tokenizer.backbone.vision_model.encoder.layers.4.self_attn.out_proj.bias": "model-00004-of-00005.safetensors",
820
+ "visual_tokenizer.backbone.vision_model.encoder.layers.4.self_attn.out_proj.weight": "model-00004-of-00005.safetensors",
821
+ "visual_tokenizer.backbone.vision_model.encoder.layers.4.self_attn.q_proj.bias": "model-00004-of-00005.safetensors",
822
+ "visual_tokenizer.backbone.vision_model.encoder.layers.4.self_attn.q_proj.weight": "model-00004-of-00005.safetensors",
823
+ "visual_tokenizer.backbone.vision_model.encoder.layers.4.self_attn.v_proj.bias": "model-00004-of-00005.safetensors",
824
+ "visual_tokenizer.backbone.vision_model.encoder.layers.4.self_attn.v_proj.weight": "model-00004-of-00005.safetensors",
825
+ "visual_tokenizer.backbone.vision_model.encoder.layers.5.layer_norm1.bias": "model-00004-of-00005.safetensors",
826
+ "visual_tokenizer.backbone.vision_model.encoder.layers.5.layer_norm1.weight": "model-00004-of-00005.safetensors",
827
+ "visual_tokenizer.backbone.vision_model.encoder.layers.5.layer_norm2.bias": "model-00004-of-00005.safetensors",
828
+ "visual_tokenizer.backbone.vision_model.encoder.layers.5.layer_norm2.weight": "model-00004-of-00005.safetensors",
829
+ "visual_tokenizer.backbone.vision_model.encoder.layers.5.mlp.fc1.bias": "model-00004-of-00005.safetensors",
830
+ "visual_tokenizer.backbone.vision_model.encoder.layers.5.mlp.fc1.weight": "model-00004-of-00005.safetensors",
831
+ "visual_tokenizer.backbone.vision_model.encoder.layers.5.mlp.fc2.bias": "model-00004-of-00005.safetensors",
832
+ "visual_tokenizer.backbone.vision_model.encoder.layers.5.mlp.fc2.weight": "model-00004-of-00005.safetensors",
833
+ "visual_tokenizer.backbone.vision_model.encoder.layers.5.self_attn.k_proj.bias": "model-00004-of-00005.safetensors",
834
+ "visual_tokenizer.backbone.vision_model.encoder.layers.5.self_attn.k_proj.weight": "model-00004-of-00005.safetensors",
835
+ "visual_tokenizer.backbone.vision_model.encoder.layers.5.self_attn.out_proj.bias": "model-00004-of-00005.safetensors",
836
+ "visual_tokenizer.backbone.vision_model.encoder.layers.5.self_attn.out_proj.weight": "model-00004-of-00005.safetensors",
837
+ "visual_tokenizer.backbone.vision_model.encoder.layers.5.self_attn.q_proj.bias": "model-00004-of-00005.safetensors",
838
+ "visual_tokenizer.backbone.vision_model.encoder.layers.5.self_attn.q_proj.weight": "model-00004-of-00005.safetensors",
839
+ "visual_tokenizer.backbone.vision_model.encoder.layers.5.self_attn.v_proj.bias": "model-00004-of-00005.safetensors",
840
+ "visual_tokenizer.backbone.vision_model.encoder.layers.5.self_attn.v_proj.weight": "model-00004-of-00005.safetensors",
841
+ "visual_tokenizer.backbone.vision_model.encoder.layers.6.layer_norm1.bias": "model-00004-of-00005.safetensors",
842
+ "visual_tokenizer.backbone.vision_model.encoder.layers.6.layer_norm1.weight": "model-00004-of-00005.safetensors",
843
+ "visual_tokenizer.backbone.vision_model.encoder.layers.6.layer_norm2.bias": "model-00004-of-00005.safetensors",
844
+ "visual_tokenizer.backbone.vision_model.encoder.layers.6.layer_norm2.weight": "model-00004-of-00005.safetensors",
845
+ "visual_tokenizer.backbone.vision_model.encoder.layers.6.mlp.fc1.bias": "model-00004-of-00005.safetensors",
846
+ "visual_tokenizer.backbone.vision_model.encoder.layers.6.mlp.fc1.weight": "model-00004-of-00005.safetensors",
847
+ "visual_tokenizer.backbone.vision_model.encoder.layers.6.mlp.fc2.bias": "model-00004-of-00005.safetensors",
848
+ "visual_tokenizer.backbone.vision_model.encoder.layers.6.mlp.fc2.weight": "model-00004-of-00005.safetensors",
849
+ "visual_tokenizer.backbone.vision_model.encoder.layers.6.self_attn.k_proj.bias": "model-00004-of-00005.safetensors",
850
+ "visual_tokenizer.backbone.vision_model.encoder.layers.6.self_attn.k_proj.weight": "model-00004-of-00005.safetensors",
851
+ "visual_tokenizer.backbone.vision_model.encoder.layers.6.self_attn.out_proj.bias": "model-00004-of-00005.safetensors",
852
+ "visual_tokenizer.backbone.vision_model.encoder.layers.6.self_attn.out_proj.weight": "model-00004-of-00005.safetensors",
853
+ "visual_tokenizer.backbone.vision_model.encoder.layers.6.self_attn.q_proj.bias": "model-00004-of-00005.safetensors",
854
+ "visual_tokenizer.backbone.vision_model.encoder.layers.6.self_attn.q_proj.weight": "model-00004-of-00005.safetensors",
855
+ "visual_tokenizer.backbone.vision_model.encoder.layers.6.self_attn.v_proj.bias": "model-00004-of-00005.safetensors",
856
+ "visual_tokenizer.backbone.vision_model.encoder.layers.6.self_attn.v_proj.weight": "model-00004-of-00005.safetensors",
857
+ "visual_tokenizer.backbone.vision_model.encoder.layers.7.layer_norm1.bias": "model-00004-of-00005.safetensors",
858
+ "visual_tokenizer.backbone.vision_model.encoder.layers.7.layer_norm1.weight": "model-00004-of-00005.safetensors",
859
+ "visual_tokenizer.backbone.vision_model.encoder.layers.7.layer_norm2.bias": "model-00004-of-00005.safetensors",
860
+ "visual_tokenizer.backbone.vision_model.encoder.layers.7.layer_norm2.weight": "model-00004-of-00005.safetensors",
861
+ "visual_tokenizer.backbone.vision_model.encoder.layers.7.mlp.fc1.bias": "model-00004-of-00005.safetensors",
862
+ "visual_tokenizer.backbone.vision_model.encoder.layers.7.mlp.fc1.weight": "model-00004-of-00005.safetensors",
863
+ "visual_tokenizer.backbone.vision_model.encoder.layers.7.mlp.fc2.bias": "model-00004-of-00005.safetensors",
864
+ "visual_tokenizer.backbone.vision_model.encoder.layers.7.mlp.fc2.weight": "model-00004-of-00005.safetensors",
865
+ "visual_tokenizer.backbone.vision_model.encoder.layers.7.self_attn.k_proj.bias": "model-00004-of-00005.safetensors",
866
+ "visual_tokenizer.backbone.vision_model.encoder.layers.7.self_attn.k_proj.weight": "model-00004-of-00005.safetensors",
867
+ "visual_tokenizer.backbone.vision_model.encoder.layers.7.self_attn.out_proj.bias": "model-00004-of-00005.safetensors",
868
+ "visual_tokenizer.backbone.vision_model.encoder.layers.7.self_attn.out_proj.weight": "model-00004-of-00005.safetensors",
869
+ "visual_tokenizer.backbone.vision_model.encoder.layers.7.self_attn.q_proj.bias": "model-00004-of-00005.safetensors",
870
+ "visual_tokenizer.backbone.vision_model.encoder.layers.7.self_attn.q_proj.weight": "model-00004-of-00005.safetensors",
871
+ "visual_tokenizer.backbone.vision_model.encoder.layers.7.self_attn.v_proj.bias": "model-00004-of-00005.safetensors",
872
+ "visual_tokenizer.backbone.vision_model.encoder.layers.7.self_attn.v_proj.weight": "model-00004-of-00005.safetensors",
873
+ "visual_tokenizer.backbone.vision_model.encoder.layers.8.layer_norm1.bias": "model-00004-of-00005.safetensors",
874
+ "visual_tokenizer.backbone.vision_model.encoder.layers.8.layer_norm1.weight": "model-00004-of-00005.safetensors",
875
+ "visual_tokenizer.backbone.vision_model.encoder.layers.8.layer_norm2.bias": "model-00004-of-00005.safetensors",
876
+ "visual_tokenizer.backbone.vision_model.encoder.layers.8.layer_norm2.weight": "model-00004-of-00005.safetensors",
877
+ "visual_tokenizer.backbone.vision_model.encoder.layers.8.mlp.fc1.bias": "model-00004-of-00005.safetensors",
878
+ "visual_tokenizer.backbone.vision_model.encoder.layers.8.mlp.fc1.weight": "model-00004-of-00005.safetensors",
879
+ "visual_tokenizer.backbone.vision_model.encoder.layers.8.mlp.fc2.bias": "model-00004-of-00005.safetensors",
880
+ "visual_tokenizer.backbone.vision_model.encoder.layers.8.mlp.fc2.weight": "model-00004-of-00005.safetensors",
881
+ "visual_tokenizer.backbone.vision_model.encoder.layers.8.self_attn.k_proj.bias": "model-00004-of-00005.safetensors",
882
+ "visual_tokenizer.backbone.vision_model.encoder.layers.8.self_attn.k_proj.weight": "model-00004-of-00005.safetensors",
883
+ "visual_tokenizer.backbone.vision_model.encoder.layers.8.self_attn.out_proj.bias": "model-00004-of-00005.safetensors",
884
+ "visual_tokenizer.backbone.vision_model.encoder.layers.8.self_attn.out_proj.weight": "model-00004-of-00005.safetensors",
885
+ "visual_tokenizer.backbone.vision_model.encoder.layers.8.self_attn.q_proj.bias": "model-00004-of-00005.safetensors",
886
+ "visual_tokenizer.backbone.vision_model.encoder.layers.8.self_attn.q_proj.weight": "model-00004-of-00005.safetensors",
887
+ "visual_tokenizer.backbone.vision_model.encoder.layers.8.self_attn.v_proj.bias": "model-00004-of-00005.safetensors",
888
+ "visual_tokenizer.backbone.vision_model.encoder.layers.8.self_attn.v_proj.weight": "model-00004-of-00005.safetensors",
889
+ "visual_tokenizer.backbone.vision_model.encoder.layers.9.layer_norm1.bias": "model-00004-of-00005.safetensors",
890
+ "visual_tokenizer.backbone.vision_model.encoder.layers.9.layer_norm1.weight": "model-00004-of-00005.safetensors",
891
+ "visual_tokenizer.backbone.vision_model.encoder.layers.9.layer_norm2.bias": "model-00004-of-00005.safetensors",
892
+ "visual_tokenizer.backbone.vision_model.encoder.layers.9.layer_norm2.weight": "model-00004-of-00005.safetensors",
893
+ "visual_tokenizer.backbone.vision_model.encoder.layers.9.mlp.fc1.bias": "model-00004-of-00005.safetensors",
894
+ "visual_tokenizer.backbone.vision_model.encoder.layers.9.mlp.fc1.weight": "model-00004-of-00005.safetensors",
895
+ "visual_tokenizer.backbone.vision_model.encoder.layers.9.mlp.fc2.bias": "model-00004-of-00005.safetensors",
896
+ "visual_tokenizer.backbone.vision_model.encoder.layers.9.mlp.fc2.weight": "model-00004-of-00005.safetensors",
897
+ "visual_tokenizer.backbone.vision_model.encoder.layers.9.self_attn.k_proj.bias": "model-00004-of-00005.safetensors",
898
+ "visual_tokenizer.backbone.vision_model.encoder.layers.9.self_attn.k_proj.weight": "model-00004-of-00005.safetensors",
899
+ "visual_tokenizer.backbone.vision_model.encoder.layers.9.self_attn.out_proj.bias": "model-00004-of-00005.safetensors",
900
+ "visual_tokenizer.backbone.vision_model.encoder.layers.9.self_attn.out_proj.weight": "model-00004-of-00005.safetensors",
901
+ "visual_tokenizer.backbone.vision_model.encoder.layers.9.self_attn.q_proj.bias": "model-00004-of-00005.safetensors",
902
+ "visual_tokenizer.backbone.vision_model.encoder.layers.9.self_attn.q_proj.weight": "model-00004-of-00005.safetensors",
903
+ "visual_tokenizer.backbone.vision_model.encoder.layers.9.self_attn.v_proj.bias": "model-00004-of-00005.safetensors",
904
+ "visual_tokenizer.backbone.vision_model.encoder.layers.9.self_attn.v_proj.weight": "model-00004-of-00005.safetensors",
905
+ "visual_tokenizer.backbone.vision_model.head.attention.in_proj_bias": "model-00004-of-00005.safetensors",
906
+ "visual_tokenizer.backbone.vision_model.head.attention.in_proj_weight": "model-00004-of-00005.safetensors",
907
+ "visual_tokenizer.backbone.vision_model.head.attention.out_proj.bias": "model-00004-of-00005.safetensors",
908
+ "visual_tokenizer.backbone.vision_model.head.attention.out_proj.weight": "model-00004-of-00005.safetensors",
909
+ "visual_tokenizer.backbone.vision_model.head.layernorm.bias": "model-00004-of-00005.safetensors",
910
+ "visual_tokenizer.backbone.vision_model.head.layernorm.weight": "model-00004-of-00005.safetensors",
911
+ "visual_tokenizer.backbone.vision_model.head.mlp.fc1.bias": "model-00004-of-00005.safetensors",
912
+ "visual_tokenizer.backbone.vision_model.head.mlp.fc1.weight": "model-00004-of-00005.safetensors",
913
+ "visual_tokenizer.backbone.vision_model.head.mlp.fc2.bias": "model-00004-of-00005.safetensors",
914
+ "visual_tokenizer.backbone.vision_model.head.mlp.fc2.weight": "model-00004-of-00005.safetensors",
915
+ "visual_tokenizer.backbone.vision_model.head.probe": "model-00004-of-00005.safetensors",
916
+ "visual_tokenizer.backbone.vision_model.post_layernorm.bias": "model-00004-of-00005.safetensors",
917
+ "visual_tokenizer.backbone.vision_model.post_layernorm.weight": "model-00004-of-00005.safetensors",
918
+ "visual_tokenizer.head.0.weight": "model-00005-of-00005.safetensors",
919
+ "visual_tokenizer.head.1.bias": "model-00005-of-00005.safetensors",
920
+ "visual_tokenizer.head.1.weight": "model-00005-of-00005.safetensors",
921
+ "vte.weight": "model-00005-of-00005.safetensors"
922
+ }
923
+ }
modeling_ovis.py ADDED
@@ -0,0 +1,596 @@
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
1
+ import logging
2
+ import os
3
+ from importlib import import_module
4
+ from typing import List, Callable, Union, Optional, Dict
5
+
6
+ import PIL.Image
7
+ import torch
8
+ from torch import Tensor
9
+ from torch.nn import init
10
+ from torch.nn.functional import softmax, gumbel_softmax, pad
11
+ from transformers import PreTrainedModel, AutoModel, AutoTokenizer, AutoModelForCausalLM, AutoImageProcessor
12
+ from transformers import SiglipImageProcessor, SiglipVisionModel
13
+ from transformers.cache_utils import HybridCache
14
+ from transformers.generation.utils import GenerateOutput
15
+
16
+ from .configuration_ovis import BaseVisualTokenizerConfig, SiglipVisualTokenizerConfig
17
+ from .configuration_ovis import OvisConfig, ConversationFormatter
18
+ from .configuration_ovis import IGNORE_ID, IMAGE_ATOM_ID, IMAGE_INDICATOR_IDS, IMAGE_TOKEN_ID
19
+
20
+
21
+ # ----------------------------------------------------------------------
22
+ # Visual Tokenizer
23
+ # ----------------------------------------------------------------------
24
+ class BaseVisualTokenizer(PreTrainedModel):
25
+ base_model_prefix = "backbone"
26
+ main_input_name = None
27
+ _image_processor_class = None
28
+ _image_processor_kwargs = {}
29
+ _backbone_class = None
30
+ _backbone_name_or_path = None
31
+
32
+ def __init__(self, config: BaseVisualTokenizerConfig, *inputs, **kwargs):
33
+ super().__init__(config, *inputs, **kwargs)
34
+ self.image_processor = AutoImageProcessor.from_pretrained(kwargs['image_processor_name_or_path'])
35
+ self.backbone = AutoModel.from_config(self.config.backbone_config)
36
+ head_dim = self.config.vocab_size - len(IMAGE_INDICATOR_IDS) # reserved tokens for IMAGE_INDICATORS
37
+ self.head = torch.nn.Sequential(
38
+ torch.nn.Linear(
39
+ self.backbone.config.hidden_size * self.config.hidden_stride * self.config.hidden_stride, head_dim,
40
+ bias=False
41
+ ),
42
+ torch.nn.LayerNorm(head_dim)
43
+ )
44
+
45
+ assert all((self.image_processor.do_resize,
46
+ not getattr(self.image_processor, 'do_center_crop', False),
47
+ self.image_processor.do_rescale,
48
+ self.image_processor.do_normalize
49
+ )), f"image_processor `{self.image_processor}` is not supported currently"
50
+
51
+ def get_backbone(self):
52
+ return self.backbone
53
+
54
+ def get_image_processor(self):
55
+ return self.image_processor
56
+
57
+ def mock_input(self):
58
+ height, width = self.get_image_size()
59
+ return torch.zeros(1, 3, height, width), self.construct_image_placeholders((1, 1))
60
+
61
+ def get_head(self):
62
+ return self.head
63
+
64
+ def get_image_size(self):
65
+ raise NotImplementedError
66
+
67
+ @staticmethod
68
+ def construct_image_placeholders(grid):
69
+ image_placeholders = [IMAGE_INDICATOR_IDS[0], IMAGE_ATOM_ID, IMAGE_INDICATOR_IDS[1]]
70
+ if grid[0] * grid[1] > 1:
71
+ for r in range(grid[0]):
72
+ for c in range(grid[1]):
73
+ image_placeholders.append(IMAGE_ATOM_ID)
74
+ if c < grid[1] - 1:
75
+ image_placeholders.append(IMAGE_INDICATOR_IDS[2])
76
+ if r < grid[0] - 1:
77
+ image_placeholders.append(IMAGE_INDICATOR_IDS[3])
78
+ image_placeholders.append(IMAGE_INDICATOR_IDS[4])
79
+ return image_placeholders
80
+
81
+ def preprocess_image(self, image: PIL.Image.Image, max_partition=9, covering_threshold=0.9, convert_to_rgb=True):
82
+ def _preprocess(img: PIL.Image.Image, side):
83
+ # first resize and preprocess
84
+ w, h = img.size
85
+ if w == h:
86
+ new_width = new_height = side
87
+ elif w > h:
88
+ new_width = side
89
+ new_height = int(h / w * new_width)
90
+ else:
91
+ new_height = side
92
+ new_width = int(w / h * new_height)
93
+ new_size = dict(height=new_height, width=new_width)
94
+ pixel_values = self.image_processor.preprocess(img, size=new_size, return_tensors='pt')['pixel_values']
95
+
96
+ # then pad to square
97
+ square_values = torch.zeros([1, 3, side, side], dtype=pixel_values.dtype, device=pixel_values.device)
98
+ new_height, new_width = pixel_values.shape[2:]
99
+ if new_height == new_width:
100
+ square_values[:, :, :, :] = pixel_values
101
+ elif new_height > new_width:
102
+ from_index = (side - new_width) // 2
103
+ square_values[:, :, :, from_index:from_index + new_width] = pixel_values
104
+ else:
105
+ from_index = (side - new_height) // 2
106
+ square_values[:, :, from_index:from_index + new_height, :] = pixel_values
107
+
108
+ return square_values
109
+
110
+ def _partition(img, grid):
111
+ w, h = img.size
112
+ row_height = h // grid[0]
113
+ col_width = w // grid[1]
114
+
115
+ partition = []
116
+ for row in range(grid[0]):
117
+ for col in range(grid[1]):
118
+ left = col * col_width
119
+ upper = row * row_height
120
+ right = w if col == grid[1] - 1 else (col + 1) * col_width
121
+ lower = h if row == grid[0] - 1 else (row + 1) * row_height
122
+ partition.append((left, upper, right, lower))
123
+
124
+ return partition
125
+
126
+ def _covering_area(left, upper, right, lower, side):
127
+ w = right - left
128
+ h = lower - upper
129
+ w, h = max(w, h), min(w, h)
130
+ if w > side:
131
+ h = h / w * side
132
+ w = side
133
+ return w * h
134
+
135
+ def _get_best_grid(img, side):
136
+ img_area = img.size[0] * img.size[1]
137
+
138
+ candidate_grids = []
139
+ for i in range(1, max_partition + 1):
140
+ for j in range(1, max_partition + 1):
141
+ if i * j <= max_partition:
142
+ candidate_grids.append((i, j))
143
+
144
+ all_grids = []
145
+ good_grids = []
146
+ for grid in candidate_grids:
147
+ partition = _partition(img, grid)
148
+ covering_ratio = sum([_covering_area(*p, side) for p in partition]) / img_area
149
+ assert covering_ratio <= 1.0
150
+ all_grids.append((grid, covering_ratio))
151
+ if covering_ratio > covering_threshold:
152
+ good_grids.append((grid, covering_ratio))
153
+
154
+ if len(good_grids) > 0:
155
+ # pick the good partition with minimum #sub_images and break the tie using covering_ratio
156
+ return sorted(good_grids, key=lambda x: (x[0][0] * x[0][1], -x[1]))[0][0]
157
+ else:
158
+ # pick the partition with maximum covering_ratio and break the tie using #sub_images
159
+ return sorted(all_grids, key=lambda x: (-x[1], x[0][0] * x[0][1]))[0][0]
160
+
161
+ if convert_to_rgb and image.mode != 'RGB':
162
+ image = image.convert('RGB')
163
+
164
+ sides = self.get_image_size()
165
+ if sides[0] != sides[1]:
166
+ raise ValueError('get_image_size() returns non-square size')
167
+ side = sides[0]
168
+ grid = _get_best_grid(image, side)
169
+ partition = _partition(image, grid)
170
+ crops = [image.crop(p) for p in partition]
171
+ if len(crops) > 1:
172
+ crops.insert(0, image)
173
+ pixel_values = torch.cat([_preprocess(crop, side) for crop in crops], dim=0)
174
+ image_placeholders = self.construct_image_placeholders(grid)
175
+ return pixel_values, image_placeholders
176
+
177
+ def tokenize(self, logits):
178
+ def st_argmax(y_soft, dim): # straight-through softmax
179
+ index = y_soft.max(dim, keepdim=True)[1]
180
+ y_hard = torch.zeros_like(y_soft, memory_format=torch.legacy_contiguous_format).scatter_(dim, index, 1.0)
181
+ ret = y_hard - y_soft.detach() + y_soft
182
+ return ret
183
+
184
+ if self.config.tokenize_function == 'softmax':
185
+ tokens = softmax(logits, dim=-1)
186
+ elif self.config.tokenize_function == 'gumbel_argmax':
187
+ tokens = gumbel_softmax(logits, tau=self.config.tau, hard=True)
188
+ elif self.config.tokenize_function == 'st_argmax':
189
+ tokens = st_argmax(logits, dim=-1)
190
+ else:
191
+ raise ValueError(
192
+ f'Invalid `max_type`, expected softmax or gumbel_argmax or st_argmax, but got {self.config.tokenize_function}')
193
+ return tokens
194
+
195
+ def encode(self, pixel_values):
196
+ output = self.backbone(pixel_values, output_hidden_states=True, return_dict=True)
197
+ features = output.hidden_states[-1]
198
+ if self.config.drop_cls_token:
199
+ features = features[:, 1:, :]
200
+
201
+ # merge number of `hidden_stride * hidden_stride` hidden states together to reduce token sequence length
202
+ # e.g., for hidden_stride=3, this leads to a token length reduction: 729 -> 81 for siglip
203
+ if self.config.hidden_stride > 1:
204
+ n, l, d = features.shape # this `d` maybe different from the above `d
205
+ sqrt_l = int(l ** 0.5)
206
+ assert sqrt_l ** 2 == l, "The token sequence length should be a perfect square."
207
+ features = features.reshape(n, sqrt_l, sqrt_l, d)
208
+ pl = (self.config.hidden_stride - (sqrt_l % self.config.hidden_stride)) % self.config.hidden_stride
209
+ features = pad(features, (0, 0, 0, pl, 0, pl), "constant", 0)
210
+ sqrt_l += pl
211
+ features = features.reshape(n, sqrt_l // self.config.hidden_stride, self.config.hidden_stride,
212
+ sqrt_l // self.config.hidden_stride, self.config.hidden_stride, d)
213
+ features = features.permute(0, 1, 3, 2, 4, 5) # [n, sqrt_l/hs, sqrt_l/hs, hs, hs, d]
214
+ features = features.flatten(3) # [n, sqrt_l/hs, sqrt_l/hs, hs*hs*d]
215
+ features = features.reshape(
216
+ n, -1, self.config.hidden_stride * self.config.hidden_stride * d)
217
+
218
+ return features
219
+
220
+ def forward(self, pixel_values) -> torch.Tensor: # [BatchSize, ImageShape] -> [BatchSize, #Token, VocabSize]
221
+ features = self.encode(pixel_values)
222
+ logits = self.head(features)
223
+ tokens = self.tokenize(logits)
224
+ # tokens' shape is [BatchSize, #Token, VocabSize-5], so padding with [BatchSize, #Token, 5], after
225
+ # which, tokens' shape should become [BatchSize, #Token, VocabSize]
226
+ batch_size, token_len, _ = tokens.shape
227
+ padding_tensor = torch.zeros(size=(batch_size, token_len, len(IMAGE_INDICATOR_IDS)),
228
+ dtype=tokens.dtype,
229
+ device=tokens.device,
230
+ layout=tokens.layout,
231
+ requires_grad=False)
232
+ tokens = torch.cat((tokens, padding_tensor), dim=2)
233
+ return tokens
234
+
235
+
236
+ class SiglipVisualTokenizer(BaseVisualTokenizer):
237
+ config_class = SiglipVisualTokenizerConfig
238
+ supports_gradient_checkpointing = True
239
+ _no_split_modules = ["SiglipVisionTransformer"]
240
+ _image_processor_class = SiglipImageProcessor
241
+ _image_processor_kwargs = {}
242
+ _backbone_class = SiglipVisionModel
243
+ _backbone_name_or_path = "google/siglip-so400m-patch14-384"
244
+
245
+ def get_image_size(self):
246
+ height = self.image_processor.size["height"]
247
+ width = self.image_processor.size["width"]
248
+ return height, width
249
+
250
+
251
+ AutoModel.register(SiglipVisualTokenizerConfig, SiglipVisualTokenizer)
252
+
253
+
254
+ # ----------------------------------------------------------------------
255
+ # Ovis
256
+ # ----------------------------------------------------------------------
257
+ class VisualEmbedding(torch.nn.Embedding):
258
+ def forward(self, visual_tokens: Tensor) -> Tensor:
259
+ if visual_tokens.dtype in [torch.int8, torch.int16, torch.int32, torch.int64, torch.long]:
260
+ return super().forward(visual_tokens)
261
+ return torch.matmul(visual_tokens, self.weight)
262
+
263
+ def reset_parameters(self, mean=0., std=1.) -> None:
264
+ init.normal_(self.weight, mean=mean, std=std)
265
+ self._fill_padding_idx_with_zero()
266
+
267
+
268
+ class OvisPreTrainedModel(PreTrainedModel):
269
+ config_class = OvisConfig
270
+ base_model_prefix = "ovis"
271
+
272
+
273
+ class Ovis(OvisPreTrainedModel):
274
+
275
+ def __init__(self, config: OvisConfig, *inputs, **kwargs):
276
+ super().__init__(config, *inputs, **kwargs)
277
+ attn_kwargs = dict()
278
+ if self.config.llm_attn_implementation:
279
+ attn_kwargs['attn_implementation'] = self.config.llm_attn_implementation
280
+ self.llm = AutoModelForCausalLM.from_config(self.config.llm_config, **attn_kwargs)
281
+ assert self.config.hidden_size == self.llm.config.hidden_size, "hidden size mismatch"
282
+ self.text_tokenizer = AutoTokenizer.from_pretrained(self.config.name_or_path)
283
+ self.visual_tokenizer = AutoModel.from_config(self.config.visual_tokenizer_config,
284
+ image_processor_name_or_path=self.config.name_or_path)
285
+ self.vte = VisualEmbedding(
286
+ self.config.visual_tokenizer_config.vocab_size,
287
+ self.config.hidden_size,
288
+ device=self.visual_tokenizer.device,
289
+ dtype=self.visual_tokenizer.dtype
290
+ )
291
+
292
+ def _merge_modules(modules_list: tuple):
293
+ merged_modules = []
294
+ for modules in modules_list:
295
+ merged_modules.extend(modules if modules else [])
296
+ return merged_modules
297
+
298
+ self._no_split_modules = _merge_modules((self.llm._no_split_modules, self.visual_tokenizer._no_split_modules))
299
+ self._skip_keys_device_placement = self.llm._skip_keys_device_placement
300
+ self._keep_in_fp32_modules = _merge_modules(
301
+ (self.llm._keep_in_fp32_modules, self.visual_tokenizer._keep_in_fp32_modules))
302
+ self.is_parallelizable = all((self.llm.is_parallelizable, self.visual_tokenizer.is_parallelizable))
303
+ self.supports_gradient_checkpointing = all(
304
+ (self.llm.supports_gradient_checkpointing, self.visual_tokenizer.supports_gradient_checkpointing))
305
+ self._supports_flash_attn_2 = all(
306
+ (self.llm._supports_flash_attn_2, self.visual_tokenizer._supports_flash_attn_2))
307
+ self._supports_sdpa = all((self.llm._supports_sdpa, self.visual_tokenizer._supports_sdpa))
308
+
309
+ def get_text_tokenizer(self):
310
+ return self.text_tokenizer
311
+
312
+ def get_visual_tokenizer(self):
313
+ return self.visual_tokenizer
314
+
315
+ def tie_weights(self):
316
+ if not self.config.disable_tie_weight:
317
+ self.get_llm().tie_weights()
318
+
319
+ def get_llm(self):
320
+ return self.llm
321
+
322
+ def get_vte(self):
323
+ return self.vte
324
+
325
+ def get_wte(self):
326
+ return self.llm.get_input_embeddings()
327
+
328
+ def get_conversation_formatter(self) -> ConversationFormatter:
329
+ if getattr(self, 'conversation_formatter', None) is None:
330
+ self.conversation_formatter = getattr(import_module(".configuration_ovis", __package__),
331
+ self.config.conversation_formatter_class)(self.text_tokenizer)
332
+ return self.conversation_formatter
333
+
334
+ def forward(
335
+ self,
336
+ input_ids: torch.Tensor,
337
+ attention_mask: torch.Tensor,
338
+ labels: Optional[torch.Tensor],
339
+ pixel_values: List[Optional[torch.Tensor]],
340
+ **kwargs
341
+ ):
342
+ assert self.training, "`forward` can only be used in training. For inference, use `generate`."
343
+ _, inputs_embeds, labels, attention_mask = self.merge_multimodal(
344
+ text_input_ids=input_ids,
345
+ text_attention_masks=attention_mask,
346
+ text_labels=labels,
347
+ pixel_values=pixel_values
348
+ )
349
+ return self.llm(inputs_embeds=inputs_embeds, labels=labels, attention_mask=attention_mask, **kwargs)
350
+
351
+ def merge_multimodal(
352
+ self,
353
+ text_input_ids: torch.Tensor,
354
+ text_attention_masks: torch.Tensor,
355
+ text_labels: Optional[torch.Tensor],
356
+ pixel_values: List[Optional[torch.Tensor]]
357
+ ):
358
+ input_device = text_input_ids.device
359
+ visual_vocab_szie = self.get_visual_tokenizer().config.vocab_size
360
+ visual_indicator_embeds = self.get_vte()(
361
+ torch.tensor(
362
+ list(range(visual_vocab_szie - 5, visual_vocab_szie)),
363
+ dtype=torch.long,
364
+ device=self.get_visual_tokenizer().device
365
+ )
366
+ ).to(device=input_device)
367
+
368
+ if self.training:
369
+ # When training, to be compatible with deepspeed zero, each sample has to include pixel_value tensor.
370
+ # For text-only sample, one can simply use a full zero tensor as pixel_value, which will be ignored
371
+ # (see below in this function); so, the gradient will not be affected.
372
+ num_images = [x.shape[0] for x in pixel_values]
373
+ visual_tokens = self.visual_tokenizer(torch.cat([x for x in pixel_values], dim=0))
374
+ visual_embeds = torch.split(self.get_vte()(visual_tokens).to(dtype=self.dtype, device=input_device),
375
+ split_size_or_sections=num_images, dim=0)
376
+ visual_input_ids = torch.split(torch.argmax(visual_tokens, dim=-1).to(device=input_device),
377
+ split_size_or_sections=num_images, dim=0)
378
+ visual_labels = [torch.full(x.shape, IGNORE_ID, dtype=torch.long, device=input_device) for x in
379
+ visual_input_ids]
380
+ else:
381
+ # When inference, sample can include only text with `None` pixel_value
382
+ num_images = [x.shape[0] if x is not None else 0 for x in pixel_values]
383
+ if sum(num_images) > 0:
384
+ visual_tokens = self.visual_tokenizer(torch.cat([x for x in pixel_values if x is not None], dim=0))
385
+ visual_embeds = torch.split(self.get_vte()(visual_tokens).to(dtype=self.dtype, device=input_device),
386
+ split_size_or_sections=num_images, dim=0)
387
+ visual_input_ids = torch.split(torch.argmax(visual_tokens, dim=-1).to(device=input_device),
388
+ split_size_or_sections=num_images, dim=0)
389
+ visual_labels = [torch.full(x.shape, IGNORE_ID, dtype=torch.long, device=input_device) for x in
390
+ visual_input_ids]
391
+ else:
392
+ # just placeholders
393
+ visual_embeds = [None] * len(num_images)
394
+ visual_input_ids = [None] * len(num_images)
395
+ visual_labels = [None] * len(num_images)
396
+ # just placeholders
397
+ text_labels = torch.full(text_input_ids.shape, IGNORE_ID, dtype=torch.long, device=input_device)
398
+
399
+ input_embeds = []
400
+ attention_masks = []
401
+ labels = []
402
+ for text_input_id, text_label, text_attention_mask, visual_embed, visual_input_id, visual_label in zip(
403
+ text_input_ids, text_labels, text_attention_masks, visual_embeds, visual_input_ids, visual_labels
404
+ ):
405
+ placeholder_token_mask = torch.lt(text_input_id, 0)
406
+ text_embed = self.get_wte()(torch.masked_fill(text_input_id, placeholder_token_mask, 0))
407
+ for i, indicator_id in enumerate(IMAGE_INDICATOR_IDS):
408
+ text_embed[text_input_id == indicator_id] = visual_indicator_embeds[i]
409
+ image_atom_positions = torch.where(torch.eq(text_input_id, IMAGE_ATOM_ID))[0].tolist()
410
+ if len(image_atom_positions) > 0:
411
+ input_embed_parts = []
412
+ attention_mask_parts = []
413
+ label_parts = []
414
+ prev_image_atom_position = -1
415
+ for index, image_atom_position in enumerate(image_atom_positions):
416
+ input_embed_parts.append(
417
+ text_embed[prev_image_atom_position + 1:image_atom_position, :])
418
+ label_parts.append(
419
+ text_label[prev_image_atom_position + 1:image_atom_position])
420
+ attention_mask_parts.append(
421
+ text_attention_mask[prev_image_atom_position + 1:image_atom_position])
422
+ input_embed_parts.append(visual_embed[index])
423
+ attention_mask_parts.append(
424
+ torch.ones_like(visual_label[index], dtype=torch.bool))
425
+ label_parts.append(visual_label[index])
426
+ prev_image_atom_position = image_atom_position
427
+ if prev_image_atom_position + 1 < text_input_id.shape[0]:
428
+ input_embed_parts.append(
429
+ text_embed[prev_image_atom_position + 1:, :])
430
+ attention_mask_parts.append(
431
+ text_attention_mask[prev_image_atom_position + 1:])
432
+ label_parts.append(
433
+ text_label[prev_image_atom_position + 1:])
434
+ input_embed = torch.cat(input_embed_parts, dim=0)
435
+ attention_mask = torch.cat(attention_mask_parts, dim=0)
436
+ label = torch.cat(label_parts, dim=0)
437
+ else:
438
+ input_embed = text_embed
439
+ attention_mask = text_attention_mask
440
+ label = text_label
441
+ if self.training:
442
+ # Make visual_embed & visual_indicator_embeds involved in the backward graph,
443
+ # to be compatible with deepspeed zero and ddp.
444
+ input_embed += torch.sum(visual_embed * 0.0) + torch.sum(visual_indicator_embeds * 0.0)
445
+ input_embeds.append(input_embed)
446
+ attention_masks.append(attention_mask)
447
+ labels.append(label)
448
+
449
+ if self.training: # padding to self.config.multimodal_max_length for increased training speed
450
+ padding_size = max(0, self.config.multimodal_max_length - len(input_embeds[0]))
451
+ input_embeds[0] = torch.nn.ConstantPad2d((0, 0, 0, padding_size), 0.0)(input_embeds[0])
452
+ attention_masks[0] = torch.nn.ConstantPad1d((0, padding_size), False)(attention_masks[0])
453
+ labels[0] = torch.nn.ConstantPad1d((0, padding_size), IGNORE_ID)(labels[0])
454
+ batch_input_embeds = torch.nn.utils.rnn.pad_sequence(input_embeds, batch_first=True, padding_value=0.0)[:,
455
+ :self.config.multimodal_max_length, :]
456
+ batch_attention_mask = torch.nn.utils.rnn.pad_sequence(attention_masks, batch_first=True, padding_value=False)[
457
+ :,
458
+ :self.config.multimodal_max_length]
459
+ batch_labels = torch.nn.utils.rnn.pad_sequence(labels, batch_first=True, padding_value=IGNORE_ID)[:,
460
+ :self.config.multimodal_max_length]
461
+
462
+ return visual_input_ids, batch_input_embeds, batch_labels, batch_attention_mask
463
+
464
+ def preprocess_inputs(
465
+ self,
466
+ text_or_conversations: Union[List[Dict], str],
467
+ images: Optional[List[PIL.Image.Image]],
468
+ max_partition=9,
469
+ generation_preface='',
470
+ return_labels=False,
471
+ propagate_exception=True
472
+ ):
473
+ # convert text to conversations
474
+ if isinstance(text_or_conversations, str):
475
+ conversations = [{
476
+ "from": "human",
477
+ "value": text_or_conversations
478
+ }]
479
+ elif isinstance(text_or_conversations, list):
480
+ conversations = text_or_conversations
481
+ else:
482
+ raise ValueError(f'Invalid type of `text_or_conversations`, expected `List[Dict]` or `str`,'
483
+ f' but got {type(text_or_conversations)}')
484
+
485
+ # format conversations
486
+ prompt, raw_input_ids, raw_labels = self.get_conversation_formatter().format(
487
+ conversations, generation_preface=generation_preface)
488
+
489
+ # place image placeholders
490
+ input_ids = []
491
+ labels = []
492
+ pixel_values = []
493
+ invalidate_label = False
494
+ image_token_indices = [i for i, v in enumerate(raw_input_ids) if v == IMAGE_TOKEN_ID]
495
+ last_image_token_index = -1
496
+ for i in range(len(image_token_indices)):
497
+ head = 0 if i == 0 else image_token_indices[i - 1] + 1
498
+ tail = image_token_indices[i]
499
+ last_image_token_index = tail
500
+ input_ids.extend(raw_input_ids[head:tail])
501
+ labels.extend(raw_labels[head:tail])
502
+ try:
503
+ image = images[i]
504
+ raw_pixel_values, image_placeholders = self.visual_tokenizer.preprocess_image(
505
+ image, max_partition=max_partition)
506
+ except Exception as e:
507
+ if propagate_exception:
508
+ raise e
509
+ logging.exception(e)
510
+ invalidate_label = True
511
+ raw_pixel_values, image_placeholders = self.visual_tokenizer.mock_input()
512
+ input_ids.extend(image_placeholders)
513
+ labels.extend([IGNORE_ID] * len(image_placeholders))
514
+ pixel_values.append(raw_pixel_values)
515
+ input_ids.extend(raw_input_ids[last_image_token_index + 1:])
516
+ labels.extend(raw_labels[last_image_token_index + 1:])
517
+
518
+ # return tensors
519
+ input_ids = torch.tensor(input_ids, dtype=torch.long)
520
+ labels = torch.tensor([IGNORE_ID] * len(labels) if invalidate_label else labels, dtype=torch.long)
521
+ pixel_values = torch.cat(pixel_values, dim=0) if len(pixel_values) > 0 else None
522
+
523
+ if return_labels:
524
+ return prompt, input_ids, pixel_values, labels
525
+ else:
526
+ return prompt, input_ids, pixel_values
527
+
528
+ def save_pretrained(
529
+ self,
530
+ save_directory: Union[str, os.PathLike],
531
+ is_main_process: bool = True,
532
+ state_dict: Optional[dict] = None,
533
+ save_function: Callable = torch.save,
534
+ push_to_hub: bool = False,
535
+ max_shard_size: Union[int, str] = "5GB",
536
+ safe_serialization: bool = True,
537
+ variant: Optional[str] = None,
538
+ token: Optional[Union[str, bool]] = None,
539
+ save_peft_format: bool = True,
540
+ **kwargs
541
+ ):
542
+ super().save_pretrained(save_directory,
543
+ is_main_process=is_main_process,
544
+ state_dict=state_dict,
545
+ save_function=save_function,
546
+ safe_serialization=safe_serialization)
547
+ self.get_text_tokenizer().save_pretrained(save_directory)
548
+ self.get_visual_tokenizer().get_image_processor().save_pretrained(save_directory)
549
+
550
+ def _get_hybrid_cache_for_llm(self, max_batch_size: int, max_cache_len: int):
551
+ cache_cls = HybridCache
552
+ llm = self.get_llm()
553
+
554
+ need_new_cache = (
555
+ not hasattr(llm, "_cache")
556
+ or (not isinstance(llm._cache, cache_cls))
557
+ or llm._cache.max_batch_size != max_batch_size
558
+ or llm._cache.max_cache_len < max_cache_len
559
+ )
560
+
561
+ if need_new_cache:
562
+ if hasattr(llm.config, "_pre_quantization_dtype"):
563
+ cache_dtype = llm.config._pre_quantization_dtype
564
+ else:
565
+ cache_dtype = llm.dtype
566
+ llm._cache = cache_cls(
567
+ config=llm.config,
568
+ max_batch_size=max_batch_size,
569
+ max_cache_len=max_cache_len,
570
+ device=llm.device,
571
+ dtype=cache_dtype,
572
+ )
573
+ else:
574
+ llm._cache.reset()
575
+ return llm._cache
576
+
577
+ # TODO: support batch generation
578
+ def generate(
579
+ self,
580
+ inputs: Optional[torch.Tensor] = None,
581
+ **kwargs
582
+ ) -> Union[GenerateOutput, torch.LongTensor]:
583
+ assert inputs.shape[0] == 1, 'Currently, only support `batch_size=1`'
584
+ _, inputs_embeds, labels, attention_mask = self.merge_multimodal(
585
+ text_input_ids=inputs,
586
+ text_attention_masks=kwargs.pop('attention_mask'),
587
+ text_labels=None,
588
+ pixel_values=kwargs.pop('pixel_values')
589
+ )
590
+ if getattr(self.generation_config, 'cache_implementation') == 'hybrid': # mainly for Gemma2
591
+ kwargs['past_key_values'] = self._get_hybrid_cache_for_llm(
592
+ getattr(kwargs, "num_beams", 1), kwargs['max_new_tokens'] + inputs_embeds.shape[-2])
593
+ self.get_llm()._supports_cache_class = True
594
+ kwargs['cache_implementation'] = None
595
+
596
+ return self.llm.generate(inputs=None, inputs_embeds=inputs_embeds, attention_mask=attention_mask, **kwargs)
preprocessor_config.json ADDED
@@ -0,0 +1,24 @@
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
1
+ {
2
+ "do_convert_rgb": null,
3
+ "do_normalize": true,
4
+ "do_rescale": true,
5
+ "do_resize": true,
6
+ "image_mean": [
7
+ 0.5,
8
+ 0.5,
9
+ 0.5
10
+ ],
11
+ "image_processor_type": "SiglipImageProcessor",
12
+ "image_std": [
13
+ 0.5,
14
+ 0.5,
15
+ 0.5
16
+ ],
17
+ "processor_class": "SiglipProcessor",
18
+ "resample": 3,
19
+ "rescale_factor": 0.00392156862745098,
20
+ "size": {
21
+ "height": 384,
22
+ "width": 384
23
+ }
24
+ }
special_tokens_map.json ADDED
@@ -0,0 +1,34 @@
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
1
+ {
2
+ "additional_special_tokens": [
3
+ "<start_of_turn>",
4
+ "<end_of_turn>"
5
+ ],
6
+ "bos_token": {
7
+ "content": "<bos>",
8
+ "lstrip": false,
9
+ "normalized": false,
10
+ "rstrip": false,
11
+ "single_word": false
12
+ },
13
+ "eos_token": {
14
+ "content": "<eos>",
15
+ "lstrip": false,
16
+ "normalized": false,
17
+ "rstrip": false,
18
+ "single_word": false
19
+ },
20
+ "pad_token": {
21
+ "content": "<pad>",
22
+ "lstrip": false,
23
+ "normalized": false,
24
+ "rstrip": false,
25
+ "single_word": false
26
+ },
27
+ "unk_token": {
28
+ "content": "<unk>",
29
+ "lstrip": false,
30
+ "normalized": false,
31
+ "rstrip": false,
32
+ "single_word": false
33
+ }
34
+ }
tokenizer.json ADDED
@@ -0,0 +1,3 @@
 
 
 
 
1
+ version https://git-lfs.github.com/spec/v1
2
+ oid sha256:7da53ca29fb16f6b2489482fc0bc6a394162cdab14d12764a1755ebc583fea79
3
+ size 17518525
tokenizer.model ADDED
@@ -0,0 +1,3 @@
 
 
 
 
1
+ version https://git-lfs.github.com/spec/v1
2
+ oid sha256:61a7b147390c64585d6c3543dd6fc636906c9af3865a5548f27f31aee1d4c8e2
3
+ size 4241003
tokenizer_config.json ADDED
@@ -0,0 +1,1757 @@
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
1
+ {
2
+ "add_bos_token": true,
3
+ "add_eos_token": false,
4
+ "added_tokens_decoder": {
5
+ "0": {
6
+ "content": "<pad>",
7
+ "lstrip": false,
8
+ "normalized": false,
9
+ "rstrip": false,
10
+ "single_word": false,
11
+ "special": true
12
+ },
13
+ "1": {
14
+ "content": "<eos>",
15
+ "lstrip": false,
16
+ "normalized": false,
17
+ "rstrip": false,
18
+ "single_word": false,
19
+ "special": true
20
+ },
21
+ "2": {
22
+ "content": "<bos>",
23
+ "lstrip": false,
24
+ "normalized": false,
25
+ "rstrip": false,
26
+ "single_word": false,
27
+ "special": true
28
+ },
29
+ "3": {
30
+ "content": "<unk>",
31
+ "lstrip": false,
32
+ "normalized": false,
33
+ "rstrip": false,
34
+ "single_word": false,
35
+ "special": true
36
+ },
37
+ "4": {
38
+ "content": "<mask>",
39
+ "lstrip": false,
40
+ "normalized": false,
41
+ "rstrip": false,
42
+ "single_word": false,
43
+ "special": false
44
+ },
45
+ "5": {
46
+ "content": "<2mass>",
47
+ "lstrip": false,
48
+ "normalized": false,
49
+ "rstrip": false,
50
+ "single_word": false,
51
+ "special": false
52
+ },
53
+ "6": {
54
+ "content": "[@BOS@]",
55
+ "lstrip": false,
56
+ "normalized": false,
57
+ "rstrip": false,
58
+ "single_word": false,
59
+ "special": false
60
+ },
61
+ "7": {
62
+ "content": "<unused0>",
63
+ "lstrip": false,
64
+ "normalized": false,
65
+ "rstrip": false,
66
+ "single_word": false,
67
+ "special": false
68
+ },
69
+ "8": {
70
+ "content": "<unused1>",
71
+ "lstrip": false,
72
+ "normalized": false,
73
+ "rstrip": false,
74
+ "single_word": false,
75
+ "special": false
76
+ },
77
+ "9": {
78
+ "content": "<unused2>",
79
+ "lstrip": false,
80
+ "normalized": false,
81
+ "rstrip": false,
82
+ "single_word": false,
83
+ "special": false
84
+ },
85
+ "10": {
86
+ "content": "<unused3>",
87
+ "lstrip": false,
88
+ "normalized": false,
89
+ "rstrip": false,
90
+ "single_word": false,
91
+ "special": false
92
+ },
93
+ "11": {
94
+ "content": "<unused4>",
95
+ "lstrip": false,
96
+ "normalized": false,
97
+ "rstrip": false,
98
+ "single_word": false,
99
+ "special": false
100
+ },
101
+ "12": {
102
+ "content": "<unused5>",
103
+ "lstrip": false,
104
+ "normalized": false,
105
+ "rstrip": false,
106
+ "single_word": false,
107
+ "special": false
108
+ },
109
+ "13": {
110
+ "content": "<unused6>",
111
+ "lstrip": false,
112
+ "normalized": false,
113
+ "rstrip": false,
114
+ "single_word": false,
115
+ "special": false
116
+ },
117
+ "14": {
118
+ "content": "<unused7>",
119
+ "lstrip": false,
120
+ "normalized": false,
121
+ "rstrip": false,
122
+ "single_word": false,
123
+ "special": false
124
+ },
125
+ "15": {
126
+ "content": "<unused8>",
127
+ "lstrip": false,
128
+ "normalized": false,
129
+ "rstrip": false,
130
+ "single_word": false,
131
+ "special": false
132
+ },
133
+ "16": {
134
+ "content": "<unused9>",
135
+ "lstrip": false,
136
+ "normalized": false,
137
+ "rstrip": false,
138
+ "single_word": false,
139
+ "special": false
140
+ },
141
+ "17": {
142
+ "content": "<unused10>",
143
+ "lstrip": false,
144
+ "normalized": false,
145
+ "rstrip": false,
146
+ "single_word": false,
147
+ "special": false
148
+ },
149
+ "18": {
150
+ "content": "<unused11>",
151
+ "lstrip": false,
152
+ "normalized": false,
153
+ "rstrip": false,
154
+ "single_word": false,
155
+ "special": false
156
+ },
157
+ "19": {
158
+ "content": "<unused12>",
159
+ "lstrip": false,
160
+ "normalized": false,
161
+ "rstrip": false,
162
+ "single_word": false,
163
+ "special": false
164
+ },
165
+ "20": {
166
+ "content": "<unused13>",
167
+ "lstrip": false,
168
+ "normalized": false,
169
+ "rstrip": false,
170
+ "single_word": false,
171
+ "special": false
172
+ },
173
+ "21": {
174
+ "content": "<unused14>",
175
+ "lstrip": false,
176
+ "normalized": false,
177
+ "rstrip": false,
178
+ "single_word": false,
179
+ "special": false
180
+ },
181
+ "22": {
182
+ "content": "<unused15>",
183
+ "lstrip": false,
184
+ "normalized": false,
185
+ "rstrip": false,
186
+ "single_word": false,
187
+ "special": false
188
+ },
189
+ "23": {
190
+ "content": "<unused16>",
191
+ "lstrip": false,
192
+ "normalized": false,
193
+ "rstrip": false,
194
+ "single_word": false,
195
+ "special": false
196
+ },
197
+ "24": {
198
+ "content": "<unused17>",
199
+ "lstrip": false,
200
+ "normalized": false,
201
+ "rstrip": false,
202
+ "single_word": false,
203
+ "special": false
204
+ },
205
+ "25": {
206
+ "content": "<unused18>",
207
+ "lstrip": false,
208
+ "normalized": false,
209
+ "rstrip": false,
210
+ "single_word": false,
211
+ "special": false
212
+ },
213
+ "26": {
214
+ "content": "<unused19>",
215
+ "lstrip": false,
216
+ "normalized": false,
217
+ "rstrip": false,
218
+ "single_word": false,
219
+ "special": false
220
+ },
221
+ "27": {
222
+ "content": "<unused20>",
223
+ "lstrip": false,
224
+ "normalized": false,
225
+ "rstrip": false,
226
+ "single_word": false,
227
+ "special": false
228
+ },
229
+ "28": {
230
+ "content": "<unused21>",
231
+ "lstrip": false,
232
+ "normalized": false,
233
+ "rstrip": false,
234
+ "single_word": false,
235
+ "special": false
236
+ },
237
+ "29": {
238
+ "content": "<unused22>",
239
+ "lstrip": false,
240
+ "normalized": false,
241
+ "rstrip": false,
242
+ "single_word": false,
243
+ "special": false
244
+ },
245
+ "30": {
246
+ "content": "<unused23>",
247
+ "lstrip": false,
248
+ "normalized": false,
249
+ "rstrip": false,
250
+ "single_word": false,
251
+ "special": false
252
+ },
253
+ "31": {
254
+ "content": "<unused24>",
255
+ "lstrip": false,
256
+ "normalized": false,
257
+ "rstrip": false,
258
+ "single_word": false,
259
+ "special": false
260
+ },
261
+ "32": {
262
+ "content": "<unused25>",
263
+ "lstrip": false,
264
+ "normalized": false,
265
+ "rstrip": false,
266
+ "single_word": false,
267
+ "special": false
268
+ },
269
+ "33": {
270
+ "content": "<unused26>",
271
+ "lstrip": false,
272
+ "normalized": false,
273
+ "rstrip": false,
274
+ "single_word": false,
275
+ "special": false
276
+ },
277
+ "34": {
278
+ "content": "<unused27>",
279
+ "lstrip": false,
280
+ "normalized": false,
281
+ "rstrip": false,
282
+ "single_word": false,
283
+ "special": false
284
+ },
285
+ "35": {
286
+ "content": "<unused28>",
287
+ "lstrip": false,
288
+ "normalized": false,
289
+ "rstrip": false,
290
+ "single_word": false,
291
+ "special": false
292
+ },
293
+ "36": {
294
+ "content": "<unused29>",
295
+ "lstrip": false,
296
+ "normalized": false,
297
+ "rstrip": false,
298
+ "single_word": false,
299
+ "special": false
300
+ },
301
+ "37": {
302
+ "content": "<unused30>",
303
+ "lstrip": false,
304
+ "normalized": false,
305
+ "rstrip": false,
306
+ "single_word": false,
307
+ "special": false
308
+ },
309
+ "38": {
310
+ "content": "<unused31>",
311
+ "lstrip": false,
312
+ "normalized": false,
313
+ "rstrip": false,
314
+ "single_word": false,
315
+ "special": false
316
+ },
317
+ "39": {
318
+ "content": "<unused32>",
319
+ "lstrip": false,
320
+ "normalized": false,
321
+ "rstrip": false,
322
+ "single_word": false,
323
+ "special": false
324
+ },
325
+ "40": {
326
+ "content": "<unused33>",
327
+ "lstrip": false,
328
+ "normalized": false,
329
+ "rstrip": false,
330
+ "single_word": false,
331
+ "special": false
332
+ },
333
+ "41": {
334
+ "content": "<unused34>",
335
+ "lstrip": false,
336
+ "normalized": false,
337
+ "rstrip": false,
338
+ "single_word": false,
339
+ "special": false
340
+ },
341
+ "42": {
342
+ "content": "<unused35>",
343
+ "lstrip": false,
344
+ "normalized": false,
345
+ "rstrip": false,
346
+ "single_word": false,
347
+ "special": false
348
+ },
349
+ "43": {
350
+ "content": "<unused36>",
351
+ "lstrip": false,
352
+ "normalized": false,
353
+ "rstrip": false,
354
+ "single_word": false,
355
+ "special": false
356
+ },
357
+ "44": {
358
+ "content": "<unused37>",
359
+ "lstrip": false,
360
+ "normalized": false,
361
+ "rstrip": false,
362
+ "single_word": false,
363
+ "special": false
364
+ },
365
+ "45": {
366
+ "content": "<unused38>",
367
+ "lstrip": false,
368
+ "normalized": false,
369
+ "rstrip": false,
370
+ "single_word": false,
371
+ "special": false
372
+ },
373
+ "46": {
374
+ "content": "<unused39>",
375
+ "lstrip": false,
376
+ "normalized": false,
377
+ "rstrip": false,
378
+ "single_word": false,
379
+ "special": false
380
+ },
381
+ "47": {
382
+ "content": "<unused40>",
383
+ "lstrip": false,
384
+ "normalized": false,
385
+ "rstrip": false,
386
+ "single_word": false,
387
+ "special": false
388
+ },
389
+ "48": {
390
+ "content": "<unused41>",
391
+ "lstrip": false,
392
+ "normalized": false,
393
+ "rstrip": false,
394
+ "single_word": false,
395
+ "special": false
396
+ },
397
+ "49": {
398
+ "content": "<unused42>",
399
+ "lstrip": false,
400
+ "normalized": false,
401
+ "rstrip": false,
402
+ "single_word": false,
403
+ "special": false
404
+ },
405
+ "50": {
406
+ "content": "<unused43>",
407
+ "lstrip": false,
408
+ "normalized": false,
409
+ "rstrip": false,
410
+ "single_word": false,
411
+ "special": false
412
+ },
413
+ "51": {
414
+ "content": "<unused44>",
415
+ "lstrip": false,
416
+ "normalized": false,
417
+ "rstrip": false,
418
+ "single_word": false,
419
+ "special": false
420
+ },
421
+ "52": {
422
+ "content": "<unused45>",
423
+ "lstrip": false,
424
+ "normalized": false,
425
+ "rstrip": false,
426
+ "single_word": false,
427
+ "special": false
428
+ },
429
+ "53": {
430
+ "content": "<unused46>",
431
+ "lstrip": false,
432
+ "normalized": false,
433
+ "rstrip": false,
434
+ "single_word": false,
435
+ "special": false
436
+ },
437
+ "54": {
438
+ "content": "<unused47>",
439
+ "lstrip": false,
440
+ "normalized": false,
441
+ "rstrip": false,
442
+ "single_word": false,
443
+ "special": false
444
+ },
445
+ "55": {
446
+ "content": "<unused48>",
447
+ "lstrip": false,
448
+ "normalized": false,
449
+ "rstrip": false,
450
+ "single_word": false,
451
+ "special": false
452
+ },
453
+ "56": {
454
+ "content": "<unused49>",
455
+ "lstrip": false,
456
+ "normalized": false,
457
+ "rstrip": false,
458
+ "single_word": false,
459
+ "special": false
460
+ },
461
+ "57": {
462
+ "content": "<unused50>",
463
+ "lstrip": false,
464
+ "normalized": false,
465
+ "rstrip": false,
466
+ "single_word": false,
467
+ "special": false
468
+ },
469
+ "58": {
470
+ "content": "<unused51>",
471
+ "lstrip": false,
472
+ "normalized": false,
473
+ "rstrip": false,
474
+ "single_word": false,
475
+ "special": false
476
+ },
477
+ "59": {
478
+ "content": "<unused52>",
479
+ "lstrip": false,
480
+ "normalized": false,
481
+ "rstrip": false,
482
+ "single_word": false,
483
+ "special": false
484
+ },
485
+ "60": {
486
+ "content": "<unused53>",
487
+ "lstrip": false,
488
+ "normalized": false,
489
+ "rstrip": false,
490
+ "single_word": false,
491
+ "special": false
492
+ },
493
+ "61": {
494
+ "content": "<unused54>",
495
+ "lstrip": false,
496
+ "normalized": false,
497
+ "rstrip": false,
498
+ "single_word": false,
499
+ "special": false
500
+ },
501
+ "62": {
502
+ "content": "<unused55>",
503
+ "lstrip": false,
504
+ "normalized": false,
505
+ "rstrip": false,
506
+ "single_word": false,
507
+ "special": false
508
+ },
509
+ "63": {
510
+ "content": "<unused56>",
511
+ "lstrip": false,
512
+ "normalized": false,
513
+ "rstrip": false,
514
+ "single_word": false,
515
+ "special": false
516
+ },
517
+ "64": {
518
+ "content": "<unused57>",
519
+ "lstrip": false,
520
+ "normalized": false,
521
+ "rstrip": false,
522
+ "single_word": false,
523
+ "special": false
524
+ },
525
+ "65": {
526
+ "content": "<unused58>",
527
+ "lstrip": false,
528
+ "normalized": false,
529
+ "rstrip": false,
530
+ "single_word": false,
531
+ "special": false
532
+ },
533
+ "66": {
534
+ "content": "<unused59>",
535
+ "lstrip": false,
536
+ "normalized": false,
537
+ "rstrip": false,
538
+ "single_word": false,
539
+ "special": false
540
+ },
541
+ "67": {
542
+ "content": "<unused60>",
543
+ "lstrip": false,
544
+ "normalized": false,
545
+ "rstrip": false,
546
+ "single_word": false,
547
+ "special": false
548
+ },
549
+ "68": {
550
+ "content": "<unused61>",
551
+ "lstrip": false,
552
+ "normalized": false,
553
+ "rstrip": false,
554
+ "single_word": false,
555
+ "special": false
556
+ },
557
+ "69": {
558
+ "content": "<unused62>",
559
+ "lstrip": false,
560
+ "normalized": false,
561
+ "rstrip": false,
562
+ "single_word": false,
563
+ "special": false
564
+ },
565
+ "70": {
566
+ "content": "<unused63>",
567
+ "lstrip": false,
568
+ "normalized": false,
569
+ "rstrip": false,
570
+ "single_word": false,
571
+ "special": false
572
+ },
573
+ "71": {
574
+ "content": "<unused64>",
575
+ "lstrip": false,
576
+ "normalized": false,
577
+ "rstrip": false,
578
+ "single_word": false,
579
+ "special": false
580
+ },
581
+ "72": {
582
+ "content": "<unused65>",
583
+ "lstrip": false,
584
+ "normalized": false,
585
+ "rstrip": false,
586
+ "single_word": false,
587
+ "special": false
588
+ },
589
+ "73": {
590
+ "content": "<unused66>",
591
+ "lstrip": false,
592
+ "normalized": false,
593
+ "rstrip": false,
594
+ "single_word": false,
595
+ "special": false
596
+ },
597
+ "74": {
598
+ "content": "<unused67>",
599
+ "lstrip": false,
600
+ "normalized": false,
601
+ "rstrip": false,
602
+ "single_word": false,
603
+ "special": false
604
+ },
605
+ "75": {
606
+ "content": "<unused68>",
607
+ "lstrip": false,
608
+ "normalized": false,
609
+ "rstrip": false,
610
+ "single_word": false,
611
+ "special": false
612
+ },
613
+ "76": {
614
+ "content": "<unused69>",
615
+ "lstrip": false,
616
+ "normalized": false,
617
+ "rstrip": false,
618
+ "single_word": false,
619
+ "special": false
620
+ },
621
+ "77": {
622
+ "content": "<unused70>",
623
+ "lstrip": false,
624
+ "normalized": false,
625
+ "rstrip": false,
626
+ "single_word": false,
627
+ "special": false
628
+ },
629
+ "78": {
630
+ "content": "<unused71>",
631
+ "lstrip": false,
632
+ "normalized": false,
633
+ "rstrip": false,
634
+ "single_word": false,
635
+ "special": false
636
+ },
637
+ "79": {
638
+ "content": "<unused72>",
639
+ "lstrip": false,
640
+ "normalized": false,
641
+ "rstrip": false,
642
+ "single_word": false,
643
+ "special": false
644
+ },
645
+ "80": {
646
+ "content": "<unused73>",
647
+ "lstrip": false,
648
+ "normalized": false,
649
+ "rstrip": false,
650
+ "single_word": false,
651
+ "special": false
652
+ },
653
+ "81": {
654
+ "content": "<unused74>",
655
+ "lstrip": false,
656
+ "normalized": false,
657
+ "rstrip": false,
658
+ "single_word": false,
659
+ "special": false
660
+ },
661
+ "82": {
662
+ "content": "<unused75>",
663
+ "lstrip": false,
664
+ "normalized": false,
665
+ "rstrip": false,
666
+ "single_word": false,
667
+ "special": false
668
+ },
669
+ "83": {
670
+ "content": "<unused76>",
671
+ "lstrip": false,
672
+ "normalized": false,
673
+ "rstrip": false,
674
+ "single_word": false,
675
+ "special": false
676
+ },
677
+ "84": {
678
+ "content": "<unused77>",
679
+ "lstrip": false,
680
+ "normalized": false,
681
+ "rstrip": false,
682
+ "single_word": false,
683
+ "special": false
684
+ },
685
+ "85": {
686
+ "content": "<unused78>",
687
+ "lstrip": false,
688
+ "normalized": false,
689
+ "rstrip": false,
690
+ "single_word": false,
691
+ "special": false
692
+ },
693
+ "86": {
694
+ "content": "<unused79>",
695
+ "lstrip": false,
696
+ "normalized": false,
697
+ "rstrip": false,
698
+ "single_word": false,
699
+ "special": false
700
+ },
701
+ "87": {
702
+ "content": "<unused80>",
703
+ "lstrip": false,
704
+ "normalized": false,
705
+ "rstrip": false,
706
+ "single_word": false,
707
+ "special": false
708
+ },
709
+ "88": {
710
+ "content": "<unused81>",
711
+ "lstrip": false,
712
+ "normalized": false,
713
+ "rstrip": false,
714
+ "single_word": false,
715
+ "special": false
716
+ },
717
+ "89": {
718
+ "content": "<unused82>",
719
+ "lstrip": false,
720
+ "normalized": false,
721
+ "rstrip": false,
722
+ "single_word": false,
723
+ "special": false
724
+ },
725
+ "90": {
726
+ "content": "<unused83>",
727
+ "lstrip": false,
728
+ "normalized": false,
729
+ "rstrip": false,
730
+ "single_word": false,
731
+ "special": false
732
+ },
733
+ "91": {
734
+ "content": "<unused84>",
735
+ "lstrip": false,
736
+ "normalized": false,
737
+ "rstrip": false,
738
+ "single_word": false,
739
+ "special": false
740
+ },
741
+ "92": {
742
+ "content": "<unused85>",
743
+ "lstrip": false,
744
+ "normalized": false,
745
+ "rstrip": false,
746
+ "single_word": false,
747
+ "special": false
748
+ },
749
+ "93": {
750
+ "content": "<unused86>",
751
+ "lstrip": false,
752
+ "normalized": false,
753
+ "rstrip": false,
754
+ "single_word": false,
755
+ "special": false
756
+ },
757
+ "94": {
758
+ "content": "<unused87>",
759
+ "lstrip": false,
760
+ "normalized": false,
761
+ "rstrip": false,
762
+ "single_word": false,
763
+ "special": false
764
+ },
765
+ "95": {
766
+ "content": "<unused88>",
767
+ "lstrip": false,
768
+ "normalized": false,
769
+ "rstrip": false,
770
+ "single_word": false,
771
+ "special": false
772
+ },
773
+ "96": {
774
+ "content": "<unused89>",
775
+ "lstrip": false,
776
+ "normalized": false,
777
+ "rstrip": false,
778
+ "single_word": false,
779
+ "special": false
780
+ },
781
+ "97": {
782
+ "content": "<unused90>",
783
+ "lstrip": false,
784
+ "normalized": false,
785
+ "rstrip": false,
786
+ "single_word": false,
787
+ "special": false
788
+ },
789
+ "98": {
790
+ "content": "<unused91>",
791
+ "lstrip": false,
792
+ "normalized": false,
793
+ "rstrip": false,
794
+ "single_word": false,
795
+ "special": false
796
+ },
797
+ "99": {
798
+ "content": "<unused92>",
799
+ "lstrip": false,
800
+ "normalized": false,
801
+ "rstrip": false,
802
+ "single_word": false,
803
+ "special": false
804
+ },
805
+ "100": {
806
+ "content": "<unused93>",
807
+ "lstrip": false,
808
+ "normalized": false,
809
+ "rstrip": false,
810
+ "single_word": false,
811
+ "special": false
812
+ },
813
+ "101": {
814
+ "content": "<unused94>",
815
+ "lstrip": false,
816
+ "normalized": false,
817
+ "rstrip": false,
818
+ "single_word": false,
819
+ "special": false
820
+ },
821
+ "102": {
822
+ "content": "<unused95>",
823
+ "lstrip": false,
824
+ "normalized": false,
825
+ "rstrip": false,
826
+ "single_word": false,
827
+ "special": false
828
+ },
829
+ "103": {
830
+ "content": "<unused96>",
831
+ "lstrip": false,
832
+ "normalized": false,
833
+ "rstrip": false,
834
+ "single_word": false,
835
+ "special": false
836
+ },
837
+ "104": {
838
+ "content": "<unused97>",
839
+ "lstrip": false,
840
+ "normalized": false,
841
+ "rstrip": false,
842
+ "single_word": false,
843
+ "special": false
844
+ },
845
+ "105": {
846
+ "content": "<unused98>",
847
+ "lstrip": false,
848
+ "normalized": false,
849
+ "rstrip": false,
850
+ "single_word": false,
851
+ "special": false
852
+ },
853
+ "106": {
854
+ "content": "<start_of_turn>",
855
+ "lstrip": false,
856
+ "normalized": false,
857
+ "rstrip": false,
858
+ "single_word": false,
859
+ "special": true
860
+ },
861
+ "107": {
862
+ "content": "<end_of_turn>",
863
+ "lstrip": false,
864
+ "normalized": false,
865
+ "rstrip": false,
866
+ "single_word": false,
867
+ "special": true
868
+ },
869
+ "108": {
870
+ "content": "\n",
871
+ "lstrip": false,
872
+ "normalized": false,
873
+ "rstrip": false,
874
+ "single_word": false,
875
+ "special": false
876
+ },
877
+ "109": {
878
+ "content": "\n\n",
879
+ "lstrip": false,
880
+ "normalized": false,
881
+ "rstrip": false,
882
+ "single_word": false,
883
+ "special": false
884
+ },
885
+ "110": {
886
+ "content": "\n\n\n",
887
+ "lstrip": false,
888
+ "normalized": false,
889
+ "rstrip": false,
890
+ "single_word": false,
891
+ "special": false
892
+ },
893
+ "111": {
894
+ "content": "\n\n\n\n",
895
+ "lstrip": false,
896
+ "normalized": false,
897
+ "rstrip": false,
898
+ "single_word": false,
899
+ "special": false
900
+ },
901
+ "112": {
902
+ "content": "\n\n\n\n\n",
903
+ "lstrip": false,
904
+ "normalized": false,
905
+ "rstrip": false,
906
+ "single_word": false,
907
+ "special": false
908
+ },
909
+ "113": {
910
+ "content": "\n\n\n\n\n\n",
911
+ "lstrip": false,
912
+ "normalized": false,
913
+ "rstrip": false,
914
+ "single_word": false,
915
+ "special": false
916
+ },
917
+ "114": {
918
+ "content": "\n\n\n\n\n\n\n",
919
+ "lstrip": false,
920
+ "normalized": false,
921
+ "rstrip": false,
922
+ "single_word": false,
923
+ "special": false
924
+ },
925
+ "115": {
926
+ "content": "\n\n\n\n\n\n\n\n",
927
+ "lstrip": false,
928
+ "normalized": false,
929
+ "rstrip": false,
930
+ "single_word": false,
931
+ "special": false
932
+ },
933
+ "116": {
934
+ "content": "\n\n\n\n\n\n\n\n\n",
935
+ "lstrip": false,
936
+ "normalized": false,
937
+ "rstrip": false,
938
+ "single_word": false,
939
+ "special": false
940
+ },
941
+ "117": {
942
+ "content": "\n\n\n\n\n\n\n\n\n\n",
943
+ "lstrip": false,
944
+ "normalized": false,
945
+ "rstrip": false,
946
+ "single_word": false,
947
+ "special": false
948
+ },
949
+ "118": {
950
+ "content": "\n\n\n\n\n\n\n\n\n\n\n",
951
+ "lstrip": false,
952
+ "normalized": false,
953
+ "rstrip": false,
954
+ "single_word": false,
955
+ "special": false
956
+ },
957
+ "119": {
958
+ "content": "\n\n\n\n\n\n\n\n\n\n\n\n",
959
+ "lstrip": false,
960
+ "normalized": false,
961
+ "rstrip": false,
962
+ "single_word": false,
963
+ "special": false
964
+ },
965
+ "120": {
966
+ "content": "\n\n\n\n\n\n\n\n\n\n\n\n\n",
967
+ "lstrip": false,
968
+ "normalized": false,
969
+ "rstrip": false,
970
+ "single_word": false,
971
+ "special": false
972
+ },
973
+ "121": {
974
+ "content": "\n\n\n\n\n\n\n\n\n\n\n\n\n\n",
975
+ "lstrip": false,
976
+ "normalized": false,
977
+ "rstrip": false,
978
+ "single_word": false,
979
+ "special": false
980
+ },
981
+ "122": {
982
+ "content": "\n\n\n\n\n\n\n\n\n\n\n\n\n\n\n",
983
+ "lstrip": false,
984
+ "normalized": false,
985
+ "rstrip": false,
986
+ "single_word": false,
987
+ "special": false
988
+ },
989
+ "123": {
990
+ "content": "\n\n\n\n\n\n\n\n\n\n\n\n\n\n\n\n",
991
+ "lstrip": false,
992
+ "normalized": false,
993
+ "rstrip": false,
994
+ "single_word": false,
995
+ "special": false
996
+ },
997
+ "124": {
998
+ "content": "\n\n\n\n\n\n\n\n\n\n\n\n\n\n\n\n\n",
999
+ "lstrip": false,
1000
+ "normalized": false,
1001
+ "rstrip": false,
1002
+ "single_word": false,
1003
+ "special": false
1004
+ },
1005
+ "125": {
1006
+ "content": "\n\n\n\n\n\n\n\n\n\n\n\n\n\n\n\n\n\n",
1007
+ "lstrip": false,
1008
+ "normalized": false,
1009
+ "rstrip": false,
1010
+ "single_word": false,
1011
+ "special": false
1012
+ },
1013
+ "126": {
1014
+ "content": "\n\n\n\n\n\n\n\n\n\n\n\n\n\n\n\n\n\n\n",
1015
+ "lstrip": false,
1016
+ "normalized": false,
1017
+ "rstrip": false,
1018
+ "single_word": false,
1019
+ "special": false
1020
+ },
1021
+ "127": {
1022
+ "content": "\n\n\n\n\n\n\n\n\n\n\n\n\n\n\n\n\n\n\n\n",
1023
+ "lstrip": false,
1024
+ "normalized": false,
1025
+ "rstrip": false,
1026
+ "single_word": false,
1027
+ "special": false
1028
+ },
1029
+ "128": {
1030
+ "content": "\n\n\n\n\n\n\n\n\n\n\n\n\n\n\n\n\n\n\n\n\n",
1031
+ "lstrip": false,
1032
+ "normalized": false,
1033
+ "rstrip": false,
1034
+ "single_word": false,
1035
+ "special": false
1036
+ },
1037
+ "129": {
1038
+ "content": "\n\n\n\n\n\n\n\n\n\n\n\n\n\n\n\n\n\n\n\n\n\n",
1039
+ "lstrip": false,
1040
+ "normalized": false,
1041
+ "rstrip": false,
1042
+ "single_word": false,
1043
+ "special": false
1044
+ },
1045
+ "130": {
1046
+ "content": "\n\n\n\n\n\n\n\n\n\n\n\n\n\n\n\n\n\n\n\n\n\n\n",
1047
+ "lstrip": false,
1048
+ "normalized": false,
1049
+ "rstrip": false,
1050
+ "single_word": false,
1051
+ "special": false
1052
+ },
1053
+ "131": {
1054
+ "content": "\n\n\n\n\n\n\n\n\n\n\n\n\n\n\n\n\n\n\n\n\n\n\n\n",
1055
+ "lstrip": false,
1056
+ "normalized": false,
1057
+ "rstrip": false,
1058
+ "single_word": false,
1059
+ "special": false
1060
+ },
1061
+ "132": {
1062
+ "content": "\n\n\n\n\n\n\n\n\n\n\n\n\n\n\n\n\n\n\n\n\n\n\n\n\n",
1063
+ "lstrip": false,
1064
+ "normalized": false,
1065
+ "rstrip": false,
1066
+ "single_word": false,
1067
+ "special": false
1068
+ },
1069
+ "133": {
1070
+ "content": "\n\n\n\n\n\n\n\n\n\n\n\n\n\n\n\n\n\n\n\n\n\n\n\n\n\n",
1071
+ "lstrip": false,
1072
+ "normalized": false,
1073
+ "rstrip": false,
1074
+ "single_word": false,
1075
+ "special": false
1076
+ },
1077
+ "134": {
1078
+ "content": "\n\n\n\n\n\n\n\n\n\n\n\n\n\n\n\n\n\n\n\n\n\n\n\n\n\n\n",
1079
+ "lstrip": false,
1080
+ "normalized": false,
1081
+ "rstrip": false,
1082
+ "single_word": false,
1083
+ "special": false
1084
+ },
1085
+ "135": {
1086
+ "content": "\n\n\n\n\n\n\n\n\n\n\n\n\n\n\n\n\n\n\n\n\n\n\n\n\n\n\n\n",
1087
+ "lstrip": false,
1088
+ "normalized": false,
1089
+ "rstrip": false,
1090
+ "single_word": false,
1091
+ "special": false
1092
+ },
1093
+ "136": {
1094
+ "content": "\n\n\n\n\n\n\n\n\n\n\n\n\n\n\n\n\n\n\n\n\n\n\n\n\n\n\n\n\n",
1095
+ "lstrip": false,
1096
+ "normalized": false,
1097
+ "rstrip": false,
1098
+ "single_word": false,
1099
+ "special": false
1100
+ },
1101
+ "137": {
1102
+ "content": "\n\n\n\n\n\n\n\n\n\n\n\n\n\n\n\n\n\n\n\n\n\n\n\n\n\n\n\n\n\n",
1103
+ "lstrip": false,
1104
+ "normalized": false,
1105
+ "rstrip": false,
1106
+ "single_word": false,
1107
+ "special": false
1108
+ },
1109
+ "138": {
1110
+ "content": "\n\n\n\n\n\n\n\n\n\n\n\n\n\n\n\n\n\n\n\n\n\n\n\n\n\n\n\n\n\n\n",
1111
+ "lstrip": false,
1112
+ "normalized": false,
1113
+ "rstrip": false,
1114
+ "single_word": false,
1115
+ "special": false
1116
+ },
1117
+ "139": {
1118
+ "content": "▁▁",
1119
+ "lstrip": false,
1120
+ "normalized": false,
1121
+ "rstrip": false,
1122
+ "single_word": false,
1123
+ "special": false
1124
+ },
1125
+ "140": {
1126
+ "content": "▁▁▁",
1127
+ "lstrip": false,
1128
+ "normalized": false,
1129
+ "rstrip": false,
1130
+ "single_word": false,
1131
+ "special": false
1132
+ },
1133
+ "141": {
1134
+ "content": "▁▁▁▁",
1135
+ "lstrip": false,
1136
+ "normalized": false,
1137
+ "rstrip": false,
1138
+ "single_word": false,
1139
+ "special": false
1140
+ },
1141
+ "142": {
1142
+ "content": "▁▁▁▁▁",
1143
+ "lstrip": false,
1144
+ "normalized": false,
1145
+ "rstrip": false,
1146
+ "single_word": false,
1147
+ "special": false
1148
+ },
1149
+ "143": {
1150
+ "content": "▁▁▁▁▁▁",
1151
+ "lstrip": false,
1152
+ "normalized": false,
1153
+ "rstrip": false,
1154
+ "single_word": false,
1155
+ "special": false
1156
+ },
1157
+ "144": {
1158
+ "content": "▁▁▁▁▁▁▁",
1159
+ "lstrip": false,
1160
+ "normalized": false,
1161
+ "rstrip": false,
1162
+ "single_word": false,
1163
+ "special": false
1164
+ },
1165
+ "145": {
1166
+ "content": "▁▁▁▁▁▁▁▁",
1167
+ "lstrip": false,
1168
+ "normalized": false,
1169
+ "rstrip": false,
1170
+ "single_word": false,
1171
+ "special": false
1172
+ },
1173
+ "146": {
1174
+ "content": "▁▁▁▁▁▁▁▁▁",
1175
+ "lstrip": false,
1176
+ "normalized": false,
1177
+ "rstrip": false,
1178
+ "single_word": false,
1179
+ "special": false
1180
+ },
1181
+ "147": {
1182
+ "content": "▁▁▁▁▁▁▁▁▁▁",
1183
+ "lstrip": false,
1184
+ "normalized": false,
1185
+ "rstrip": false,
1186
+ "single_word": false,
1187
+ "special": false
1188
+ },
1189
+ "148": {
1190
+ "content": "▁▁▁▁▁▁▁▁▁▁▁",
1191
+ "lstrip": false,
1192
+ "normalized": false,
1193
+ "rstrip": false,
1194
+ "single_word": false,
1195
+ "special": false
1196
+ },
1197
+ "149": {
1198
+ "content": "▁▁▁▁▁▁▁▁▁▁▁▁",
1199
+ "lstrip": false,
1200
+ "normalized": false,
1201
+ "rstrip": false,
1202
+ "single_word": false,
1203
+ "special": false
1204
+ },
1205
+ "150": {
1206
+ "content": "▁▁▁▁▁▁▁▁▁▁▁▁▁",
1207
+ "lstrip": false,
1208
+ "normalized": false,
1209
+ "rstrip": false,
1210
+ "single_word": false,
1211
+ "special": false
1212
+ },
1213
+ "151": {
1214
+ "content": "▁▁▁▁▁▁▁▁▁▁▁▁▁▁",
1215
+ "lstrip": false,
1216
+ "normalized": false,
1217
+ "rstrip": false,
1218
+ "single_word": false,
1219
+ "special": false
1220
+ },
1221
+ "152": {
1222
+ "content": "▁▁▁▁▁▁▁▁▁▁▁▁▁▁▁",
1223
+ "lstrip": false,
1224
+ "normalized": false,
1225
+ "rstrip": false,
1226
+ "single_word": false,
1227
+ "special": false
1228
+ },
1229
+ "153": {
1230
+ "content": "▁▁▁▁▁▁▁▁▁▁▁▁▁▁▁▁",
1231
+ "lstrip": false,
1232
+ "normalized": false,
1233
+ "rstrip": false,
1234
+ "single_word": false,
1235
+ "special": false
1236
+ },
1237
+ "154": {
1238
+ "content": "▁▁▁▁▁▁▁▁▁▁▁▁▁▁▁▁▁",
1239
+ "lstrip": false,
1240
+ "normalized": false,
1241
+ "rstrip": false,
1242
+ "single_word": false,
1243
+ "special": false
1244
+ },
1245
+ "155": {
1246
+ "content": "▁▁▁▁▁▁▁▁▁▁▁▁▁▁▁▁▁▁",
1247
+ "lstrip": false,
1248
+ "normalized": false,
1249
+ "rstrip": false,
1250
+ "single_word": false,
1251
+ "special": false
1252
+ },
1253
+ "156": {
1254
+ "content": "▁▁▁▁▁▁▁▁▁▁▁▁▁▁▁▁▁▁▁",
1255
+ "lstrip": false,
1256
+ "normalized": false,
1257
+ "rstrip": false,
1258
+ "single_word": false,
1259
+ "special": false
1260
+ },
1261
+ "157": {
1262
+ "content": "▁▁▁▁▁▁▁▁▁▁▁▁▁▁▁▁▁▁▁▁",
1263
+ "lstrip": false,
1264
+ "normalized": false,
1265
+ "rstrip": false,
1266
+ "single_word": false,
1267
+ "special": false
1268
+ },
1269
+ "158": {
1270
+ "content": "▁▁▁▁▁▁▁▁▁▁▁▁▁▁▁▁▁▁▁▁▁",
1271
+ "lstrip": false,
1272
+ "normalized": false,
1273
+ "rstrip": false,
1274
+ "single_word": false,
1275
+ "special": false
1276
+ },
1277
+ "159": {
1278
+ "content": "▁▁▁▁▁▁▁▁▁▁▁▁▁▁▁▁▁▁▁▁▁▁",
1279
+ "lstrip": false,
1280
+ "normalized": false,
1281
+ "rstrip": false,
1282
+ "single_word": false,
1283
+ "special": false
1284
+ },
1285
+ "160": {
1286
+ "content": "▁▁▁▁▁▁▁▁▁▁▁▁▁▁▁▁▁▁▁▁▁▁▁",
1287
+ "lstrip": false,
1288
+ "normalized": false,
1289
+ "rstrip": false,
1290
+ "single_word": false,
1291
+ "special": false
1292
+ },
1293
+ "161": {
1294
+ "content": "▁▁▁▁▁▁▁▁▁▁▁▁▁▁▁▁▁▁▁▁▁▁▁▁",
1295
+ "lstrip": false,
1296
+ "normalized": false,
1297
+ "rstrip": false,
1298
+ "single_word": false,
1299
+ "special": false
1300
+ },
1301
+ "162": {
1302
+ "content": "▁▁▁▁▁▁▁▁▁▁▁▁▁▁▁▁▁▁▁▁▁▁▁▁▁",
1303
+ "lstrip": false,
1304
+ "normalized": false,
1305
+ "rstrip": false,
1306
+ "single_word": false,
1307
+ "special": false
1308
+ },
1309
+ "163": {
1310
+ "content": "▁▁▁▁▁▁▁▁▁▁▁▁▁▁▁▁▁▁▁▁▁▁▁▁▁▁",
1311
+ "lstrip": false,
1312
+ "normalized": false,
1313
+ "rstrip": false,
1314
+ "single_word": false,
1315
+ "special": false
1316
+ },
1317
+ "164": {
1318
+ "content": "▁▁▁▁▁▁▁▁▁▁▁▁▁▁▁▁▁▁▁▁▁▁▁▁▁▁▁",
1319
+ "lstrip": false,
1320
+ "normalized": false,
1321
+ "rstrip": false,
1322
+ "single_word": false,
1323
+ "special": false
1324
+ },
1325
+ "165": {
1326
+ "content": "▁▁▁▁▁▁▁▁▁▁▁▁▁▁▁▁▁▁▁▁▁▁▁▁▁▁▁▁",
1327
+ "lstrip": false,
1328
+ "normalized": false,
1329
+ "rstrip": false,
1330
+ "single_word": false,
1331
+ "special": false
1332
+ },
1333
+ "166": {
1334
+ "content": "▁▁▁▁▁▁▁▁▁▁▁▁▁▁▁▁▁▁▁▁▁▁▁▁▁▁▁▁▁",
1335
+ "lstrip": false,
1336
+ "normalized": false,
1337
+ "rstrip": false,
1338
+ "single_word": false,
1339
+ "special": false
1340
+ },
1341
+ "167": {
1342
+ "content": "▁▁▁▁▁▁▁▁▁▁▁▁▁▁▁▁▁▁▁▁▁▁▁▁▁▁▁▁▁▁",
1343
+ "lstrip": false,
1344
+ "normalized": false,
1345
+ "rstrip": false,
1346
+ "single_word": false,
1347
+ "special": false
1348
+ },
1349
+ "168": {
1350
+ "content": "▁▁▁▁▁▁▁▁▁▁▁▁▁▁▁▁▁▁▁▁▁▁▁▁▁▁▁▁▁▁▁",
1351
+ "lstrip": false,
1352
+ "normalized": false,
1353
+ "rstrip": false,
1354
+ "single_word": false,
1355
+ "special": false
1356
+ },
1357
+ "169": {
1358
+ "content": "<table>",
1359
+ "lstrip": false,
1360
+ "normalized": false,
1361
+ "rstrip": false,
1362
+ "single_word": false,
1363
+ "special": false
1364
+ },
1365
+ "170": {
1366
+ "content": "<caption>",
1367
+ "lstrip": false,
1368
+ "normalized": false,
1369
+ "rstrip": false,
1370
+ "single_word": false,
1371
+ "special": false
1372
+ },
1373
+ "171": {
1374
+ "content": "<thead>",
1375
+ "lstrip": false,
1376
+ "normalized": false,
1377
+ "rstrip": false,
1378
+ "single_word": false,
1379
+ "special": false
1380
+ },
1381
+ "172": {
1382
+ "content": "<tbody>",
1383
+ "lstrip": false,
1384
+ "normalized": false,
1385
+ "rstrip": false,
1386
+ "single_word": false,
1387
+ "special": false
1388
+ },
1389
+ "173": {
1390
+ "content": "<tfoot>",
1391
+ "lstrip": false,
1392
+ "normalized": false,
1393
+ "rstrip": false,
1394
+ "single_word": false,
1395
+ "special": false
1396
+ },
1397
+ "174": {
1398
+ "content": "<tr>",
1399
+ "lstrip": false,
1400
+ "normalized": false,
1401
+ "rstrip": false,
1402
+ "single_word": false,
1403
+ "special": false
1404
+ },
1405
+ "175": {
1406
+ "content": "<th>",
1407
+ "lstrip": false,
1408
+ "normalized": false,
1409
+ "rstrip": false,
1410
+ "single_word": false,
1411
+ "special": false
1412
+ },
1413
+ "176": {
1414
+ "content": "<td>",
1415
+ "lstrip": false,
1416
+ "normalized": false,
1417
+ "rstrip": false,
1418
+ "single_word": false,
1419
+ "special": false
1420
+ },
1421
+ "177": {
1422
+ "content": "</table>",
1423
+ "lstrip": false,
1424
+ "normalized": false,
1425
+ "rstrip": false,
1426
+ "single_word": false,
1427
+ "special": false
1428
+ },
1429
+ "178": {
1430
+ "content": "</caption>",
1431
+ "lstrip": false,
1432
+ "normalized": false,
1433
+ "rstrip": false,
1434
+ "single_word": false,
1435
+ "special": false
1436
+ },
1437
+ "179": {
1438
+ "content": "</thead>",
1439
+ "lstrip": false,
1440
+ "normalized": false,
1441
+ "rstrip": false,
1442
+ "single_word": false,
1443
+ "special": false
1444
+ },
1445
+ "180": {
1446
+ "content": "</tbody>",
1447
+ "lstrip": false,
1448
+ "normalized": false,
1449
+ "rstrip": false,
1450
+ "single_word": false,
1451
+ "special": false
1452
+ },
1453
+ "181": {
1454
+ "content": "</tfoot>",
1455
+ "lstrip": false,
1456
+ "normalized": false,
1457
+ "rstrip": false,
1458
+ "single_word": false,
1459
+ "special": false
1460
+ },
1461
+ "182": {
1462
+ "content": "</tr>",
1463
+ "lstrip": false,
1464
+ "normalized": false,
1465
+ "rstrip": false,
1466
+ "single_word": false,
1467
+ "special": false
1468
+ },
1469
+ "183": {
1470
+ "content": "</th>",
1471
+ "lstrip": false,
1472
+ "normalized": false,
1473
+ "rstrip": false,
1474
+ "single_word": false,
1475
+ "special": false
1476
+ },
1477
+ "184": {
1478
+ "content": "</td>",
1479
+ "lstrip": false,
1480
+ "normalized": false,
1481
+ "rstrip": false,
1482
+ "single_word": false,
1483
+ "special": false
1484
+ },
1485
+ "185": {
1486
+ "content": "<h1>",
1487
+ "lstrip": false,
1488
+ "normalized": false,
1489
+ "rstrip": false,
1490
+ "single_word": false,
1491
+ "special": false
1492
+ },
1493
+ "186": {
1494
+ "content": "<h2>",
1495
+ "lstrip": false,
1496
+ "normalized": false,
1497
+ "rstrip": false,
1498
+ "single_word": false,
1499
+ "special": false
1500
+ },
1501
+ "187": {
1502
+ "content": "<h3>",
1503
+ "lstrip": false,
1504
+ "normalized": false,
1505
+ "rstrip": false,
1506
+ "single_word": false,
1507
+ "special": false
1508
+ },
1509
+ "188": {
1510
+ "content": "<h4>",
1511
+ "lstrip": false,
1512
+ "normalized": false,
1513
+ "rstrip": false,
1514
+ "single_word": false,
1515
+ "special": false
1516
+ },
1517
+ "189": {
1518
+ "content": "<h5>",
1519
+ "lstrip": false,
1520
+ "normalized": false,
1521
+ "rstrip": false,
1522
+ "single_word": false,
1523
+ "special": false
1524
+ },
1525
+ "190": {
1526
+ "content": "<h6>",
1527
+ "lstrip": false,
1528
+ "normalized": false,
1529
+ "rstrip": false,
1530
+ "single_word": false,
1531
+ "special": false
1532
+ },
1533
+ "191": {
1534
+ "content": "<blockquote>",
1535
+ "lstrip": false,
1536
+ "normalized": false,
1537
+ "rstrip": false,
1538
+ "single_word": false,
1539
+ "special": false
1540
+ },
1541
+ "192": {
1542
+ "content": "</h1>",
1543
+ "lstrip": false,
1544
+ "normalized": false,
1545
+ "rstrip": false,
1546
+ "single_word": false,
1547
+ "special": false
1548
+ },
1549
+ "193": {
1550
+ "content": "</h2>",
1551
+ "lstrip": false,
1552
+ "normalized": false,
1553
+ "rstrip": false,
1554
+ "single_word": false,
1555
+ "special": false
1556
+ },
1557
+ "194": {
1558
+ "content": "</h3>",
1559
+ "lstrip": false,
1560
+ "normalized": false,
1561
+ "rstrip": false,
1562
+ "single_word": false,
1563
+ "special": false
1564
+ },
1565
+ "195": {
1566
+ "content": "</h4>",
1567
+ "lstrip": false,
1568
+ "normalized": false,
1569
+ "rstrip": false,
1570
+ "single_word": false,
1571
+ "special": false
1572
+ },
1573
+ "196": {
1574
+ "content": "</h5>",
1575
+ "lstrip": false,
1576
+ "normalized": false,
1577
+ "rstrip": false,
1578
+ "single_word": false,
1579
+ "special": false
1580
+ },
1581
+ "197": {
1582
+ "content": "</h6>",
1583
+ "lstrip": false,
1584
+ "normalized": false,
1585
+ "rstrip": false,
1586
+ "single_word": false,
1587
+ "special": false
1588
+ },
1589
+ "198": {
1590
+ "content": "</blockquote>",
1591
+ "lstrip": false,
1592
+ "normalized": false,
1593
+ "rstrip": false,
1594
+ "single_word": false,
1595
+ "special": false
1596
+ },
1597
+ "199": {
1598
+ "content": "<strong>",
1599
+ "lstrip": false,
1600
+ "normalized": false,
1601
+ "rstrip": false,
1602
+ "single_word": false,
1603
+ "special": false
1604
+ },
1605
+ "200": {
1606
+ "content": "<em>",
1607
+ "lstrip": false,
1608
+ "normalized": false,
1609
+ "rstrip": false,
1610
+ "single_word": false,
1611
+ "special": false
1612
+ },
1613
+ "201": {
1614
+ "content": "<b>",
1615
+ "lstrip": false,
1616
+ "normalized": false,
1617
+ "rstrip": false,
1618
+ "single_word": false,
1619
+ "special": false
1620
+ },
1621
+ "202": {
1622
+ "content": "<i>",
1623
+ "lstrip": false,
1624
+ "normalized": false,
1625
+ "rstrip": false,
1626
+ "single_word": false,
1627
+ "special": false
1628
+ },
1629
+ "203": {
1630
+ "content": "<u>",
1631
+ "lstrip": false,
1632
+ "normalized": false,
1633
+ "rstrip": false,
1634
+ "single_word": false,
1635
+ "special": false
1636
+ },
1637
+ "204": {
1638
+ "content": "<s>",
1639
+ "lstrip": false,
1640
+ "normalized": false,
1641
+ "rstrip": false,
1642
+ "single_word": false,
1643
+ "special": false
1644
+ },
1645
+ "205": {
1646
+ "content": "<sub>",
1647
+ "lstrip": false,
1648
+ "normalized": false,
1649
+ "rstrip": false,
1650
+ "single_word": false,
1651
+ "special": false
1652
+ },
1653
+ "206": {
1654
+ "content": "<sup>",
1655
+ "lstrip": false,
1656
+ "normalized": false,
1657
+ "rstrip": false,
1658
+ "single_word": false,
1659
+ "special": false
1660
+ },
1661
+ "207": {
1662
+ "content": "<code>",
1663
+ "lstrip": false,
1664
+ "normalized": false,
1665
+ "rstrip": false,
1666
+ "single_word": false,
1667
+ "special": false
1668
+ },
1669
+ "208": {
1670
+ "content": "</strong>",
1671
+ "lstrip": false,
1672
+ "normalized": false,
1673
+ "rstrip": false,
1674
+ "single_word": false,
1675
+ "special": false
1676
+ },
1677
+ "209": {
1678
+ "content": "</em>",
1679
+ "lstrip": false,
1680
+ "normalized": false,
1681
+ "rstrip": false,
1682
+ "single_word": false,
1683
+ "special": false
1684
+ },
1685
+ "210": {
1686
+ "content": "</b>",
1687
+ "lstrip": false,
1688
+ "normalized": false,
1689
+ "rstrip": false,
1690
+ "single_word": false,
1691
+ "special": false
1692
+ },
1693
+ "211": {
1694
+ "content": "</i>",
1695
+ "lstrip": false,
1696
+ "normalized": false,
1697
+ "rstrip": false,
1698
+ "single_word": false,
1699
+ "special": false
1700
+ },
1701
+ "212": {
1702
+ "content": "</u>",
1703
+ "lstrip": false,
1704
+ "normalized": false,
1705
+ "rstrip": false,
1706
+ "single_word": false,
1707
+ "special": false
1708
+ },
1709
+ "213": {
1710
+ "content": "</s>",
1711
+ "lstrip": false,
1712
+ "normalized": false,
1713
+ "rstrip": false,
1714
+ "single_word": false,
1715
+ "special": false
1716
+ },
1717
+ "214": {
1718
+ "content": "</sub>",
1719
+ "lstrip": false,
1720
+ "normalized": false,
1721
+ "rstrip": false,
1722
+ "single_word": false,
1723
+ "special": false
1724
+ },
1725
+ "215": {
1726
+ "content": "</sup>",
1727
+ "lstrip": false,
1728
+ "normalized": false,
1729
+ "rstrip": false,
1730
+ "single_word": false,
1731
+ "special": false
1732
+ },
1733
+ "216": {
1734
+ "content": "</code>",
1735
+ "lstrip": false,
1736
+ "normalized": false,
1737
+ "rstrip": false,
1738
+ "single_word": false,
1739
+ "special": false
1740
+ }
1741
+ },
1742
+ "additional_special_tokens": [
1743
+ "<start_of_turn>",
1744
+ "<end_of_turn>"
1745
+ ],
1746
+ "bos_token": "<bos>",
1747
+ "chat_template": "{{ bos_token }}{% if messages[0]['role'] == 'system' %}{{ raise_exception('System role not supported') }}{% endif %}{% for message in messages %}{% if (message['role'] == 'user') != (loop.index0 % 2 == 0) %}{{ raise_exception('Conversation roles must alternate user/assistant/user/assistant/...') }}{% endif %}{% if (message['role'] == 'assistant') %}{% set role = 'model' %}{% else %}{% set role = message['role'] %}{% endif %}{{ '<start_of_turn>' + role + '\n' + message['content'] | trim + '<end_of_turn>\n' }}{% endfor %}{% if add_generation_prompt %}{{'<start_of_turn>model\n'}}{% endif %}",
1748
+ "clean_up_tokenization_spaces": false,
1749
+ "eos_token": "<eos>",
1750
+ "model_max_length": 1000000000000000019884624838656,
1751
+ "pad_token": "<pad>",
1752
+ "sp_model_kwargs": {},
1753
+ "spaces_between_special_tokens": false,
1754
+ "tokenizer_class": "GemmaTokenizer",
1755
+ "unk_token": "<unk>",
1756
+ "use_default_system_prompt": false
1757
+ }