Hugging Face's logo

isaiahbjork
/
ShowUI-2B-mlx-4-bit

MLX
Safetensors
qwen2_vl
GUI agents
vision-language-action model
computer use
8-bit precision
Model card Files Community
isaiahbjork commited on
Commit
61d892b
·
verified ·
1 Parent(s): 68b5983

Update config.json

Browse files
Files changed (1) hide show
  1. config.json +347 -43
config.json CHANGED
@@ -1,48 +1,352 @@
1
  {
2
- "_name_or_path": "Qwen/Qwen2-VL-2B-Instruct",
3
- "architectures": [
4
- "Qwen2VLForConditionalGeneration"
5
- ],
6
- "attention_dropout": 0.0,
7
- "bos_token_id": 151643,
8
- "eos_token_id": 151645,
9
- "hidden_act": "silu",
10
- "hidden_size": 1536,
11
- "image_token_id": 151655,
12
- "initializer_range": 0.02,
13
- "intermediate_size": 8960,
14
- "max_position_embeddings": 32768,
15
- "max_window_layers": 28,
16
- "model_type": "qwen2_vl",
17
- "num_attention_heads": 12,
18
- "num_hidden_layers": 28,
19
- "num_key_value_heads": 2,
20
- "rms_norm_eps": 1e-06,
21
- "rope_scaling": {
22
- "mrope_section": [
23
- 16,
24
- 24,
25
- 24
26
  ],
27
- "type": "mrope"
28
- },
29
- "rope_theta": 1000000.0,
30
- "sliding_window": 32768,
31
- "tie_word_embeddings": true,
32
- "tokenizer_model_max_length": 4096,
33
- "torch_dtype": "bfloat16",
34
- "transformers_version": "4.45.0.dev0",
35
- "use_cache": false,
36
- "use_sliding_window": false,
37
- "video_token_id": 151656,
38
- "vision_config": {
 
 
 
 
 
39
  "hidden_size": 1536,
40
- "in_chans": 3,
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
41
  "model_type": "qwen2_vl",
42
- "spatial_patch_size": 14
43
- },
44
- "vision_end_token_id": 151653,
45
- "vision_start_token_id": 151652,
46
- "vision_token_id": 151654,
47
- "vocab_size": 151936
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
48
  }
 
1
  {
2
+ "_attn_implementation_autoset": true,
3
+ "add_cross_attention": false,
4
+ "architectures": [
5
+ "Qwen2VLForConditionalGeneration"
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
6
  ],
7
+ "attention_dropout": 0.0,
8
+ "bad_words_ids": null,
9
+ "begin_suppress_tokens": null,
10
+ "bos_token_id": 151643,
11
+ "chunk_size_feed_forward": 0,
12
+ "cross_attention_hidden_size": null,
13
+ "decoder_start_token_id": null,
14
+ "diversity_penalty": 0.0,
15
+ "do_sample": false,
16
+ "early_stopping": false,
17
+ "encoder_no_repeat_ngram_size": 0,
18
+ "eos_token_id": 151645,
19
+ "exponential_decay_length_penalty": null,
20
+ "finetuning_task": null,
21
+ "forced_bos_token_id": null,
22
+ "forced_eos_token_id": null,
23
+ "hidden_act": "silu",
24
  "hidden_size": 1536,
25
+ "id2label": {
26
+ "0": "LABEL_0",
27
+ "1": "LABEL_1"
28
+ },
29
+ "image_token_id": 151655,
30
+ "initializer_range": 0.02,
31
+ "intermediate_size": 8960,
32
+ "is_decoder": false,
33
+ "is_encoder_decoder": false,
34
+ "label2id": {
35
+ "LABEL_0": 0,
36
+ "LABEL_1": 1
37
+ },
38
+ "length_penalty": 1.0,
39
+ "max_length": 20,
40
+ "max_position_embeddings": 32768,
41
+ "max_window_layers": 28,
42
+ "min_length": 0,
43
  "model_type": "qwen2_vl",
44
+ "no_repeat_ngram_size": 0,
45
+ "num_attention_heads": 12,
46
+ "num_beam_groups": 1,
47
+ "num_beams": 1,
48
+ "num_hidden_layers": 28,
49
+ "num_key_value_heads": 2,
50
+ "num_return_sequences": 1,
51
+ "output_attentions": false,
52
+ "output_hidden_states": false,
53
+ "output_scores": false,
54
+ "pad_token_id": null,
55
+ "prefix": null,
56
+ "problem_type": null,
57
+ "pruned_heads": {},
58
+ "quantization": {
59
+ "group_size": 64,
60
+ "bits": 4
61
+ },
62
+ "remove_invalid_values": false,
63
+ "repetition_penalty": 1.0,
64
+ "return_dict": true,
65
+ "return_dict_in_generate": false,
66
+ "rms_norm_eps": 1e-06,
67
+ "rope_scaling": {
68
+ "mrope_section": [
69
+ 16,
70
+ 24,
71
+ 24
72
+ ],
73
+ "type": "default",
74
+ "rope_type": "default"
75
+ },
76
+ "rope_theta": 1000000.0,
77
+ "rope_traditional": true,
78
+ "sep_token_id": null,
79
+ "sliding_window": 32768,
80
+ "suppress_tokens": null,
81
+ "task_specific_params": null,
82
+ "temperature": 1.0,
83
+ "text_config": {
84
+ "vision_config": {
85
+ "return_dict": true,
86
+ "output_hidden_states": false,
87
+ "output_attentions": false,
88
+ "torchscript": false,
89
+ "torch_dtype": null,
90
+ "use_bfloat16": false,
91
+ "tf_legacy_loss": false,
92
+ "pruned_heads": {},
93
+ "tie_word_embeddings": true,
94
+ "chunk_size_feed_forward": 0,
95
+ "is_encoder_decoder": false,
96
+ "is_decoder": false,
97
+ "cross_attention_hidden_size": null,
98
+ "add_cross_attention": false,
99
+ "tie_encoder_decoder": false,
100
+ "max_length": 20,
101
+ "min_length": 0,
102
+ "do_sample": false,
103
+ "early_stopping": false,
104
+ "num_beams": 1,
105
+ "num_beam_groups": 1,
106
+ "diversity_penalty": 0.0,
107
+ "temperature": 1.0,
108
+ "top_k": 50,
109
+ "top_p": 1.0,
110
+ "typical_p": 1.0,
111
+ "repetition_penalty": 1.0,
112
+ "length_penalty": 1.0,
113
+ "no_repeat_ngram_size": 0,
114
+ "encoder_no_repeat_ngram_size": 0,
115
+ "bad_words_ids": null,
116
+ "num_return_sequences": 1,
117
+ "output_scores": false,
118
+ "return_dict_in_generate": false,
119
+ "forced_bos_token_id": null,
120
+ "forced_eos_token_id": null,
121
+ "remove_invalid_values": false,
122
+ "exponential_decay_length_penalty": null,
123
+ "suppress_tokens": null,
124
+ "begin_suppress_tokens": null,
125
+ "architectures": null,
126
+ "finetuning_task": null,
127
+ "id2label": {
128
+ "0": "LABEL_0",
129
+ "1": "LABEL_1"
130
+ },
131
+ "label2id": {
132
+ "LABEL_0": 0,
133
+ "LABEL_1": 1
134
+ },
135
+ "tokenizer_class": null,
136
+ "prefix": null,
137
+ "bos_token_id": null,
138
+ "pad_token_id": null,
139
+ "eos_token_id": null,
140
+ "sep_token_id": null,
141
+ "decoder_start_token_id": null,
142
+ "task_specific_params": null,
143
+ "problem_type": null,
144
+ "_name_or_path": "",
145
+ "_attn_implementation_autoset": false,
146
+ "in_chans": 3,
147
+ "model_type": "qwen2_vl",
148
+ "spatial_patch_size": 14,
149
+ "depth": 32,
150
+ "embed_dim": 1280,
151
+ "hidden_size": 1536,
152
+ "hidden_act": "quick_gelu",
153
+ "mlp_ratio": 4,
154
+ "num_heads": 16,
155
+ "in_channels": 3,
156
+ "patch_size": 14,
157
+ "spatial_merge_size": 2,
158
+ "temporal_patch_size": 2
159
+ },
160
+ "vocab_size": 151936,
161
+ "max_position_embeddings": 32768,
162
+ "hidden_size": 1536,
163
+ "intermediate_size": 8960,
164
+ "num_hidden_layers": 28,
165
+ "num_attention_heads": 12,
166
+ "use_sliding_window": false,
167
+ "sliding_window": 32768,
168
+ "max_window_layers": 28,
169
+ "num_key_value_heads": 2,
170
+ "hidden_act": "silu",
171
+ "initializer_range": 0.02,
172
+ "rms_norm_eps": 1e-06,
173
+ "use_cache": false,
174
+ "rope_theta": 1000000.0,
175
+ "attention_dropout": 0.0,
176
+ "rope_scaling": {
177
+ "mrope_section": [
178
+ 16,
179
+ 24,
180
+ 24
181
+ ],
182
+ "type": "default",
183
+ "rope_type": "default"
184
+ },
185
+ "return_dict": true,
186
+ "output_hidden_states": false,
187
+ "output_attentions": false,
188
+ "torchscript": false,
189
+ "torch_dtype": "bfloat16",
190
+ "use_bfloat16": false,
191
+ "tf_legacy_loss": false,
192
+ "pruned_heads": {},
193
+ "tie_word_embeddings": true,
194
+ "chunk_size_feed_forward": 0,
195
+ "is_encoder_decoder": false,
196
+ "is_decoder": false,
197
+ "cross_attention_hidden_size": null,
198
+ "add_cross_attention": false,
199
+ "tie_encoder_decoder": false,
200
+ "max_length": 20,
201
+ "min_length": 0,
202
+ "do_sample": false,
203
+ "early_stopping": false,
204
+ "num_beams": 1,
205
+ "num_beam_groups": 1,
206
+ "diversity_penalty": 0.0,
207
+ "temperature": 1.0,
208
+ "top_k": 50,
209
+ "top_p": 1.0,
210
+ "typical_p": 1.0,
211
+ "repetition_penalty": 1.0,
212
+ "length_penalty": 1.0,
213
+ "no_repeat_ngram_size": 0,
214
+ "encoder_no_repeat_ngram_size": 0,
215
+ "bad_words_ids": null,
216
+ "num_return_sequences": 1,
217
+ "output_scores": false,
218
+ "return_dict_in_generate": false,
219
+ "forced_bos_token_id": null,
220
+ "forced_eos_token_id": null,
221
+ "remove_invalid_values": false,
222
+ "exponential_decay_length_penalty": null,
223
+ "suppress_tokens": null,
224
+ "begin_suppress_tokens": null,
225
+ "architectures": [
226
+ "Qwen2VLForConditionalGeneration"
227
+ ],
228
+ "finetuning_task": null,
229
+ "id2label": {
230
+ "0": "LABEL_0",
231
+ "1": "LABEL_1"
232
+ },
233
+ "label2id": {
234
+ "LABEL_0": 0,
235
+ "LABEL_1": 1
236
+ },
237
+ "tokenizer_class": null,
238
+ "prefix": null,
239
+ "bos_token_id": 151643,
240
+ "pad_token_id": null,
241
+ "eos_token_id": 151645,
242
+ "sep_token_id": null,
243
+ "decoder_start_token_id": null,
244
+ "task_specific_params": null,
245
+ "problem_type": null,
246
+ "_name_or_path": "ShiniShiho/ShowUI-2B",
247
+ "_attn_implementation_autoset": true,
248
+ "image_token_id": 151655,
249
+ "model_type": "qwen2_vl",
250
+ "tokenizer_model_max_length": 4096,
251
+ "video_token_id": 151656,
252
+ "vision_end_token_id": 151653,
253
+ "vision_start_token_id": 151652,
254
+ "vision_token_id": 151654
255
+ },
256
+ "tf_legacy_loss": false,
257
+ "tie_encoder_decoder": false,
258
+ "tie_word_embeddings": true,
259
+ "tokenizer_class": null,
260
+ "tokenizer_model_max_length": 4096,
261
+ "top_k": 50,
262
+ "top_p": 1.0,
263
+ "torch_dtype": "bfloat16",
264
+ "torchscript": false,
265
+ "transformers_version": "4.48.3",
266
+ "typical_p": 1.0,
267
+ "use_bfloat16": false,
268
+ "use_cache": false,
269
+ "use_sliding_window": false,
270
+ "video_token_id": 151656,
271
+ "vision_config": {
272
+ "return_dict": true,
273
+ "output_hidden_states": false,
274
+ "output_attentions": false,
275
+ "torchscript": false,
276
+ "torch_dtype": null,
277
+ "use_bfloat16": false,
278
+ "tf_legacy_loss": false,
279
+ "pruned_heads": {},
280
+ "tie_word_embeddings": true,
281
+ "chunk_size_feed_forward": 0,
282
+ "is_encoder_decoder": false,
283
+ "is_decoder": false,
284
+ "cross_attention_hidden_size": null,
285
+ "add_cross_attention": false,
286
+ "tie_encoder_decoder": false,
287
+ "max_length": 20,
288
+ "min_length": 0,
289
+ "do_sample": false,
290
+ "early_stopping": false,
291
+ "num_beams": 1,
292
+ "num_beam_groups": 1,
293
+ "diversity_penalty": 0.0,
294
+ "temperature": 1.0,
295
+ "top_k": 50,
296
+ "top_p": 1.0,
297
+ "typical_p": 1.0,
298
+ "repetition_penalty": 1.0,
299
+ "length_penalty": 1.0,
300
+ "no_repeat_ngram_size": 0,
301
+ "encoder_no_repeat_ngram_size": 0,
302
+ "bad_words_ids": null,
303
+ "num_return_sequences": 1,
304
+ "output_scores": false,
305
+ "return_dict_in_generate": false,
306
+ "forced_bos_token_id": null,
307
+ "forced_eos_token_id": null,
308
+ "remove_invalid_values": false,
309
+ "exponential_decay_length_penalty": null,
310
+ "suppress_tokens": null,
311
+ "begin_suppress_tokens": null,
312
+ "architectures": null,
313
+ "finetuning_task": null,
314
+ "id2label": {
315
+ "0": "LABEL_0",
316
+ "1": "LABEL_1"
317
+ },
318
+ "label2id": {
319
+ "LABEL_0": 0,
320
+ "LABEL_1": 1
321
+ },
322
+ "tokenizer_class": null,
323
+ "prefix": null,
324
+ "bos_token_id": null,
325
+ "pad_token_id": null,
326
+ "eos_token_id": null,
327
+ "sep_token_id": null,
328
+ "decoder_start_token_id": null,
329
+ "task_specific_params": null,
330
+ "problem_type": null,
331
+ "_name_or_path": "",
332
+ "_attn_implementation_autoset": false,
333
+ "transformers_version": "4.48.3",
334
+ "in_chans": 3,
335
+ "model_type": "qwen2_vl",
336
+ "spatial_patch_size": 14,
337
+ "depth": 32,
338
+ "embed_dim": 1280,
339
+ "hidden_size": 1536,
340
+ "hidden_act": "quick_gelu",
341
+ "mlp_ratio": 4,
342
+ "num_heads": 16,
343
+ "in_channels": 3,
344
+ "patch_size": 14,
345
+ "spatial_merge_size": 2,
346
+ "temporal_patch_size": 2
347
+ },
348
+ "vision_end_token_id": 151653,
349
+ "vision_start_token_id": 151652,
350
+ "vision_token_id": 151654,
351
+ "vocab_size": 151936
352
  }