Hugging Face's logo

Intel
/
Qwen3-Omni-30B-A3B-Instruct-int4-AutoRound

Safetensors
qwen3_omni_moe
Any-to-Any
4-bit precision
auto-round
Model card Files
xet
 Community
1
lianglv commited on
Commit
12df3a2
·
verified ·
1 Parent(s): 28e2127

Upload folder using huggingface_hub

Browse files
Files changed (20) hide show
  1. .gitattributes +1 -0
  2. chat_template.jinja +122 -0
  3. config.json +512 -0
  4. generation_config.json +8 -0
  5. model-00001-of-00011.safetensors +3 -0
  6. model-00002-of-00011.safetensors +3 -0
  7. model-00003-of-00011.safetensors +3 -0
  8. model-00004-of-00011.safetensors +3 -0
  9. model-00005-of-00011.safetensors +3 -0
  10. model-00006-of-00011.safetensors +3 -0
  11. model-00007-of-00011.safetensors +3 -0
  12. model-00008-of-00011.safetensors +3 -0
  13. model-00009-of-00011.safetensors +3 -0
  14. model-00010-of-00011.safetensors +3 -0
  15. model-00011-of-00011.safetensors +3 -0
  16. model.safetensors.index.json +0 -0
  17. processor_config.json +114 -0
  18. quantization_config.json +92 -0
  19. tokenizer.json +3 -0
  20. tokenizer_config.json +31 -0
.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
chat_template.jinja ADDED
@@ -0,0 +1,122 @@
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
1
+ {%- if tools %}
2
+ {{- '<|im_start|>system\n' }}
3
+ {%- if messages[0].role == 'system' %}
4
+ {%- if messages[0].content is string %}
5
+ {{- messages[0].content }}
6
+ {%- else %}
7
+ {%- for content in messages[0].content %}
8
+ {%- if content.type == 'image' or 'image' in content or 'image_url' in content %}
9
+ {{- "<|vision_start|><|image_pad|><|vision_end|>" }}
10
+ {%- elif content.type == 'audio' or 'audio' in content or 'audio_url' in content %}
11
+ {{- "<|audio_start|><|audio_pad|><|audio_end|>" }}
12
+ {%- elif content.type == 'video' or 'video' in content %}
13
+ {{- "<|vision_start|><|video_pad|><|vision_end|>" }}
14
+ {%- elif content.type == 'text' %}
15
+ {{- content.text }}
16
+ {%- endif %}
17
+ {%- endfor %}
18
+ {%- endif %}
19
+ {%- endif %}
20
+ {{- '\n\n' }}
21
+ {{- "# Tools\n\nYou may call one or more functions to assist with the user query.\n\nYou are provided with function signatures within <tools></tools> XML tags:\n<tools>" }}
22
+ {%- for tool in tools %}
23
+ {{- "\n" }}
24
+ {{- tool | tojson }}
25
+ {%- endfor %}
26
+ {{- "\n</tools>\n\nFor each function call, return a json object with function name and arguments within <tool_call></tool_call> XML tags:\n<tool_call>\n{\"name\": <function-name>, \"arguments\": <args-json-object>}\n</tool_call><|im_end|>\n" }}
27
+ {%- else %}
28
+ {%- if messages[0].role == 'system' %}
29
+ {%- if messages[0].content is string %}
30
+ {{- '<|im_start|>system\n' + messages[0].content + '<|im_end|>\n' }}
31
+ {%- else %}
32
+ {%- for content in messages[0].content %}
33
+ {%- if content.type == 'image' or 'image' in content or 'image_url' in content %}
34
+ {{- '<|im_start|>system\n' +"<|vision_start|><|image_pad|><|vision_end|>"+ '<|im_end|>\n' }}
35
+ {%- elif content.type == 'audio' or 'audio' in content or 'audio_url' in content %}
36
+ {{- '<|im_start|>system\n' +"<|audio_start|><|audio_pad|><|audio_end|>"+ '<|im_end|>\n' }}
37
+ {%- elif content.type == 'video' or 'video' in content %}
38
+ {{- '<|im_start|>system\n' +"<|vision_start|><|video_pad|><|vision_end|>"+ '<|im_end|>\n' }}
39
+ {%- elif content.type == 'text' %}
40
+ {{- '<|im_start|>system\n' +content.text+ '<|im_end|>\n' }}
41
+ {%- endif %}
42
+ {%- endfor %}
43
+ {%- endif %}
44
+ {%- endif %}
45
+ {%- endif %}
46
+ {%- set ns = namespace(multi_step_tool=true, last_query_index=messages|length - 1) %}
47
+ {%- for message in messages[::-1] %}
48
+ {%- set index = (messages|length - 1) - loop.index0 %}
49
+ {%- if ns.multi_step_tool and message.role == "user" and message.content is string and not(message.content.startswith('<tool_response>') and message.content.endswith('</tool_response>')) %}
50
+ {%- set ns.multi_step_tool = false %}
51
+ {%- set ns.last_query_index = index %}
52
+ {%- endif %}
53
+ {%- endfor %}
54
+ {%- for message in messages %}
55
+ {%- if message.content is string %}
56
+ {%- set content = message.content %}
57
+ {%- else %}
58
+ {%- set content = namespace(text="") %}
59
+ {%- for mcontent in message.content %}
60
+ {%- if mcontent.type == 'image' or 'image' in mcontent or 'image_url' in mcontent %}
61
+ {%- set content.text = content.text~"<|vision_start|><|image_pad|><|vision_end|>" %}
62
+ {%- elif mcontent.type == 'audio' or 'audio' in mcontent or 'audio_url' in mcontent %}
63
+ {%- set content.text = content.text~"<|audio_start|><|audio_pad|><|audio_end|>" %}
64
+ {%- elif mcontent.type == 'video' or 'video' in mcontent %}
65
+ {%- set content.text = content.text~"<|vision_start|><|video_pad|><|vision_end|>" %}
66
+ {%- elif mcontent.type == 'text' %}
67
+ {%- set content.text = content.text~mcontent.text %}
68
+ {%- endif %}
69
+ {%- endfor %}
70
+ {%- set content = content.text %}
71
+ {%- endif %}
72
+ {%- if (message.role == "user") or (message.role == "system" and not loop.first) %}
73
+ {{- '<|im_start|>' + message.role + '\n' + content + '<|im_end|>' + '\n' }}
74
+ {%- elif message.role == "assistant" %}
75
+ {%- set reasoning_content = "" %}
76
+ {%- if message.reasoning_content is string %}
77
+ {%- set reasoning_content = message.reasoning_content %}
78
+ {%- else %}
79
+ {%- if '</think>' in content %}
80
+ {%- set reasoning_content = content.split('</think>')[0].rstrip('\n').split('<think>')[-1].lstrip('\n') %}
81
+ {%- set content = content.split('</think>')[-1].lstrip('\n') %}
82
+ {%- endif %}
83
+ {%- endif %}
84
+ {%- if loop.index0 > ns.last_query_index %}
85
+ {%- if loop.last or (not loop.last and reasoning_content) %}
86
+ {{- '<|im_start|>' + message.role + '\n<think>\n' + reasoning_content.strip("\n") + '\n</think>\n\n' + content.lstrip('\n') }}
87
+ {%- else %}
88
+ {{- '<|im_start|>' + message.role + '\n' + content }}
89
+ {%- endif %}
90
+ {%- else %}
91
+ {{- '<|im_start|>' + message.role + '\n' + content }}
92
+ {%- endif %}
93
+ {%- if message.tool_calls %}
94
+ {%- for tool_call in message.tool_calls %}
95
+ {%- if (loop.first and content) or (not loop.first) %}{{- '\n' }}{%- endif %}
96
+ {%- if tool_call.function %}
97
+ {%- set tool_call = tool_call.function %}
98
+ {%- endif %}
99
+ {{- '<tool_call>\n{"name": "' }}
100
+ {{- tool_call.name }}
101
+ {{- '", "arguments": ' }}
102
+ {%- if tool_call.arguments is string %}
103
+ {{- tool_call.arguments }}
104
+ {%- else %}
105
+ {{- tool_call.arguments | tojson }}
106
+ {%- endif %}
107
+ {{- '}\n</tool_call>' }}
108
+ {%- endfor %}
109
+ {%- endif %}
110
+ {{- '<|im_end|>\n' }}
111
+ {%- elif message.role == "tool" %}
112
+ {%- if loop.first or (messages[loop.index0 - 1].role != "tool") %}{{- '<|im_start|>user' }}{%- endif %}
113
+ {{- '\n<tool_response>\n' }}
114
+ {{- content }}
115
+ {{- '\n</tool_response>' }}
116
+ {%- if loop.last or (messages[loop.index0 + 1].role != "tool") %}{{- '<|im_end|>\n' }}{%- endif %}
117
+ {%- endif %}
118
+ {%- endfor %}
119
+ {%- if add_generation_prompt %}
120
+ {{- '<|im_start|>assistant\n' }}
121
+ {%- if enable_thinking is defined and enable_thinking is false %}{{- '<think>\n\n</think>\n\n' }}{%- endif %}
122
+ {%- endif %}
config.json ADDED
@@ -0,0 +1,512 @@
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
1
+ {
2
+ "architectures": [
3
+ "Qwen3OmniMoeForConditionalGeneration"
4
+ ],
5
+ "assistant_token_id": 77091,
6
+ "code2wav_config": {
7
+ "attention_bias": false,
8
+ "attention_dropout": 0.0,
9
+ "codebook_dim": 512,
10
+ "codebook_size": 2048,
11
+ "decoder_dim": 1536,
12
+ "dtype": "bfloat16",
13
+ "hidden_act": "silu",
14
+ "hidden_size": 1024,
15
+ "initializer_range": 0.02,
16
+ "intermediate_size": 3072,
17
+ "layer_scale_initial_scale": 0.01,
18
+ "max_position_embeddings": 8000,
19
+ "model_type": "",
20
+ "num_attention_heads": 16,
21
+ "num_hidden_layers": 8,
22
+ "num_key_value_heads": 16,
23
+ "num_quantizers": 16,
24
+ "num_semantic_quantizers": 1,
25
+ "rms_norm_eps": 1e-05,
26
+ "rope_parameters": {
27
+ "rope_theta": 10000,
28
+ "rope_type": "default"
29
+ },
30
+ "semantic_codebook_size": 4096,
31
+ "sliding_window": 72,
32
+ "upsample_rates": [
33
+ 8,
34
+ 5,
35
+ 4,
36
+ 3
37
+ ],
38
+ "upsampling_ratios": [
39
+ 2,
40
+ 2
41
+ ],
42
+ "vector_quantization_hidden_dimension": 512
43
+ },
44
+ "dtype": "bfloat16",
45
+ "enable_audio_output": true,
46
+ "im_end_token_id": 151645,
47
+ "im_start_token_id": 151644,
48
+ "initializer_range": 0.02,
49
+ "model_type": "qwen3_omni_moe",
50
+ "quantization_config": {
51
+ "autoround_version": "0.12.0",
52
+ "bits": 4,
53
+ "block_name_to_quantize": "thinker.model.layers,talker.model.layers",
54
+ "data_type": "int",
55
+ "extra_config": {
56
+ "talker.model.layers.0.mlp.shared_expert_gate": {
57
+ "bits": 16,
58
+ "data_type": "fp"
59
+ },
60
+ "talker.model.layers.1.mlp.shared_expert_gate": {
61
+ "bits": 16,
62
+ "data_type": "fp"
63
+ },
64
+ "talker.model.layers.10.mlp.shared_expert_gate": {
65
+ "bits": 16,
66
+ "data_type": "fp"
67
+ },
68
+ "talker.model.layers.11.mlp.shared_expert_gate": {
69
+ "bits": 16,
70
+ "data_type": "fp"
71
+ },
72
+ "talker.model.layers.12.mlp.shared_expert_gate": {
73
+ "bits": 16,
74
+ "data_type": "fp"
75
+ },
76
+ "talker.model.layers.13.mlp.shared_expert_gate": {
77
+ "bits": 16,
78
+ "data_type": "fp"
79
+ },
80
+ "talker.model.layers.14.mlp.shared_expert_gate": {
81
+ "bits": 16,
82
+ "data_type": "fp"
83
+ },
84
+ "talker.model.layers.15.mlp.shared_expert_gate": {
85
+ "bits": 16,
86
+ "data_type": "fp"
87
+ },
88
+ "talker.model.layers.16.mlp.shared_expert_gate": {
89
+ "bits": 16,
90
+ "data_type": "fp"
91
+ },
92
+ "talker.model.layers.17.mlp.shared_expert_gate": {
93
+ "bits": 16,
94
+ "data_type": "fp"
95
+ },
96
+ "talker.model.layers.18.mlp.shared_expert_gate": {
97
+ "bits": 16,
98
+ "data_type": "fp"
99
+ },
100
+ "talker.model.layers.19.mlp.shared_expert_gate": {
101
+ "bits": 16,
102
+ "data_type": "fp"
103
+ },
104
+ "talker.model.layers.2.mlp.shared_expert_gate": {
105
+ "bits": 16,
106
+ "data_type": "fp"
107
+ },
108
+ "talker.model.layers.3.mlp.shared_expert_gate": {
109
+ "bits": 16,
110
+ "data_type": "fp"
111
+ },
112
+ "talker.model.layers.4.mlp.shared_expert_gate": {
113
+ "bits": 16,
114
+ "data_type": "fp"
115
+ },
116
+ "talker.model.layers.5.mlp.shared_expert_gate": {
117
+ "bits": 16,
118
+ "data_type": "fp"
119
+ },
120
+ "talker.model.layers.6.mlp.shared_expert_gate": {
121
+ "bits": 16,
122
+ "data_type": "fp"
123
+ },
124
+ "talker.model.layers.7.mlp.shared_expert_gate": {
125
+ "bits": 16,
126
+ "data_type": "fp"
127
+ },
128
+ "talker.model.layers.8.mlp.shared_expert_gate": {
129
+ "bits": 16,
130
+ "data_type": "fp"
131
+ },
132
+ "talker.model.layers.9.mlp.shared_expert_gate": {
133
+ "bits": 16,
134
+ "data_type": "fp"
135
+ }
136
+ },
137
+ "group_size": 128,
138
+ "packing_format": "auto_round:auto_gptq",
139
+ "quant_method": "auto-round",
140
+ "sym": true
141
+ },
142
+ "system_token_id": 8948,
143
+ "talker_config": {
144
+ "accept_hidden_layer": 24,
145
+ "audio_end_token_id": 151670,
146
+ "audio_start_token_id": 151669,
147
+ "audio_token_id": 151675,
148
+ "code_predictor_config": {
149
+ "_name_or_path": "",
150
+ "add_cross_attention": false,
151
+ "architectures": null,
152
+ "attention_bias": false,
153
+ "attention_dropout": 0,
154
+ "bos_token_id": null,
155
+ "chunk_size_feed_forward": 0,
156
+ "cross_attention_hidden_size": null,
157
+ "decoder_start_token_id": null,
158
+ "dtype": null,
159
+ "eos_token_id": null,
160
+ "finetuning_task": null,
161
+ "head_dim": 128,
162
+ "hidden_act": "silu",
163
+ "hidden_size": 1024,
164
+ "id2label": {
165
+ "0": "LABEL_0",
166
+ "1": "LABEL_1"
167
+ },
168
+ "initializer_range": 0.02,
169
+ "intermediate_size": 3072,
170
+ "is_decoder": false,
171
+ "is_encoder_decoder": false,
172
+ "label2id": {
173
+ "LABEL_0": 0,
174
+ "LABEL_1": 1
175
+ },
176
+ "layer_types": [
177
+ "full_attention",
178
+ "full_attention",
179
+ "full_attention",
180
+ "full_attention",
181
+ "full_attention"
182
+ ],
183
+ "max_position_embeddings": 32768,
184
+ "max_window_layers": 28,
185
+ "model_type": "qwen3_omni_moe_talker_code_predictor",
186
+ "num_attention_heads": 16,
187
+ "num_code_groups": 16,
188
+ "num_hidden_layers": 5,
189
+ "num_key_value_heads": 8,
190
+ "output_attentions": false,
191
+ "output_hidden_states": false,
192
+ "pad_token_id": null,
193
+ "prefix": null,
194
+ "problem_type": null,
195
+ "pruned_heads": {},
196
+ "return_dict": true,
197
+ "rms_norm_eps": 1e-06,
198
+ "rope_parameters": {
199
+ "rope_theta": 1000000,
200
+ "rope_type": "default"
201
+ },
202
+ "sep_token_id": null,
203
+ "sliding_window": null,
204
+ "task_specific_params": null,
205
+ "tf_legacy_loss": false,
206
+ "tie_encoder_decoder": false,
207
+ "tie_word_embeddings": false,
208
+ "tokenizer_class": null,
209
+ "torchscript": false,
210
+ "use_bfloat16": false,
211
+ "use_cache": true,
212
+ "use_sliding_window": false,
213
+ "vocab_size": 2048
214
+ },
215
+ "codec_bos_id": 2149,
216
+ "codec_eos_token_id": 2150,
217
+ "codec_nothink_id": 2155,
218
+ "codec_pad_id": 2148,
219
+ "codec_think_bos_id": 2156,
220
+ "codec_think_eos_id": 2157,
221
+ "dtype": "bfloat16",
222
+ "image_token_id": 151655,
223
+ "initializer_range": 0.02,
224
+ "model_type": "",
225
+ "num_code_groups": 16,
226
+ "output_router_logits": false,
227
+ "position_id_per_seconds": 13,
228
+ "seconds_per_chunk": 2,
229
+ "spatial_merge_size": 2,
230
+ "speaker_id": {
231
+ "aiden": 2303,
232
+ "chelsie": 2301,
233
+ "ethan": 2302
234
+ },
235
+ "text_config": {
236
+ "_name_or_path": "",
237
+ "architectures": null,
238
+ "attention_bias": false,
239
+ "attention_dropout": 0,
240
+ "bos_token_id": null,
241
+ "chunk_size_feed_forward": 0,
242
+ "decoder_sparse_step": 1,
243
+ "dtype": null,
244
+ "eos_token_id": null,
245
+ "head_dim": 128,
246
+ "hidden_act": "silu",
247
+ "hidden_size": 1024,
248
+ "id2label": {
249
+ "0": "LABEL_0",
250
+ "1": "LABEL_1"
251
+ },
252
+ "initializer_range": 0.02,
253
+ "intermediate_size": 2048,
254
+ "is_encoder_decoder": false,
255
+ "label2id": {
256
+ "LABEL_0": 0,
257
+ "LABEL_1": 1
258
+ },
259
+ "max_position_embeddings": 65536,
260
+ "mlp_only_layers": [],
261
+ "model_type": "qwen3_omni_moe_talker_text",
262
+ "moe_intermediate_size": 384,
263
+ "norm_topk_prob": true,
264
+ "num_attention_heads": 16,
265
+ "num_experts_per_tok": 6,
266
+ "num_hidden_layers": 20,
267
+ "num_key_value_heads": 2,
268
+ "num_local_experts": 128,
269
+ "output_attentions": false,
270
+ "output_hidden_states": false,
271
+ "output_router_logits": false,
272
+ "pad_token_id": null,
273
+ "problem_type": null,
274
+ "return_dict": true,
275
+ "rms_norm_eps": 1e-06,
276
+ "rope_parameters": {
277
+ "interleaved": true,
278
+ "mrope_section": [
279
+ 24,
280
+ 20,
281
+ 20
282
+ ],
283
+ "rope_theta": 1000000,
284
+ "rope_type": "default",
285
+ "type": "default"
286
+ },
287
+ "router_aux_loss_coef": 0.001,
288
+ "shared_expert_intermediate_size": 768,
289
+ "sliding_window": null,
290
+ "tie_word_embeddings": false,
291
+ "use_cache": true,
292
+ "use_sliding_window": false,
293
+ "vocab_size": 3072
294
+ },
295
+ "thinker_hidden_size": 2048,
296
+ "video_token_id": 151656,
297
+ "vision_start_token_id": 151652
298
+ },
299
+ "thinker_config": {
300
+ "audio_config": {
301
+ "_name_or_path": "",
302
+ "activation_dropout": 0,
303
+ "activation_function": "gelu",
304
+ "add_cross_attention": false,
305
+ "architectures": null,
306
+ "attention_dropout": 0,
307
+ "bos_token_id": null,
308
+ "chunk_size_feed_forward": 0,
309
+ "conv_chunksize": 500,
310
+ "cross_attention_hidden_size": null,
311
+ "d_model": 1280,
312
+ "decoder_start_token_id": null,
313
+ "downsample_hidden_size": 480,
314
+ "dropout": 0,
315
+ "dtype": null,
316
+ "encoder_attention_heads": 20,
317
+ "encoder_ffn_dim": 5120,
318
+ "encoder_layers": 32,
319
+ "eos_token_id": null,
320
+ "finetuning_task": null,
321
+ "id2label": {
322
+ "0": "LABEL_0",
323
+ "1": "LABEL_1"
324
+ },
325
+ "initializer_range": 0.02,
326
+ "is_decoder": false,
327
+ "is_encoder_decoder": false,
328
+ "label2id": {
329
+ "LABEL_0": 0,
330
+ "LABEL_1": 1
331
+ },
332
+ "max_source_positions": 1500,
333
+ "model_type": "qwen3_omni_moe_audio_encoder",
334
+ "n_window": 50,
335
+ "n_window_infer": 800,
336
+ "num_hidden_layers": 32,
337
+ "num_mel_bins": 128,
338
+ "output_attentions": false,
339
+ "output_dim": 2048,
340
+ "output_hidden_states": false,
341
+ "pad_token_id": null,
342
+ "prefix": null,
343
+ "problem_type": null,
344
+ "pruned_heads": {},
345
+ "return_dict": true,
346
+ "scale_embedding": false,
347
+ "sep_token_id": null,
348
+ "task_specific_params": null,
349
+ "tf_legacy_loss": false,
350
+ "tie_encoder_decoder": false,
351
+ "tie_word_embeddings": true,
352
+ "tokenizer_class": null,
353
+ "torchscript": false,
354
+ "use_bfloat16": false
355
+ },
356
+ "audio_end_token_id": 151670,
357
+ "audio_start_token_id": 151669,
358
+ "audio_token_id": 151675,
359
+ "dtype": "bfloat16",
360
+ "image_token_id": 151655,
361
+ "initializer_range": 0.02,
362
+ "model_type": "qwen3_omni_moe_thinker",
363
+ "position_id_per_seconds": 13,
364
+ "seconds_per_chunk": 2,
365
+ "text_config": {
366
+ "_name_or_path": "",
367
+ "add_cross_attention": false,
368
+ "architectures": null,
369
+ "attention_bias": false,
370
+ "attention_dropout": 0.0,
371
+ "bos_token_id": null,
372
+ "chunk_size_feed_forward": 0,
373
+ "cross_attention_hidden_size": null,
374
+ "decoder_sparse_step": 1,
375
+ "decoder_start_token_id": null,
376
+ "dtype": null,
377
+ "eos_token_id": null,
378
+ "finetuning_task": null,
379
+ "head_dim": 128,
380
+ "hidden_act": "silu",
381
+ "hidden_size": 2048,
382
+ "id2label": {
383
+ "0": "LABEL_0",
384
+ "1": "LABEL_1"
385
+ },
386
+ "initializer_range": 0.02,
387
+ "intermediate_size": 768,
388
+ "is_decoder": false,
389
+ "is_encoder_decoder": false,
390
+ "label2id": {
391
+ "LABEL_0": 0,
392
+ "LABEL_1": 1
393
+ },
394
+ "max_position_embeddings": 65536,
395
+ "mlp_only_layers": [],
396
+ "model_type": "qwen3_omni_moe_text",
397
+ "moe_intermediate_size": 768,
398
+ "norm_topk_prob": true,
399
+ "num_attention_heads": 32,
400
+ "num_experts": 128,
401
+ "num_experts_per_tok": 8,
402
+ "num_hidden_layers": 48,
403
+ "num_key_value_heads": 4,
404
+ "output_attentions": false,
405
+ "output_hidden_states": false,
406
+ "output_router_logits": false,
407
+ "pad_token_id": null,
408
+ "prefix": null,
409
+ "problem_type": null,
410
+ "pruned_heads": {},
411
+ "return_dict": true,
412
+ "rms_norm_eps": 1e-06,
413
+ "rope_parameters": {
414
+ "interleaved": true,
415
+ "mrope_interleaved": true,
416
+ "mrope_section": [
417
+ 24,
418
+ 20,
419
+ 20
420
+ ],
421
+ "rope_theta": 1000000,
422
+ "rope_type": "default",
423
+ "type": "default"
424
+ },
425
+ "router_aux_loss_coef": 0.001,
426
+ "sep_token_id": null,
427
+ "shared_expert_intermediate_size": 0,
428
+ "sliding_window": null,
429
+ "task_specific_params": null,
430
+ "tf_legacy_loss": false,
431
+ "tie_encoder_decoder": false,
432
+ "tie_word_embeddings": false,
433
+ "tokenizer_class": null,
434
+ "torchscript": false,
435
+ "use_bfloat16": false,
436
+ "use_cache": true,
437
+ "use_qk_norm": true,
438
+ "use_sliding_window": false,
439
+ "vocab_size": 152064
440
+ },
441
+ "tie_word_embeddings": false,
442
+ "user_token_id": 872,
443
+ "video_token_id": 151656,
444
+ "vision_config": {
445
+ "_name_or_path": "",
446
+ "add_cross_attention": false,
447
+ "apply_vit_abs_pos_embed": true,
448
+ "architectures": null,
449
+ "bos_token_id": null,
450
+ "chunk_size_feed_forward": 0,
451
+ "cross_attention_hidden_size": null,
452
+ "decoder_start_token_id": null,
453
+ "deepstack_visual_indexes": [
454
+ 8,
455
+ 16,
456
+ 24
457
+ ],
458
+ "depth": 27,
459
+ "dtype": null,
460
+ "eos_token_id": null,
461
+ "finetuning_task": null,
462
+ "hidden_act": "gelu_pytorch_tanh",
463
+ "hidden_size": 1152,
464
+ "id2label": {
465
+ "0": "LABEL_0",
466
+ "1": "LABEL_1"
467
+ },
468
+ "image_size": 768,
469
+ "in_channels": 3,
470
+ "in_chans": 3,
471
+ "initializer_range": 0.02,
472
+ "intermediate_size": 4304,
473
+ "is_decoder": false,
474
+ "is_encoder_decoder": false,
475
+ "label2id": {
476
+ "LABEL_0": 0,
477
+ "LABEL_1": 1
478
+ },
479
+ "model_type": "qwen3_omni_moe_vision_encoder",
480
+ "num_heads": 16,
481
+ "num_position_embeddings": 2304,
482
+ "out_hidden_size": 2048,
483
+ "output_attentions": false,
484
+ "output_hidden_states": false,
485
+ "pad_token_id": null,
486
+ "patch_size": 16,
487
+ "prefix": null,
488
+ "problem_type": null,
489
+ "pruned_heads": {},
490
+ "return_dict": true,
491
+ "sep_token_id": null,
492
+ "spatial_merge_size": 2,
493
+ "spatial_patch_size": 16,
494
+ "task_specific_params": null,
495
+ "temporal_patch_size": 2,
496
+ "tf_legacy_loss": false,
497
+ "tie_encoder_decoder": false,
498
+ "tie_word_embeddings": true,
499
+ "tokenizer_class": null,
500
+ "tokens_per_second": 2,
501
+ "torchscript": false,
502
+ "use_bfloat16": false
503
+ },
504
+ "vision_end_token_id": 151653,
505
+ "vision_start_token_id": 151652
506
+ },
507
+ "transformers_version": "5.3.0",
508
+ "tts_bos_token_id": 151672,
509
+ "tts_eos_token_id": 151673,
510
+ "tts_pad_token_id": 151671,
511
+ "user_token_id": 872
512
+ }
generation_config.json ADDED
@@ -0,0 +1,8 @@
 
 
 
 
 
 
 
 
 
1
+ {
2
+ "talker_max_new_tokens": 4096,
3
+ "talker_repetition_penalty": 1.05,
4
+ "talker_temperature": 0.9,
5
+ "talker_top_k": 50,
6
+ "talker_top_p": 1.0,
7
+ "transformers_version": "5.3.0"
8
+ }
model-00001-of-00011.safetensors ADDED
@@ -0,0 +1,3 @@
 
 
 
 
1
+ version https://git-lfs.github.com/spec/v1
2
+ oid sha256:63e49d4bca7a13935f202d2b7b05fb5316c4e91e149e580098525ff6390955c4
3
+ size 2148102080
model-00002-of-00011.safetensors ADDED
@@ -0,0 +1,3 @@
 
 
 
 
1
+ version https://git-lfs.github.com/spec/v1
2
+ oid sha256:160e5cc5001d39401ce3cd1bae6d641723d9d61418bd4d4e593e6648002f2e67
3
+ size 2148114544
model-00003-of-00011.safetensors ADDED
@@ -0,0 +1,3 @@
 
 
 
 
1
+ version https://git-lfs.github.com/spec/v1
2
+ oid sha256:fce8217ee4c24e825758bb9f5993cd7cceb287541585cca9e03e7e327b6f4ac1
3
+ size 2148405552
model-00004-of-00011.safetensors ADDED
@@ -0,0 +1,3 @@
 
 
 
 
1
+ version https://git-lfs.github.com/spec/v1
2
+ oid sha256:eea4e4dbe3df67bdf5de662d1278c0a51582f63e13c695357c23a3639c56c647
3
+ size 2148118400
model-00005-of-00011.safetensors ADDED
@@ -0,0 +1,3 @@
 
 
 
 
1
+ version https://git-lfs.github.com/spec/v1
2
+ oid sha256:aa22554cc7d721138d869d45263409ed34d2befe27ed5a15cd1bbedb70448fe4
3
+ size 2148118472
model-00006-of-00011.safetensors ADDED
@@ -0,0 +1,3 @@
 
 
 
 
1
+ version https://git-lfs.github.com/spec/v1
2
+ oid sha256:2c77659b03f91e9476531e35a07acdf4657ad3ed6f6ba6337c1b725db807ff1f
3
+ size 2148405448
model-00007-of-00011.safetensors ADDED
@@ -0,0 +1,3 @@
 
 
 
 
1
+ version https://git-lfs.github.com/spec/v1
2
+ oid sha256:32c0b65b800654536b569555ad1b23ade5dd212f6beb445d9b2d13df643bdac6
3
+ size 2148118472
model-00008-of-00011.safetensors ADDED
@@ -0,0 +1,3 @@
 
 
 
 
1
+ version https://git-lfs.github.com/spec/v1
2
+ oid sha256:af69f2ffe25a06da5c567b8fc4ab6b9ffb3ca2666decfc0d08cce51a7296fe8a
3
+ size 2150651144
model-00009-of-00011.safetensors ADDED
@@ -0,0 +1,3 @@
 
 
 
 
1
+ version https://git-lfs.github.com/spec/v1
2
+ oid sha256:a5d55bb4dee37aaee597826c4c994fd78413d817346c4cc91c35deb1831290b5
3
+ size 2147104056
model-00010-of-00011.safetensors ADDED
@@ -0,0 +1,3 @@
 
 
 
 
1
+ version https://git-lfs.github.com/spec/v1
2
+ oid sha256:ad4569d9cbef997fe7eddf7319334fbb5bc9ad6bfe7147e1c7e8142d93e1d783
3
+ size 2146742568
model-00011-of-00011.safetensors ADDED
@@ -0,0 +1,3 @@
 
 
 
 
1
+ version https://git-lfs.github.com/spec/v1
2
+ oid sha256:05fe65d1facd7b85c56cf7d4958221fb5277ba31c5f97d9944005bcfb515e0a8
3
+ size 134151562
model.safetensors.index.json ADDED
The diff for this file is too large to render. See raw diff
 
processor_config.json ADDED
@@ -0,0 +1,114 @@
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
1
+ {
2
+ "feature_extractor": {
3
+ "chunk_length": 30,
4
+ "dither": 0.0,
5
+ "feature_extractor_type": "WhisperFeatureExtractor",
6
+ "feature_size": 128,
7
+ "hop_length": 160,
8
+ "image_mean": [
9
+ 0.5,
10
+ 0.5,
11
+ 0.5
12
+ ],
13
+ "image_processor_type": "Qwen2VLImageProcessor",
14
+ "image_std": [
15
+ 0.5,
16
+ 0.5,
17
+ 0.5
18
+ ],
19
+ "max_pixels": 12845056,
20
+ "merge_size": 2,
21
+ "min_pixels": 3136,
22
+ "n_fft": 400,
23
+ "n_samples": 480000,
24
+ "nb_max_frames": 3000,
25
+ "padding_side": "right",
26
+ "padding_value": 0.0,
27
+ "patch_size": 16,
28
+ "return_attention_mask": true,
29
+ "sampling_rate": 16000,
30
+ "temporal_patch_size": 2
31
+ },
32
+ "image_processor": {
33
+ "data_format": "channels_first",
34
+ "dither": 0.0,
35
+ "do_convert_rgb": true,
36
+ "do_normalize": true,
37
+ "do_rescale": true,
38
+ "do_resize": true,
39
+ "feature_size": 128,
40
+ "hop_length": 160,
41
+ "image_mean": [
42
+ 0.5,
43
+ 0.5,
44
+ 0.5
45
+ ],
46
+ "image_processor_type": "Qwen2VLImageProcessorFast",
47
+ "image_std": [
48
+ 0.5,
49
+ 0.5,
50
+ 0.5
51
+ ],
52
+ "merge_size": 2,
53
+ "n_fft": 400,
54
+ "n_samples": 4800000,
55
+ "nb_max_frames": 30000,
56
+ "padding_side": "right",
57
+ "padding_value": 0.0,
58
+ "patch_size": 16,
59
+ "resample": 3,
60
+ "rescale_factor": 0.00392156862745098,
61
+ "return_attention_mask": true,
62
+ "sampling_rate": 16000,
63
+ "size": {
64
+ "longest_edge": 12845056,
65
+ "shortest_edge": 3136
66
+ },
67
+ "temporal_patch_size": 2
68
+ },
69
+ "processor_class": "Qwen3OmniMoeProcessor",
70
+ "video_processor": {
71
+ "data_format": "channels_first",
72
+ "default_to_square": true,
73
+ "dither": 0.0,
74
+ "do_convert_rgb": true,
75
+ "do_normalize": true,
76
+ "do_rescale": true,
77
+ "do_resize": true,
78
+ "do_sample_frames": false,
79
+ "feature_extractor_type": "WhisperFeatureExtractor",
80
+ "feature_size": 128,
81
+ "hop_length": 160,
82
+ "image_mean": [
83
+ 0.5,
84
+ 0.5,
85
+ 0.5
86
+ ],
87
+ "image_processor_type": "Qwen2VLImageProcessor",
88
+ "image_std": [
89
+ 0.5,
90
+ 0.5,
91
+ 0.5
92
+ ],
93
+ "max_frames": 768,
94
+ "merge_size": 2,
95
+ "min_frames": 4,
96
+ "n_fft": 400,
97
+ "n_samples": 4800000,
98
+ "nb_max_frames": 30000,
99
+ "padding_side": "right",
100
+ "padding_value": 0.0,
101
+ "patch_size": 16,
102
+ "resample": 3,
103
+ "rescale_factor": 0.00392156862745098,
104
+ "return_attention_mask": true,
105
+ "return_metadata": false,
106
+ "sampling_rate": 16000,
107
+ "size": {
108
+ "longest_edge": 12845056,
109
+ "shortest_edge": 3136
110
+ },
111
+ "temporal_patch_size": 2,
112
+ "video_processor_type": "Qwen2VLVideoProcessor"
113
+ }
114
+ }
quantization_config.json ADDED
@@ -0,0 +1,92 @@
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
1
+ {
2
+ "bits": 4,
3
+ "data_type": "int",
4
+ "group_size": 128,
5
+ "sym": true,
6
+ "autoround_version": "0.12.0",
7
+ "block_name_to_quantize": "thinker.model.layers,talker.model.layers",
8
+ "quant_method": "auto-round",
9
+ "packing_format": "auto_round:auto_gptq",
10
+ "extra_config": {
11
+ "talker.model.layers.0.mlp.shared_expert_gate": {
12
+ "bits": 16,
13
+ "data_type": "fp"
14
+ },
15
+ "talker.model.layers.1.mlp.shared_expert_gate": {
16
+ "bits": 16,
17
+ "data_type": "fp"
18
+ },
19
+ "talker.model.layers.2.mlp.shared_expert_gate": {
20
+ "bits": 16,
21
+ "data_type": "fp"
22
+ },
23
+ "talker.model.layers.3.mlp.shared_expert_gate": {
24
+ "bits": 16,
25
+ "data_type": "fp"
26
+ },
27
+ "talker.model.layers.4.mlp.shared_expert_gate": {
28
+ "bits": 16,
29
+ "data_type": "fp"
30
+ },
31
+ "talker.model.layers.5.mlp.shared_expert_gate": {
32
+ "bits": 16,
33
+ "data_type": "fp"
34
+ },
35
+ "talker.model.layers.6.mlp.shared_expert_gate": {
36
+ "bits": 16,
37
+ "data_type": "fp"
38
+ },
39
+ "talker.model.layers.7.mlp.shared_expert_gate": {
40
+ "bits": 16,
41
+ "data_type": "fp"
42
+ },
43
+ "talker.model.layers.8.mlp.shared_expert_gate": {
44
+ "bits": 16,
45
+ "data_type": "fp"
46
+ },
47
+ "talker.model.layers.9.mlp.shared_expert_gate": {
48
+ "bits": 16,
49
+ "data_type": "fp"
50
+ },
51
+ "talker.model.layers.10.mlp.shared_expert_gate": {
52
+ "bits": 16,
53
+ "data_type": "fp"
54
+ },
55
+ "talker.model.layers.11.mlp.shared_expert_gate": {
56
+ "bits": 16,
57
+ "data_type": "fp"
58
+ },
59
+ "talker.model.layers.12.mlp.shared_expert_gate": {
60
+ "bits": 16,
61
+ "data_type": "fp"
62
+ },
63
+ "talker.model.layers.13.mlp.shared_expert_gate": {
64
+ "bits": 16,
65
+ "data_type": "fp"
66
+ },
67
+ "talker.model.layers.14.mlp.shared_expert_gate": {
68
+ "bits": 16,
69
+ "data_type": "fp"
70
+ },
71
+ "talker.model.layers.15.mlp.shared_expert_gate": {
72
+ "bits": 16,
73
+ "data_type": "fp"
74
+ },
75
+ "talker.model.layers.16.mlp.shared_expert_gate": {
76
+ "bits": 16,
77
+ "data_type": "fp"
78
+ },
79
+ "talker.model.layers.17.mlp.shared_expert_gate": {
80
+ "bits": 16,
81
+ "data_type": "fp"
82
+ },
83
+ "talker.model.layers.18.mlp.shared_expert_gate": {
84
+ "bits": 16,
85
+ "data_type": "fp"
86
+ },
87
+ "talker.model.layers.19.mlp.shared_expert_gate": {
88
+ "bits": 16,
89
+ "data_type": "fp"
90
+ }
91
+ }
92
+ }
tokenizer.json ADDED
@@ -0,0 +1,3 @@
 
 
 
 
1
+ version https://git-lfs.github.com/spec/v1
2
+ oid sha256:a7d41145c408f3062e824f965be1c29854cd809e111cdf8170aa8b0bcd5d1fab
3
+ size 11424262
tokenizer_config.json ADDED
@@ -0,0 +1,31 @@
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
1
+ {
2
+ "add_prefix_space": false,
3
+ "audio_bos_token": "<|audio_start|>",
4
+ "audio_eos_token": "<|audio_end|>",
5
+ "audio_token": "<|audio_pad|>",
6
+ "backend": "tokenizers",
7
+ "bos_token": null,
8
+ "clean_up_tokenization_spaces": false,
9
+ "eos_token": "<|im_end|>",
10
+ "errors": "replace",
11
+ "image_token": "<|image_pad|>",
12
+ "is_local": true,
13
+ "model_max_length": 131072,
14
+ "model_specific_special_tokens": {
15
+ "audio_bos_token": "<|audio_start|>",
16
+ "audio_eos_token": "<|audio_end|>",
17
+ "audio_token": "<|audio_pad|>",
18
+ "image_token": "<|image_pad|>",
19
+ "video_token": "<|video_pad|>",
20
+ "vision_bos_token": "<|vision_start|>",
21
+ "vision_eos_token": "<|vision_end|>"
22
+ },
23
+ "pad_token": "<|endoftext|>",
24
+ "processor_class": "Qwen3OmniMoeProcessor",
25
+ "split_special_tokens": false,
26
+ "tokenizer_class": "Qwen2Tokenizer",
27
+ "unk_token": null,
28
+ "video_token": "<|video_pad|>",
29
+ "vision_bos_token": "<|vision_start|>",
30
+ "vision_eos_token": "<|vision_end|>"
31
+ }