Hugging Face's logo

qnguyen3
/
colqwen2_5-v0.2-hf

Transformers
Safetensors
colqwen2_5
Model card Files
xet
Community
qnguyen3 commited on
Commit
235b460
·
verified ·
1 Parent(s): b39a114

Upload ColQwen2_5ForRetrieval

Browse files
Files changed (4) hide show
  1. config.json +218 -56
  2. model-00001-of-00002.safetensors +2 -2
  3. model-00002-of-00002.safetensors +2 -2
  4. model.safetensors.index.json +0 -0
config.json CHANGED
@@ -1,65 +1,227 @@
1
  {
2
  "architectures": [
3
- "ColQwen2_5"
4
  ],
5
- "attention_dropout": 0.0,
6
- "bos_token_id": 151643,
7
- "eos_token_id": 151645,
8
- "hidden_act": "silu",
9
- "hidden_size": 2048,
10
- "image_token_id": 151655,
11
  "initializer_range": 0.02,
12
- "intermediate_size": 11008,
13
- "max_position_embeddings": 128000,
14
- "max_window_layers": 70,
15
- "model_type": "qwen2_5_vl",
16
- "num_attention_heads": 16,
17
- "num_hidden_layers": 36,
18
- "num_key_value_heads": 2,
19
- "rms_norm_eps": 1e-06,
20
- "rope_scaling": {
21
- "mrope_section": [
22
- 16,
23
- 24,
24
- 24
25
- ],
26
- "rope_type": "default",
27
- "type": "default"
28
- },
29
- "rope_theta": 1000000.0,
30
- "sliding_window": 32768,
31
- "tie_word_embeddings": true,
32
  "torch_dtype": "bfloat16",
33
- "transformers_version": "4.51.3",
34
- "use_cache": true,
35
- "use_sliding_window": false,
36
- "video_token_id": 151656,
37
- "vision_config": {
38
- "depth": 32,
39
- "fullatt_block_indexes": [
40
- 7,
41
- 15,
42
- 23,
43
- 31
44
- ],
45
  "hidden_act": "silu",
46
- "hidden_size": 1280,
47
- "in_channels": 3,
48
- "in_chans": 3,
49
- "intermediate_size": 3420,
 
 
50
  "model_type": "qwen2_5_vl",
51
- "num_heads": 16,
52
- "out_hidden_size": 2048,
53
- "patch_size": 14,
54
- "spatial_merge_size": 2,
55
- "spatial_patch_size": 14,
56
- "temporal_patch_size": 2,
57
- "tokens_per_second": 2,
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
58
  "torch_dtype": "bfloat16",
59
- "window_size": 112
60
- },
61
- "vision_end_token_id": 151653,
62
- "vision_start_token_id": 151652,
63
- "vision_token_id": 151654,
64
- "vocab_size": 151936
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
65
  }
 
1
  {
2
  "architectures": [
3
+ "ColQwen2_5ForRetrieval"
4
  ],
5
+ "embedding_dim": 128,
 
 
 
 
 
6
  "initializer_range": 0.02,
7
+ "is_composition": false,
8
+ "model_type": "colqwen2_5",
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
9
  "torch_dtype": "bfloat16",
10
+ "transformers_version": "4.53.0.dev0",
11
+ "vlm_config": {
12
+ "_name_or_path": "Qwen/Qwen2.5-VL-3B-Instruct",
13
+ "attention_dropout": 0.0,
14
+ "bos_token_id": 151643,
15
+ "eos_token_id": 151645,
 
 
 
 
 
 
16
  "hidden_act": "silu",
17
+ "hidden_size": 2048,
18
+ "image_token_id": 151655,
19
+ "initializer_range": 0.02,
20
+ "intermediate_size": 11008,
21
+ "max_position_embeddings": 128000,
22
+ "max_window_layers": 70,
23
  "model_type": "qwen2_5_vl",
24
+ "num_attention_heads": 16,
25
+ "num_hidden_layers": 36,
26
+ "num_key_value_heads": 2,
27
+ "rms_norm_eps": 1e-06,
28
+ "rope_scaling": {
29
+ "mrope_section": [
30
+ 16,
31
+ 24,
32
+ 24
33
+ ],
34
+ "rope_type": "default",
35
+ "type": "default"
36
+ },
37
+ "rope_theta": 1000000.0,
38
+ "sliding_window": 32768,
39
+ "text_config": {
40
+ "_name_or_path": "",
41
+ "add_cross_attention": false,
42
+ "architectures": [
43
+ "ColQwen2_5"
44
+ ],
45
+ "attention_dropout": 0.0,
46
+ "bad_words_ids": null,
47
+ "begin_suppress_tokens": null,
48
+ "bos_token_id": 151643,
49
+ "chunk_size_feed_forward": 0,
50
+ "cross_attention_hidden_size": null,
51
+ "decoder_start_token_id": null,
52
+ "diversity_penalty": 0.0,
53
+ "do_sample": false,
54
+ "early_stopping": false,
55
+ "encoder_no_repeat_ngram_size": 0,
56
+ "eos_token_id": 151645,
57
+ "exponential_decay_length_penalty": null,
58
+ "finetuning_task": null,
59
+ "forced_bos_token_id": null,
60
+ "forced_eos_token_id": null,
61
+ "hidden_act": "silu",
62
+ "hidden_size": 2048,
63
+ "id2label": {
64
+ "0": "LABEL_0",
65
+ "1": "LABEL_1"
66
+ },
67
+ "image_token_id": null,
68
+ "initializer_range": 0.02,
69
+ "intermediate_size": 11008,
70
+ "is_decoder": false,
71
+ "is_encoder_decoder": false,
72
+ "label2id": {
73
+ "LABEL_0": 0,
74
+ "LABEL_1": 1
75
+ },
76
+ "length_penalty": 1.0,
77
+ "max_length": 20,
78
+ "max_position_embeddings": 128000,
79
+ "max_window_layers": 70,
80
+ "min_length": 0,
81
+ "model_type": "qwen2_5_vl_text",
82
+ "no_repeat_ngram_size": 0,
83
+ "num_attention_heads": 16,
84
+ "num_beam_groups": 1,
85
+ "num_beams": 1,
86
+ "num_hidden_layers": 36,
87
+ "num_key_value_heads": 2,
88
+ "num_return_sequences": 1,
89
+ "output_attentions": false,
90
+ "output_hidden_states": false,
91
+ "output_scores": false,
92
+ "pad_token_id": null,
93
+ "prefix": null,
94
+ "problem_type": null,
95
+ "pruned_heads": {},
96
+ "remove_invalid_values": false,
97
+ "repetition_penalty": 1.0,
98
+ "return_dict": true,
99
+ "return_dict_in_generate": false,
100
+ "rms_norm_eps": 1e-06,
101
+ "rope_scaling": {
102
+ "mrope_section": [
103
+ 16,
104
+ 24,
105
+ 24
106
+ ],
107
+ "rope_type": "default",
108
+ "type": "default"
109
+ },
110
+ "rope_theta": 1000000.0,
111
+ "sep_token_id": null,
112
+ "sliding_window": 32768,
113
+ "suppress_tokens": null,
114
+ "task_specific_params": null,
115
+ "temperature": 1.0,
116
+ "tf_legacy_loss": false,
117
+ "tie_encoder_decoder": false,
118
+ "tie_word_embeddings": true,
119
+ "tokenizer_class": null,
120
+ "top_k": 50,
121
+ "top_p": 1.0,
122
+ "torch_dtype": "bfloat16",
123
+ "torchscript": false,
124
+ "typical_p": 1.0,
125
+ "use_bfloat16": false,
126
+ "use_cache": true,
127
+ "use_sliding_window": false,
128
+ "video_token_id": null,
129
+ "vision_end_token_id": 151653,
130
+ "vision_start_token_id": 151652,
131
+ "vision_token_id": 151654,
132
+ "vocab_size": 151936
133
+ },
134
  "torch_dtype": "bfloat16",
135
+ "use_cache": true,
136
+ "use_sliding_window": false,
137
+ "video_token_id": 151656,
138
+ "vision_config": {
139
+ "_name_or_path": "",
140
+ "add_cross_attention": false,
141
+ "architectures": null,
142
+ "bad_words_ids": null,
143
+ "begin_suppress_tokens": null,
144
+ "bos_token_id": null,
145
+ "chunk_size_feed_forward": 0,
146
+ "cross_attention_hidden_size": null,
147
+ "decoder_start_token_id": null,
148
+ "depth": 32,
149
+ "diversity_penalty": 0.0,
150
+ "do_sample": false,
151
+ "early_stopping": false,
152
+ "encoder_no_repeat_ngram_size": 0,
153
+ "eos_token_id": null,
154
+ "exponential_decay_length_penalty": null,
155
+ "finetuning_task": null,
156
+ "forced_bos_token_id": null,
157
+ "forced_eos_token_id": null,
158
+ "fullatt_block_indexes": [
159
+ 7,
160
+ 15,
161
+ 23,
162
+ 31
163
+ ],
164
+ "hidden_act": "silu",
165
+ "hidden_size": 1280,
166
+ "id2label": {
167
+ "0": "LABEL_0",
168
+ "1": "LABEL_1"
169
+ },
170
+ "in_channels": 3,
171
+ "in_chans": 3,
172
+ "initializer_range": 0.02,
173
+ "intermediate_size": 3420,
174
+ "is_decoder": false,
175
+ "is_encoder_decoder": false,
176
+ "label2id": {
177
+ "LABEL_0": 0,
178
+ "LABEL_1": 1
179
+ },
180
+ "length_penalty": 1.0,
181
+ "max_length": 20,
182
+ "min_length": 0,
183
+ "model_type": "qwen2_5_vl",
184
+ "no_repeat_ngram_size": 0,
185
+ "num_beam_groups": 1,
186
+ "num_beams": 1,
187
+ "num_heads": 16,
188
+ "num_return_sequences": 1,
189
+ "out_hidden_size": 2048,
190
+ "output_attentions": false,
191
+ "output_hidden_states": false,
192
+ "output_scores": false,
193
+ "pad_token_id": null,
194
+ "patch_size": 14,
195
+ "prefix": null,
196
+ "problem_type": null,
197
+ "pruned_heads": {},
198
+ "remove_invalid_values": false,
199
+ "repetition_penalty": 1.0,
200
+ "return_dict": true,
201
+ "return_dict_in_generate": false,
202
+ "sep_token_id": null,
203
+ "spatial_merge_size": 2,
204
+ "spatial_patch_size": 14,
205
+ "suppress_tokens": null,
206
+ "task_specific_params": null,
207
+ "temperature": 1.0,
208
+ "temporal_patch_size": 2,
209
+ "tf_legacy_loss": false,
210
+ "tie_encoder_decoder": false,
211
+ "tie_word_embeddings": true,
212
+ "tokenizer_class": null,
213
+ "tokens_per_second": 2,
214
+ "top_k": 50,
215
+ "top_p": 1.0,
216
+ "torch_dtype": "bfloat16",
217
+ "torchscript": false,
218
+ "typical_p": 1.0,
219
+ "use_bfloat16": false,
220
+ "window_size": 112
221
+ },
222
+ "vision_end_token_id": 151653,
223
+ "vision_start_token_id": 151652,
224
+ "vision_token_id": 151654,
225
+ "vocab_size": 151936
226
+ }
227
  }
model-00001-of-00002.safetensors CHANGED
@@ -1,3 +1,3 @@
1
  version https://git-lfs.github.com/spec/v1
2
- oid sha256:e36caeaa49db50be5795471f0aaee792a64f15c9bf196589aee34c912016d6f7
3
- size 4997750760
 
1
  version https://git-lfs.github.com/spec/v1
2
+ oid sha256:46b4c359abc980a09b547f339ec238fae2685d2bf28e8e8a86945bdb8ae68b59
3
+ size 4997759176
model-00002-of-00002.safetensors CHANGED
@@ -1,3 +1,3 @@
1
  version https://git-lfs.github.com/spec/v1
2
- oid sha256:034a413aec84ee5504a7d85e801b34ba75033014bd2090678d77ec2d6db43d7a
3
- size 2512111904
 
1
  version https://git-lfs.github.com/spec/v1
2
+ oid sha256:f938b787dbea325fa78dc54c494c028cb06fbcf19a5d4cc1c2f44a57990adf70
3
+ size 2512115640
model.safetensors.index.json CHANGED
The diff for this file is too large to render. See raw diff