Hugging Face's logo

xianghe
/
2026_ARC

Safetensors
qwen3_omni_moe
Model card Files
xet
 Community
roycehe commited on
Commit
397f701
·
1 Parent(s): 0d00c68

Add model files and configurations

Browse files
Files changed (27) hide show
  1. .gitattributes +1 -0
  2. added_tokens.json +35 -0
  3. args.json +692 -0
  4. chat_template.jinja +122 -0
  5. config.json +557 -0
  6. infer_result/20260202-221152.jsonl +0 -0
  7. merges.txt +0 -0
  8. model-00001-of-00013.safetensors +3 -0
  9. model-00002-of-00013.safetensors +3 -0
  10. model-00003-of-00013.safetensors +3 -0
  11. model-00004-of-00013.safetensors +3 -0
  12. model-00005-of-00013.safetensors +3 -0
  13. model-00006-of-00013.safetensors +3 -0
  14. model-00007-of-00013.safetensors +3 -0
  15. model-00008-of-00013.safetensors +3 -0
  16. model-00009-of-00013.safetensors +3 -0
  17. model-00010-of-00013.safetensors +3 -0
  18. model-00011-of-00013.safetensors +3 -0
  19. model-00012-of-00013.safetensors +3 -0
  20. model-00013-of-00013.safetensors +3 -0
  21. model.safetensors.index.json +0 -0
  22. preprocessor_config.json +30 -0
  23. special_tokens_map.json +44 -0
  24. tokenizer.json +3 -0
  25. tokenizer_config.json +317 -0
  26. video_preprocessor_config.json +54 -0
  27. vocab.json +0 -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
added_tokens.json ADDED
@@ -0,0 +1,35 @@
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
1
+ {
2
+ "</think>": 151668,
3
+ "</tool_call>": 151658,
4
+ "</tool_response>": 151666,
5
+ "<think>": 151667,
6
+ "<tool_call>": 151657,
7
+ "<tool_response>": 151665,
8
+ "<tts_pad>": 151671,
9
+ "<tts_text_bos>": 151672,
10
+ "<tts_text_bos_single>": 151674,
11
+ "<tts_text_eod>": 151673,
12
+ "<|audio_end|>": 151670,
13
+ "<|audio_pad|>": 151675,
14
+ "<|audio_start|>": 151669,
15
+ "<|box_end|>": 151649,
16
+ "<|box_start|>": 151648,
17
+ "<|endoftext|>": 151643,
18
+ "<|file_sep|>": 151664,
19
+ "<|fim_middle|>": 151660,
20
+ "<|fim_pad|>": 151662,
21
+ "<|fim_prefix|>": 151659,
22
+ "<|fim_suffix|>": 151661,
23
+ "<|im_end|>": 151645,
24
+ "<|im_start|>": 151644,
25
+ "<|image_pad|>": 151655,
26
+ "<|object_ref_end|>": 151647,
27
+ "<|object_ref_start|>": 151646,
28
+ "<|quad_end|>": 151651,
29
+ "<|quad_start|>": 151650,
30
+ "<|repo_name|>": 151663,
31
+ "<|video_pad|>": 151656,
32
+ "<|vision_end|>": 151653,
33
+ "<|vision_pad|>": 151654,
34
+ "<|vision_start|>": 151652
35
+ }
args.json ADDED
@@ -0,0 +1,692 @@
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
1
+ {
2
+ "use_ray": false,
3
+ "ray_exp_name": null,
4
+ "device_groups": null,
5
+ "model": "/hifs_gz11/share_303840540/user/chenxingli/intern/xianghe/qwen3omni_instruct_GRPO/v30-20260121-012721/checkpoint-500/",
6
+ "model_type": "qwen3_omni",
7
+ "model_revision": null,
8
+ "task_type": "causal_lm",
9
+ "torch_dtype": "bfloat16",
10
+ "attn_impl": null,
11
+ "new_special_tokens": [],
12
+ "num_labels": null,
13
+ "problem_type": null,
14
+ "rope_scaling": {
15
+ "interleaved": true,
16
+ "mrope_interleaved": true,
17
+ "mrope_section": [
18
+ 24,
19
+ 20,
20
+ 20
21
+ ],
22
+ "rope_type": "default",
23
+ "type": "default"
24
+ },
25
+ "device_map": null,
26
+ "max_memory": {},
27
+ "max_model_len": null,
28
+ "local_repo_path": null,
29
+ "init_strategy": null,
30
+ "template": "qwen3_omni",
31
+ "system": "You are an audio understanding model that answers multiple choice questions based on audio content. You MUST always start your reasoning with <reasoning> and end with </reasoning> tags, followed by the exact option text, i.e., <reasoning>your detailed reasoning process</reasoning>\nselected option text",
32
+ "max_length": 4096,
33
+ "truncation_strategy": "left",
34
+ "max_pixels": null,
35
+ "agent_template": null,
36
+ "norm_bbox": null,
37
+ "use_chat_template": true,
38
+ "padding_free": true,
39
+ "padding_side": "right",
40
+ "loss_scale": "last_round",
41
+ "sequence_parallel_size": 1,
42
+ "response_prefix": null,
43
+ "template_backend": "swift",
44
+ "dataset": [
45
+ "/apdcephfs_fsgm/share_303840540/user/chenxingli/intern/xianghe/codes/SpeechLM-RL/20260123_merged_shuffled_filtered.jsonl"
46
+ ],
47
+ "val_dataset": [],
48
+ "cached_dataset": [],
49
+ "cached_val_dataset": [],
50
+ "split_dataset_ratio": 0.0,
51
+ "data_seed": 42,
52
+ "dataset_num_proc": 8,
53
+ "load_from_cache_file": false,
54
+ "dataset_shuffle": true,
55
+ "val_dataset_shuffle": false,
56
+ "streaming": false,
57
+ "interleave_prob": null,
58
+ "stopping_strategy": "first_exhausted",
59
+ "shuffle_buffer_size": 1000,
60
+ "download_mode": "reuse_dataset_if_exists",
61
+ "columns": {},
62
+ "strict": false,
63
+ "remove_unused_columns": false,
64
+ "model_name": null,
65
+ "model_author": null,
66
+ "custom_dataset_info": [],
67
+ "quant_method": null,
68
+ "quant_bits": null,
69
+ "hqq_axis": null,
70
+ "bnb_4bit_compute_dtype": "bfloat16",
71
+ "bnb_4bit_quant_type": "nf4",
72
+ "bnb_4bit_use_double_quant": true,
73
+ "bnb_4bit_quant_storage": null,
74
+ "max_new_tokens": null,
75
+ "temperature": 1.0,
76
+ "top_k": 50,
77
+ "top_p": 0.99,
78
+ "repetition_penalty": 1.0,
79
+ "num_beams": 1,
80
+ "stream": false,
81
+ "stop_words": [],
82
+ "logprobs": false,
83
+ "top_logprobs": null,
84
+ "ckpt_dir": "/hifs_gz11/share_303840540/user/chenxingli/intern/xianghe/qwen3omni_instruct_GRPO/v30-20260121-012721/checkpoint-500/",
85
+ "lora_modules": [],
86
+ "tuner_backend": "peft",
87
+ "train_type": "full",
88
+ "adapters": [],
89
+ "external_plugins": [
90
+ "/apdcephfs_fsgm/share_303840540/user/chenxingli/intern/xianghe/codes/Audio-Thinker/GRPO/plugin/plugin.py"
91
+ ],
92
+ "seed": 42,
93
+ "model_kwargs": {},
94
+ "load_args": false,
95
+ "load_data_args": false,
96
+ "packing": false,
97
+ "packing_length": null,
98
+ "packing_num_proc": 1,
99
+ "lazy_tokenize": true,
100
+ "custom_register_path": [],
101
+ "use_hf": false,
102
+ "hub_token": null,
103
+ "ddp_timeout": 18000000,
104
+ "ddp_backend": null,
105
+ "ignore_args_error": false,
106
+ "use_swift_lora": false,
107
+ "freeze_llm": false,
108
+ "freeze_vit": false,
109
+ "freeze_aligner": false,
110
+ "freeze_parameters": [],
111
+ "freeze_parameters_regex": null,
112
+ "freeze_parameters_ratio": 0.0,
113
+ "trainable_parameters": [],
114
+ "trainable_parameters_regex": null,
115
+ "adapter_load": null,
116
+ "target_modules": [
117
+ "all-linear"
118
+ ],
119
+ "target_regex": null,
120
+ "modules_to_save": [],
121
+ "lora_rank": 8,
122
+ "lora_alpha": 32,
123
+ "lora_dropout": 0.05,
124
+ "lora_bias": "none",
125
+ "lora_dtype": null,
126
+ "use_rslora": false,
127
+ "rlhf_type": "grpo",
128
+ "ref_load": null,
129
+ "ref_adapter_load": null,
130
+ "beta": 0.001,
131
+ "rpo_alpha": null,
132
+ "reference_free": false,
133
+ "label_smoothing": 0.0,
134
+ "f_divergence_type": "reverse_kl",
135
+ "loss_type": "grpo",
136
+ "desirable_weight": 1.0,
137
+ "undesirable_weight": 1.0,
138
+ "calculate_KL": null,
139
+ "center_rewards_coefficient": null,
140
+ "generation_batch_size": 896,
141
+ "steps_per_generation": 4,
142
+ "num_generations": 8,
143
+ "max_completion_length": 1024,
144
+ "importance_sampling_level": "sequence",
145
+ "tau_pos": 1.0,
146
+ "tau_neg": 1.05,
147
+ "use_gdpo": true,
148
+ "epsilon": 0.2,
149
+ "epsilon_high": null,
150
+ "delta": null,
151
+ "use_vllm": true,
152
+ "vllm_mode": "server",
153
+ "vllm_enable_prefix_caching": true,
154
+ "vllm_gpu_memory_utilization": 0.9,
155
+ "vllm_tensor_parallel_size": 1,
156
+ "vllm_max_model_len": null,
157
+ "vllm_enforce_eager": false,
158
+ "vllm_limit_mm_per_prompt": null,
159
+ "vllm_disable_cascade_attn": false,
160
+ "vllm_max_num_seqs": null,
161
+ "vllm_mm_processor_cache_gb": null,
162
+ "vllm_engine_kwargs": {},
163
+ "sleep_level": 0,
164
+ "offload_optimizer": true,
165
+ "offload_model": true,
166
+ "offload_bridge": false,
167
+ "vllm_server_base_url": null,
168
+ "vllm_server_host": [
169
+ "29.164.185.85"
170
+ ],
171
+ "vllm_server_port": [
172
+ 8989
173
+ ],
174
+ "vllm_server_timeout": 240.0,
175
+ "vllm_server_group_port": null,
176
+ "reward_funcs": [
177
+ "external_acc_reward",
178
+ "external_mixedsimilarity_reward"
179
+ ],
180
+ "reward_weights": [
181
+ 1.0,
182
+ 0.5
183
+ ],
184
+ "cosine_min_len_value_wrong": -0.5,
185
+ "cosine_max_len_value_wrong": 0.0,
186
+ "cosine_min_len_value_correct": 1.0,
187
+ "cosine_max_len_value_correct": 0.5,
188
+ "cosine_max_len": null,
189
+ "repetition_n_grams": 3,
190
+ "repetition_max_penalty": -1.0,
191
+ "soft_max_length": null,
192
+ "soft_cache_length": null,
193
+ "dynamic_sample": false,
194
+ "max_resample_times": 3,
195
+ "overlong_filter": true,
196
+ "scale_rewards": "group",
197
+ "advantage_estimator": "grpo",
198
+ "kl_in_reward": false,
199
+ "wandb_log_unique_prompts": null,
200
+ "log_completions": true,
201
+ "rollout_importance_sampling_mode": null,
202
+ "rollout_importance_sampling_threshold": 2.0,
203
+ "log_rollout_offpolicy_metrics": false,
204
+ "reward_model": null,
205
+ "reward_model_plugin": null,
206
+ "sync_ref_model": false,
207
+ "ref_model_sync_steps": 512,
208
+ "ref_model_mixup_alpha": 0.6,
209
+ "async_generate": false,
210
+ "move_model_batches": null,
211
+ "multi_turn_scheduler": null,
212
+ "max_turns": null,
213
+ "completion_length_limit_scope": "per_round",
214
+ "vllm_server_pass_dataset": false,
215
+ "log_entropy": false,
216
+ "top_entropy_quantile": 1.0,
217
+ "num_iterations": 1,
218
+ "check_model": false,
219
+ "padded_vocab_size": 152064,
220
+ "initialize_embedding": false,
221
+ "mlp_padding_free": false,
222
+ "load_safetensors": true,
223
+ "save_safetensors": true,
224
+ "ref_model": null,
225
+ "ref_adapters": [],
226
+ "merge_lora": true,
227
+ "max_shard_size": "5GB",
228
+ "dataloader_persistent_workers": true,
229
+ "dataloader_prefetch_factor": 10,
230
+ "architectures": "Qwen3OmniMoeForConditionalGeneration",
231
+ "llm_architectures": "Qwen3OmniMoeForConditionalGeneration",
232
+ "max_epochs": 2,
233
+ "enable_dft_loss": false,
234
+ "enable_channel_loss": false,
235
+ "save_strategy": "steps",
236
+ "original_max_position_embeddings": null,
237
+ "partial_rotary_factor": null,
238
+ "use_shared_expert_gate": false,
239
+ "vit_gradient_checkpointing": true,
240
+ "vit_lr": null,
241
+ "aligner_lr": null,
242
+ "gradient_checkpointing_kwargs": null,
243
+ "linear_num_value_heads": null,
244
+ "linear_num_key_heads": null,
245
+ "linear_key_head_dim": null,
246
+ "linear_value_head_dim": null,
247
+ "linear_conv_kernel_dim": null,
248
+ "layer_types": null,
249
+ "mrope_interleaved": true,
250
+ "micro_batch_size": 4,
251
+ "global_batch_size": 224,
252
+ "recompute_granularity": "selective",
253
+ "recompute_method": null,
254
+ "recompute_num_layers": null,
255
+ "recompute_modules": [
256
+ "core_attn"
257
+ ],
258
+ "use_cpu_initialization": false,
259
+ "deterministic_mode": false,
260
+ "train_iters": null,
261
+ "log_interval": 1,
262
+ "tensorboard_dir": "/hifs_gz11/share_303840540/user/chenxingli/intern/xianghe/qwen3omni_instruct_GRPO/v48-20260125-104404/runs",
263
+ "no_masked_softmax_fusion": false,
264
+ "no_bias_dropout_fusion": false,
265
+ "no_bias_swiglu_fusion": false,
266
+ "no_rope_fusion": false,
267
+ "no_gradient_accumulation_fusion": true,
268
+ "cross_entropy_loss_fusion": false,
269
+ "cross_entropy_fusion_impl": "native",
270
+ "calculate_per_token_loss": false,
271
+ "use_flash_attn": false,
272
+ "attention_backend": "flash",
273
+ "optimizer": "adam",
274
+ "optimizer_cpu_offload": true,
275
+ "optimizer_offload_fraction": 1.0,
276
+ "use_precision_aware_optimizer": true,
277
+ "main_grads_dtype": "fp32",
278
+ "main_params_dtype": "fp32",
279
+ "exp_avg_dtype": "fp32",
280
+ "exp_avg_sq_dtype": "fp32",
281
+ "dataloader_type": "cyclic",
282
+ "manual_gc": false,
283
+ "manual_gc_interval": 0,
284
+ "lr": 1e-06,
285
+ "lr_decay_style": "cosine",
286
+ "lr_decay_iters": null,
287
+ "lr_warmup_iters": 0,
288
+ "lr_warmup_fraction": 0.01,
289
+ "min_lr": 0,
290
+ "weight_decay": 0.1,
291
+ "clip_grad": 1.0,
292
+ "adam_beta1": 0.9,
293
+ "adam_beta2": 0.95,
294
+ "adam_eps": 1e-08,
295
+ "sgd_momentum": 0.9,
296
+ "save": "/hifs_gz11/share_303840540/user/chenxingli/intern/xianghe/qwen3omni_instruct_GRPO/v48-20260125-104404",
297
+ "save_interval": 50,
298
+ "save_retain_interval": null,
299
+ "no_save_optim": true,
300
+ "no_save_rng": true,
301
+ "load": null,
302
+ "no_load_optim": false,
303
+ "no_load_rng": false,
304
+ "finetune": true,
305
+ "ckpt_format": "torch_dist",
306
+ "no_initialization": true,
307
+ "auto_detect_ckpt_format": true,
308
+ "exit_on_missing_checkpoint": true,
309
+ "async_save": false,
310
+ "use_persistent_ckpt_worker": false,
311
+ "ckpt_fully_parallel_load": false,
312
+ "ckpt_assume_constant_structure": false,
313
+ "distributed_backend": "nccl",
314
+ "local_rank": 0,
315
+ "use_distributed_optimizer": true,
316
+ "tensor_model_parallel_size": 4,
317
+ "pipeline_model_parallel_size": 2,
318
+ "decoder_first_pipeline_num_layers": null,
319
+ "decoder_last_pipeline_num_layers": null,
320
+ "account_for_embedding_in_pipeline_split": false,
321
+ "account_for_loss_in_pipeline_split": false,
322
+ "sequence_parallel": true,
323
+ "context_parallel_size": 1,
324
+ "tp_comm_overlap": false,
325
+ "overlap_grad_reduce": false,
326
+ "overlap_param_gather": false,
327
+ "distributed_timeout_minutes": 300000,
328
+ "num_layers_per_virtual_pipeline_stage": null,
329
+ "num_virtual_stages_per_pipeline_rank": null,
330
+ "microbatch_group_size_per_virtual_pipeline_stage": null,
331
+ "pipeline_model_parallel_layout": null,
332
+ "num_layers": 48,
333
+ "hidden_size": 2048,
334
+ "ffn_hidden_size": null,
335
+ "num_attention_heads": 32,
336
+ "group_query_attention": true,
337
+ "num_query_groups": 4,
338
+ "softmax_type": "vanilla",
339
+ "window_size": null,
340
+ "window_attn_skip_freq": null,
341
+ "max_position_embeddings": 65536,
342
+ "position_embedding_type": "mrope",
343
+ "mrope_section": [
344
+ 24,
345
+ 20,
346
+ 20
347
+ ],
348
+ "rotary_base": 1000000,
349
+ "rotary_percent": 1.0,
350
+ "rotary_interleaved": false,
351
+ "normalization": "RMSNorm",
352
+ "norm_epsilon": 1e-06,
353
+ "swiglu": true,
354
+ "quick_geglu": false,
355
+ "activation_func_clamp_value": null,
356
+ "glu_linear_offset": 0.0,
357
+ "untie_embeddings_and_output_weights": true,
358
+ "disable_bias_linear": true,
359
+ "add_qkv_bias": false,
360
+ "attention_dropout": 0.0,
361
+ "hidden_dropout": 0.0,
362
+ "kv_channels": 128,
363
+ "qk_layernorm": true,
364
+ "qk_l2_norm": null,
365
+ "no_rope_freq": null,
366
+ "moe_apply_probs_on_input": null,
367
+ "transformer_impl": "transformer_engine",
368
+ "num_experts": 128,
369
+ "moe_layer_freq": "1",
370
+ "moe_ffn_hidden_size": 768,
371
+ "moe_shared_expert_intermediate_size": null,
372
+ "moe_router_topk": 8,
373
+ "moe_router_num_groups": null,
374
+ "moe_router_group_topk": null,
375
+ "moe_router_pre_softmax": false,
376
+ "moe_router_dtype": "fp32",
377
+ "moe_router_score_function": "softmax",
378
+ "moe_router_bias_update_rate": null,
379
+ "moe_router_enable_expert_bias": false,
380
+ "moe_router_topk_scaling_factor": null,
381
+ "moe_router_load_balancing_type": "aux_loss",
382
+ "expert_model_parallel_size": 4,
383
+ "expert_tensor_parallel_size": 1,
384
+ "moe_token_dispatcher_type": null,
385
+ "moe_enable_deepep": false,
386
+ "moe_grouped_gemm": true,
387
+ "moe_permute_fusion": true,
388
+ "moe_aux_loss_coeff": 0.0,
389
+ "moe_z_loss_coeff": null,
390
+ "moe_shared_expert_overlap": true,
391
+ "moe_layer_recompute": false,
392
+ "moe_expert_capacity_factor": null,
393
+ "moe_pad_expert_input_to_capacity": false,
394
+ "moe_token_drop_policy": null,
395
+ "multi_latent_attention": false,
396
+ "q_lora_rank": null,
397
+ "kv_lora_rank": 32,
398
+ "qk_head_dim": 128,
399
+ "qk_pos_emb_head_dim": 64,
400
+ "mtp_num_layers": null,
401
+ "mtp_loss_scaling_factor": 0.1,
402
+ "fp8_format": null,
403
+ "fp8_recipe": "delayed",
404
+ "fp8_amax_history_len": 1024,
405
+ "fp8_amax_compute_algo": "max",
406
+ "fp8_param_gather": false,
407
+ "fp16": false,
408
+ "bf16": true,
409
+ "apply_query_key_layer_scaling": false,
410
+ "attention_softmax_in_fp32": true,
411
+ "log_params_norm": false,
412
+ "log_throughput": false,
413
+ "tensorboard_log_interval": 1,
414
+ "tensorboard_queue_size": 50,
415
+ "log_timers_to_tensorboard": true,
416
+ "no_log_learning_rate_to_tensorboard": false,
417
+ "log_validation_ppl_to_tensorboard": true,
418
+ "log_memory_to_tensorboard": true,
419
+ "logging_level": null,
420
+ "wandb_project": null,
421
+ "wandb_exp_name": null,
422
+ "wandb_save_dir": null,
423
+ "eval_iters": -1,
424
+ "eval_interval": 50,
425
+ "seq_length": 4096,
426
+ "num_workers": 8,
427
+ "megatron_extra_kwargs": {},
428
+ "add_version": true,
429
+ "rank": 0,
430
+ "global_world_size": 56,
431
+ "local_world_size": 8,
432
+ "model_suffix": "checkpoint-500",
433
+ "model_info": "ModelInfo(model_type='qwen3_omni', model_dir='/hifs_gz11/share_303840540/user/chenxingli/intern/xianghe/qwen3omni_instruct_GRPO/v30-20260121-012721/checkpoint-500', torch_dtype=torch.bfloat16, max_model_len=65536, quant_method=None, quant_bits=None, rope_scaling={'interleaved': True, 'mrope_section': [24, 20, 20], 'rope_type': 'default', 'type': 'default'}, is_moe_model=True, is_multimodal=False, config=None, task_type='causal_lm', num_labels=None)",
434
+ "model_meta": "ModelMeta(model_type='qwen3_omni', model_groups=[ModelGroup(models=[Model(ms_model_id='Qwen/Qwen3-Omni-30B-A3B-Instruct', hf_model_id='Qwen/Qwen3-Omni-30B-A3B-Instruct', model_path=None, ms_revision=None, hf_revision=None), Model(ms_model_id='Qwen/Qwen3-Omni-30B-A3B-Thinking', hf_model_id='Qwen/Qwen3-Omni-30B-A3B-Thinking', model_path=None, ms_revision=None, hf_revision=None), Model(ms_model_id='Qwen/Qwen3-Omni-30B-A3B-Captioner', hf_model_id='Qwen/Qwen3-Omni-30B-A3B-Captioner', model_path=None, ms_revision=None, hf_revision=None)], ignore_patterns=None, requires=None, tags=[])], template='qwen3_omni', get_function=<function get_model_tokenizer_qwen3_omni at 0x7f613ee3c1f0>, model_arch=MultiModelKeys(arch_name='qwen3_omni', embedding=None, module_list=None, lm_head=None, q_proj=None, k_proj=None, v_proj=None, o_proj=None, attention=None, mlp=None, down_proj=None, qkv_proj=None, qk_proj=None, qa_proj=None, qb_proj=None, kv_proj=None, kva_proj=None, kvb_proj=None, language_model=['thinker.model', 'thinker.lm_head'], aligner=['thinker.audio_tower.proj1', 'thinker.audio_tower.proj2', 'thinker.visual.merger', 'thinker.visual.merger_list'], vision_tower=['thinker.audio_tower', 'thinker.visual'], generator=['talker', 'code2wav']), architectures=['Qwen3OmniMoeForConditionalGeneration'], additional_saved_files=[], torch_dtype=None, is_multimodal=True, is_reward=False, is_reranker=False, task_type=None, ignore_patterns=None, requires=['transformers>=4.57.dev0', 'soundfile', 'decord', 'qwen_omni_utils'], tags=['vision', 'video', 'audio'])",
435
+ "model_dir": "/hifs_gz11/share_303840540/user/chenxingli/intern/xianghe/qwen3omni_instruct_GRPO/v30-20260121-012721/checkpoint-500",
436
+ "_val_dataset_exists": [],
437
+ "hub": "<class 'swift.hub.hub.MSHub'>",
438
+ "megatron_model_meta": "MegatronModelMeta(megatron_model_type='qwen3_omni', model_types=['qwen3_omni'], is_multimodal=True, bridge_cls=<class 'swift.megatron.model.mm_gpt.qwen3_vl.Qwen3OmniBridge'>, model_cls=<class 'swift.megatron.model.mm_gpt.qwen3_vl.Qwen3VLGPTModel'>, get_transformer_layer_spec=None, model_provider=<function model_provider at 0x7f60d0f635b0>, visual_cls=<class 'swift.megatron.model.mm_gpt.qwen3_vl.Qwen3Omni_Vit'>, extra_args_provider=None)",
439
+ "per_device_generation_batch_size": 16,
440
+ "extra_args": {
441
+ "model_dir": "/hifs_gz11/share_303840540/user/chenxingli/intern/xianghe/qwen3omni_instruct_GRPO/v30-20260121-012721/checkpoint-500",
442
+ "is_multimodal": true,
443
+ "hf_model_type": "qwen3_omni",
444
+ "use_ray": false,
445
+ "ray_exp_name": null,
446
+ "device_groups": null,
447
+ "model": "/hifs_gz11/share_303840540/user/chenxingli/intern/xianghe/qwen3omni_instruct_GRPO/v30-20260121-012721/checkpoint-500/",
448
+ "model_type": "qwen3_omni",
449
+ "model_revision": null,
450
+ "task_type": "causal_lm",
451
+ "torch_dtype": "bfloat16",
452
+ "attn_impl": null,
453
+ "new_special_tokens": [],
454
+ "num_labels": null,
455
+ "problem_type": null,
456
+ "rope_scaling": {
457
+ "interleaved": true,
458
+ "mrope_interleaved": true,
459
+ "mrope_section": [
460
+ 24,
461
+ 20,
462
+ 20
463
+ ],
464
+ "rope_type": "default",
465
+ "type": "default"
466
+ },
467
+ "device_map": null,
468
+ "max_memory": {},
469
+ "max_model_len": null,
470
+ "local_repo_path": null,
471
+ "init_strategy": null,
472
+ "template": "qwen3_omni",
473
+ "system": "You are an audio understanding model that answers multiple choice questions based on audio content. You MUST always start your reasoning with <reasoning> and end with </reasoning> tags, followed by the exact option text, i.e., <reasoning>your detailed reasoning process</reasoning>\nselected option text",
474
+ "max_length": 4096,
475
+ "truncation_strategy": "left",
476
+ "max_pixels": null,
477
+ "agent_template": null,
478
+ "norm_bbox": null,
479
+ "use_chat_template": true,
480
+ "padding_free": true,
481
+ "padding_side": "right",
482
+ "sequence_parallel_size": 1,
483
+ "response_prefix": null,
484
+ "template_backend": "swift",
485
+ "dataset": [
486
+ "/apdcephfs_fsgm/share_303840540/user/chenxingli/intern/xianghe/codes/SpeechLM-RL/20260123_merged_shuffled_filtered.jsonl"
487
+ ],
488
+ "val_dataset": [],
489
+ "cached_dataset": [],
490
+ "cached_val_dataset": [],
491
+ "split_dataset_ratio": 0.0,
492
+ "data_seed": 42,
493
+ "dataset_num_proc": 8,
494
+ "load_from_cache_file": false,
495
+ "dataset_shuffle": true,
496
+ "val_dataset_shuffle": false,
497
+ "streaming": false,
498
+ "interleave_prob": null,
499
+ "stopping_strategy": "first_exhausted",
500
+ "shuffle_buffer_size": 1000,
501
+ "download_mode": "reuse_dataset_if_exists",
502
+ "columns": {},
503
+ "strict": false,
504
+ "remove_unused_columns": false,
505
+ "model_name": null,
506
+ "model_author": null,
507
+ "custom_dataset_info": [],
508
+ "quant_method": null,
509
+ "quant_bits": null,
510
+ "hqq_axis": null,
511
+ "bnb_4bit_compute_dtype": "bfloat16",
512
+ "bnb_4bit_quant_type": "nf4",
513
+ "bnb_4bit_use_double_quant": true,
514
+ "bnb_4bit_quant_storage": null,
515
+ "max_new_tokens": null,
516
+ "temperature": 1.0,
517
+ "top_k": 50,
518
+ "top_p": 0.99,
519
+ "repetition_penalty": 1.0,
520
+ "num_beams": 1,
521
+ "stream": false,
522
+ "stop_words": [],
523
+ "logprobs": false,
524
+ "top_logprobs": null,
525
+ "ckpt_dir": "/hifs_gz11/share_303840540/user/chenxingli/intern/xianghe/qwen3omni_instruct_GRPO/v30-20260121-012721/checkpoint-500/",
526
+ "lora_modules": [],
527
+ "tuner_backend": "peft",
528
+ "train_type": "full",
529
+ "adapters": [],
530
+ "external_plugins": [
531
+ "/apdcephfs_fsgm/share_303840540/user/chenxingli/intern/xianghe/codes/Audio-Thinker/GRPO/plugin/plugin.py"
532
+ ],
533
+ "model_kwargs": {},
534
+ "load_args": false,
535
+ "load_data_args": false,
536
+ "packing": false,
537
+ "packing_length": null,
538
+ "packing_num_proc": 1,
539
+ "lazy_tokenize": true,
540
+ "custom_register_path": [],
541
+ "use_hf": false,
542
+ "hub_token": null,
543
+ "ddp_timeout": 18000000,
544
+ "ddp_backend": null,
545
+ "ignore_args_error": false,
546
+ "use_swift_lora": false,
547
+ "freeze_llm": false,
548
+ "freeze_vit": false,
549
+ "freeze_aligner": false,
550
+ "freeze_parameters": [],
551
+ "freeze_parameters_regex": null,
552
+ "freeze_parameters_ratio": 0.0,
553
+ "trainable_parameters": [],
554
+ "trainable_parameters_regex": null,
555
+ "adapter_load": null,
556
+ "target_modules": [
557
+ "all-linear"
558
+ ],
559
+ "target_regex": null,
560
+ "modules_to_save": [],
561
+ "lora_rank": 8,
562
+ "lora_alpha": 32,
563
+ "lora_dropout": 0.05,
564
+ "lora_bias": "none",
565
+ "lora_dtype": null,
566
+ "use_rslora": false,
567
+ "rlhf_type": "grpo",
568
+ "ref_load": null,
569
+ "ref_adapter_load": null,
570
+ "beta": 0.001,
571
+ "rpo_alpha": null,
572
+ "reference_free": false,
573
+ "label_smoothing": 0.0,
574
+ "f_divergence_type": "reverse_kl",
575
+ "loss_type": "grpo",
576
+ "desirable_weight": 1.0,
577
+ "undesirable_weight": 1.0,
578
+ "calculate_KL": null,
579
+ "center_rewards_coefficient": null,
580
+ "generation_batch_size": 896,
581
+ "steps_per_generation": 4,
582
+ "num_generations": 8,
583
+ "max_completion_length": 1024,
584
+ "importance_sampling_level": "sequence",
585
+ "tau_pos": 1.0,
586
+ "tau_neg": 1.05,
587
+ "use_gdpo": true,
588
+ "epsilon": 0.2,
589
+ "epsilon_high": null,
590
+ "delta": null,
591
+ "use_vllm": true,
592
+ "vllm_mode": "server",
593
+ "vllm_enable_prefix_caching": true,
594
+ "vllm_gpu_memory_utilization": 0.9,
595
+ "vllm_tensor_parallel_size": 1,
596
+ "vllm_max_model_len": null,
597
+ "vllm_enforce_eager": false,
598
+ "vllm_limit_mm_per_prompt": null,
599
+ "vllm_disable_cascade_attn": false,
600
+ "vllm_max_num_seqs": null,
601
+ "vllm_mm_processor_cache_gb": null,
602
+ "vllm_engine_kwargs": {},
603
+ "sleep_level": 0,
604
+ "offload_optimizer": true,
605
+ "offload_model": true,
606
+ "offload_bridge": false,
607
+ "vllm_server_base_url": null,
608
+ "vllm_server_host": [
609
+ "29.164.185.85"
610
+ ],
611
+ "vllm_server_port": [
612
+ 8989
613
+ ],
614
+ "vllm_server_timeout": 240.0,
615
+ "vllm_server_group_port": null,
616
+ "reward_funcs": [
617
+ "external_acc_reward",
618
+ "external_mixedsimilarity_reward"
619
+ ],
620
+ "reward_weights": [
621
+ 1.0,
622
+ 0.5
623
+ ],
624
+ "cosine_min_len_value_wrong": -0.5,
625
+ "cosine_max_len_value_wrong": 0.0,
626
+ "cosine_min_len_value_correct": 1.0,
627
+ "cosine_max_len_value_correct": 0.5,
628
+ "cosine_max_len": null,
629
+ "repetition_n_grams": 3,
630
+ "repetition_max_penalty": -1.0,
631
+ "soft_max_length": null,
632
+ "soft_cache_length": null,
633
+ "dynamic_sample": false,
634
+ "max_resample_times": 3,
635
+ "overlong_filter": true,
636
+ "scale_rewards": "group",
637
+ "advantage_estimator": "grpo",
638
+ "kl_in_reward": false,
639
+ "wandb_log_unique_prompts": null,
640
+ "log_completions": true,
641
+ "rollout_importance_sampling_mode": null,
642
+ "rollout_importance_sampling_threshold": 2.0,
643
+ "log_rollout_offpolicy_metrics": false,
644
+ "reward_model": null,
645
+ "reward_model_plugin": null,
646
+ "sync_ref_model": false,
647
+ "ref_model_sync_steps": 512,
648
+ "ref_model_mixup_alpha": 0.6,
649
+ "async_generate": false,
650
+ "move_model_batches": null,
651
+ "multi_turn_scheduler": null,
652
+ "max_turns": null,
653
+ "completion_length_limit_scope": "per_round",
654
+ "vllm_server_pass_dataset": false,
655
+ "log_entropy": false,
656
+ "top_entropy_quantile": 1.0,
657
+ "num_iterations": 1,
658
+ "check_model": false,
659
+ "padded_vocab_size": 152064,
660
+ "initialize_embedding": false,
661
+ "mlp_padding_free": false,
662
+ "load_safetensors": true,
663
+ "save_safetensors": true,
664
+ "ref_model": null,
665
+ "ref_adapters": [],
666
+ "merge_lora": true,
667
+ "max_shard_size": "5GB",
668
+ "dataloader_persistent_workers": true,
669
+ "dataloader_prefetch_factor": 10,
670
+ "architectures": "Qwen3OmniMoeForConditionalGeneration",
671
+ "llm_architectures": "Qwen3OmniMoeForConditionalGeneration",
672
+ "max_epochs": 2,
673
+ "enable_dft_loss": false,
674
+ "enable_channel_loss": false,
675
+ "save_strategy": "steps",
676
+ "original_max_position_embeddings": null,
677
+ "partial_rotary_factor": null,
678
+ "use_shared_expert_gate": false,
679
+ "vit_gradient_checkpointing": true,
680
+ "vit_lr": null,
681
+ "aligner_lr": null,
682
+ "gradient_checkpointing_kwargs": null,
683
+ "linear_num_value_heads": null,
684
+ "linear_num_key_heads": null,
685
+ "linear_key_head_dim": null,
686
+ "linear_value_head_dim": null,
687
+ "linear_conv_kernel_dim": null,
688
+ "layer_types": null,
689
+ "mrope_interleaved": true,
690
+ "add_version": true
691
+ }
692
+ }
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 +1,557 @@
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
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
+ "hidden_act": "silu",
13
+ "hidden_size": 1024,
14
+ "intermediate_size": 3072,
15
+ "layer_scale_initial_scale": 0.01,
16
+ "max_position_embeddings": 8000,
17
+ "model_type": "",
18
+ "num_attention_heads": 16,
19
+ "num_hidden_layers": 8,
20
+ "num_key_value_heads": 16,
21
+ "num_quantizers": 16,
22
+ "num_semantic_quantizers": 1,
23
+ "rms_norm_eps": 1e-05,
24
+ "rope_theta": 10000,
25
+ "semantic_codebook_size": 4096,
26
+ "sliding_window": 72,
27
+ "upsample_rates": [
28
+ 8,
29
+ 5,
30
+ 4,
31
+ 3
32
+ ],
33
+ "upsampling_ratios": [
34
+ 2,
35
+ 2
36
+ ],
37
+ "vector_quantization_hidden_dimension": 512
38
+ },
39
+ "dtype": "bfloat16",
40
+ "enable_audio_output": false,
41
+ "hidden_size": 1024,
42
+ "im_end_token_id": 151645,
43
+ "im_start_token_id": 151644,
44
+ "keys_to_ignore_at_inference": [
45
+ "past_key_values",
46
+ "hidden_states",
47
+ "attention_mask",
48
+ "hidden_states",
49
+ "attention_mask"
50
+ ],
51
+ "model_type": "qwen3_omni_moe",
52
+ "pad_token_id": 151643,
53
+ "system_token_id": 8948,
54
+ "talker_config": {
55
+ "accept_hidden_layer": 24,
56
+ "audio_end_token_id": 151670,
57
+ "audio_start_token_id": 151669,
58
+ "audio_token_id": 151675,
59
+ "code_predictor_config": {
60
+ "_name_or_path": "",
61
+ "add_cross_attention": false,
62
+ "architectures": null,
63
+ "attention_bias": false,
64
+ "attention_dropout": 0,
65
+ "bad_words_ids": null,
66
+ "begin_suppress_tokens": null,
67
+ "bos_token_id": null,
68
+ "chunk_size_feed_forward": 0,
69
+ "cross_attention_hidden_size": null,
70
+ "decoder_start_token_id": null,
71
+ "diversity_penalty": 0.0,
72
+ "do_sample": false,
73
+ "dtype": null,
74
+ "early_stopping": false,
75
+ "encoder_no_repeat_ngram_size": 0,
76
+ "eos_token_id": null,
77
+ "exponential_decay_length_penalty": null,
78
+ "finetuning_task": null,
79
+ "forced_bos_token_id": null,
80
+ "forced_eos_token_id": null,
81
+ "head_dim": 128,
82
+ "hidden_act": "silu",
83
+ "hidden_size": 1024,
84
+ "id2label": {
85
+ "0": "LABEL_0",
86
+ "1": "LABEL_1"
87
+ },
88
+ "initializer_range": 0.02,
89
+ "intermediate_size": 3072,
90
+ "is_decoder": false,
91
+ "is_encoder_decoder": false,
92
+ "label2id": {
93
+ "LABEL_0": 0,
94
+ "LABEL_1": 1
95
+ },
96
+ "layer_types": [
97
+ "full_attention",
98
+ "full_attention",
99
+ "full_attention",
100
+ "full_attention",
101
+ "full_attention"
102
+ ],
103
+ "length_penalty": 1.0,
104
+ "max_length": 20,
105
+ "max_position_embeddings": 32768,
106
+ "max_window_layers": 28,
107
+ "min_length": 0,
108
+ "model_type": "qwen3_omni_moe_talker_code_predictor",
109
+ "no_repeat_ngram_size": 0,
110
+ "num_attention_heads": 16,
111
+ "num_beam_groups": 1,
112
+ "num_beams": 1,
113
+ "num_code_groups": 16,
114
+ "num_hidden_layers": 5,
115
+ "num_key_value_heads": 8,
116
+ "num_return_sequences": 1,
117
+ "output_attentions": false,
118
+ "output_hidden_states": false,
119
+ "output_scores": false,
120
+ "pad_token_id": null,
121
+ "prefix": null,
122
+ "problem_type": null,
123
+ "pruned_heads": {},
124
+ "remove_invalid_values": false,
125
+ "repetition_penalty": 1.0,
126
+ "return_dict": true,
127
+ "return_dict_in_generate": false,
128
+ "rms_norm_eps": 1e-06,
129
+ "rope_scaling": null,
130
+ "rope_theta": 1000000,
131
+ "sep_token_id": null,
132
+ "sliding_window": null,
133
+ "suppress_tokens": null,
134
+ "task_specific_params": null,
135
+ "temperature": 1.0,
136
+ "tf_legacy_loss": false,
137
+ "tie_encoder_decoder": false,
138
+ "tie_word_embeddings": false,
139
+ "tokenizer_class": null,
140
+ "top_k": 50,
141
+ "top_p": 1.0,
142
+ "torchscript": false,
143
+ "typical_p": 1.0,
144
+ "use_bfloat16": false,
145
+ "use_cache": true,
146
+ "use_sliding_window": false,
147
+ "vocab_size": 2048
148
+ },
149
+ "codec_bos_id": 2149,
150
+ "codec_eos_token_id": 2150,
151
+ "codec_nothink_id": 2155,
152
+ "codec_pad_id": 2148,
153
+ "codec_think_bos_id": 2156,
154
+ "codec_think_eos_id": 2157,
155
+ "image_token_id": 151655,
156
+ "model_type": "",
157
+ "num_code_groups": 16,
158
+ "output_router_logits": false,
159
+ "position_id_per_seconds": 13,
160
+ "seconds_per_chunk": 2,
161
+ "spatial_merge_size": 2,
162
+ "speaker_id": {
163
+ "aiden": 2303,
164
+ "chelsie": 2301,
165
+ "ethan": 2302
166
+ },
167
+ "text_config": {
168
+ "_name_or_path": "",
169
+ "add_cross_attention": false,
170
+ "architectures": null,
171
+ "attention_bias": false,
172
+ "attention_dropout": 0,
173
+ "bad_words_ids": null,
174
+ "begin_suppress_tokens": null,
175
+ "bos_token_id": null,
176
+ "chunk_size_feed_forward": 0,
177
+ "cross_attention_hidden_size": null,
178
+ "decoder_sparse_step": 1,
179
+ "decoder_start_token_id": null,
180
+ "diversity_penalty": 0.0,
181
+ "do_sample": false,
182
+ "dtype": "bfloat16",
183
+ "early_stopping": false,
184
+ "encoder_no_repeat_ngram_size": 0,
185
+ "eos_token_id": null,
186
+ "exponential_decay_length_penalty": null,
187
+ "finetuning_task": null,
188
+ "forced_bos_token_id": null,
189
+ "forced_eos_token_id": null,
190
+ "head_dim": 128,
191
+ "hidden_act": "silu",
192
+ "hidden_size": 1024,
193
+ "id2label": {
194
+ "0": "LABEL_0",
195
+ "1": "LABEL_1"
196
+ },
197
+ "initializer_range": 0.02,
198
+ "intermediate_size": 2048,
199
+ "is_decoder": false,
200
+ "is_encoder_decoder": false,
201
+ "label2id": {
202
+ "LABEL_0": 0,
203
+ "LABEL_1": 1
204
+ },
205
+ "length_penalty": 1.0,
206
+ "max_length": 20,
207
+ "max_position_embeddings": 65536,
208
+ "min_length": 0,
209
+ "mlp_only_layers": [],
210
+ "model_type": "qwen3_omni_moe_talker_text",
211
+ "moe_intermediate_size": 384,
212
+ "no_repeat_ngram_size": 0,
213
+ "norm_topk_prob": true,
214
+ "num_attention_heads": 16,
215
+ "num_beam_groups": 1,
216
+ "num_beams": 1,
217
+ "num_experts": 128,
218
+ "num_experts_per_tok": 6,
219
+ "num_hidden_layers": 20,
220
+ "num_key_value_heads": 2,
221
+ "num_return_sequences": 1,
222
+ "output_attentions": false,
223
+ "output_hidden_states": false,
224
+ "output_router_logits": false,
225
+ "output_scores": false,
226
+ "pad_token_id": null,
227
+ "prefix": null,
228
+ "problem_type": null,
229
+ "pruned_heads": {},
230
+ "remove_invalid_values": false,
231
+ "repetition_penalty": 1.0,
232
+ "return_dict": true,
233
+ "return_dict_in_generate": false,
234
+ "rms_norm_eps": 1e-06,
235
+ "rope_scaling": {
236
+ "interleaved": true,
237
+ "mrope_section": [
238
+ 24,
239
+ 20,
240
+ 20
241
+ ],
242
+ "rope_type": "default",
243
+ "type": "default"
244
+ },
245
+ "rope_theta": 1000000,
246
+ "router_aux_loss_coef": 0.001,
247
+ "sep_token_id": null,
248
+ "shared_expert_intermediate_size": 768,
249
+ "sliding_window": null,
250
+ "suppress_tokens": null,
251
+ "task_specific_params": null,
252
+ "temperature": 1.0,
253
+ "tie_encoder_decoder": false,
254
+ "tie_word_embeddings": false,
255
+ "tokenizer_class": null,
256
+ "top_k": 50,
257
+ "top_p": 1.0,
258
+ "torchscript": false,
259
+ "typical_p": 1.0,
260
+ "use_cache": true,
261
+ "use_sliding_window": false,
262
+ "vocab_size": 3072
263
+ },
264
+ "thinker_hidden_size": 2048,
265
+ "video_token_id": 151656,
266
+ "vision_start_token_id": 151652
267
+ },
268
+ "thinker_config": {
269
+ "audio_config": {
270
+ "_name_or_path": "",
271
+ "activation_dropout": 0,
272
+ "activation_function": "gelu",
273
+ "add_cross_attention": false,
274
+ "architectures": null,
275
+ "attention_dropout": 0,
276
+ "bad_words_ids": null,
277
+ "begin_suppress_tokens": null,
278
+ "bos_token_id": null,
279
+ "chunk_size_feed_forward": 0,
280
+ "conv_chunksize": 500,
281
+ "cross_attention_hidden_size": null,
282
+ "d_model": 1280,
283
+ "decoder_start_token_id": null,
284
+ "diversity_penalty": 0.0,
285
+ "do_sample": false,
286
+ "downsample_hidden_size": 480,
287
+ "dropout": 0,
288
+ "dtype": "bfloat16",
289
+ "early_stopping": false,
290
+ "encoder_attention_heads": 20,
291
+ "encoder_ffn_dim": 5120,
292
+ "encoder_layers": 32,
293
+ "encoder_no_repeat_ngram_size": 0,
294
+ "eos_token_id": null,
295
+ "exponential_decay_length_penalty": null,
296
+ "finetuning_task": null,
297
+ "forced_bos_token_id": null,
298
+ "forced_eos_token_id": null,
299
+ "id2label": {
300
+ "0": "LABEL_0",
301
+ "1": "LABEL_1"
302
+ },
303
+ "initializer_range": 0.02,
304
+ "is_decoder": false,
305
+ "is_encoder_decoder": false,
306
+ "label2id": {
307
+ "LABEL_0": 0,
308
+ "LABEL_1": 1
309
+ },
310
+ "length_penalty": 1.0,
311
+ "max_length": 20,
312
+ "max_source_positions": 1500,
313
+ "min_length": 0,
314
+ "model_type": "qwen3_omni_moe_audio_encoder",
315
+ "n_window": 50,
316
+ "n_window_infer": 800,
317
+ "no_repeat_ngram_size": 0,
318
+ "num_beam_groups": 1,
319
+ "num_beams": 1,
320
+ "num_hidden_layers": 32,
321
+ "num_mel_bins": 128,
322
+ "num_return_sequences": 1,
323
+ "output_attentions": false,
324
+ "output_dim": 2048,
325
+ "output_hidden_states": false,
326
+ "output_scores": false,
327
+ "pad_token_id": 151643,
328
+ "prefix": null,
329
+ "problem_type": null,
330
+ "pruned_heads": {},
331
+ "remove_invalid_values": false,
332
+ "repetition_penalty": 1.0,
333
+ "return_dict": true,
334
+ "return_dict_in_generate": false,
335
+ "scale_embedding": false,
336
+ "sep_token_id": null,
337
+ "suppress_tokens": null,
338
+ "task_specific_params": null,
339
+ "temperature": 1.0,
340
+ "tf_legacy_loss": false,
341
+ "tie_encoder_decoder": false,
342
+ "tie_word_embeddings": true,
343
+ "tokenizer_class": null,
344
+ "top_k": 50,
345
+ "top_p": 1.0,
346
+ "torchscript": false,
347
+ "typical_p": 1.0,
348
+ "use_bfloat16": false
349
+ },
350
+ "audio_end_token_id": 151670,
351
+ "audio_start_token_id": 151669,
352
+ "audio_token_index": 151675,
353
+ "dtype": "bfloat16",
354
+ "image_token_index": 151655,
355
+ "initializer_range": 0.02,
356
+ "model_type": "qwen3_omni_moe_thinker",
357
+ "pad_token_id": 151643,
358
+ "position_id_per_seconds": 13,
359
+ "seconds_per_chunk": 2,
360
+ "text_config": {
361
+ "_name_or_path": "",
362
+ "add_cross_attention": false,
363
+ "architectures": null,
364
+ "attention_bias": false,
365
+ "attention_dropout": 0.0,
366
+ "bad_words_ids": null,
367
+ "begin_suppress_tokens": null,
368
+ "bos_token_id": null,
369
+ "chunk_size_feed_forward": 0,
370
+ "cross_attention_hidden_size": null,
371
+ "decoder_sparse_step": 1,
372
+ "decoder_start_token_id": null,
373
+ "diversity_penalty": 0.0,
374
+ "do_sample": false,
375
+ "dtype": "bfloat16",
376
+ "early_stopping": false,
377
+ "encoder_no_repeat_ngram_size": 0,
378
+ "eos_token_id": null,
379
+ "exponential_decay_length_penalty": null,
380
+ "finetuning_task": null,
381
+ "forced_bos_token_id": null,
382
+ "forced_eos_token_id": null,
383
+ "head_dim": 128,
384
+ "hidden_act": "silu",
385
+ "hidden_size": 2048,
386
+ "id2label": {
387
+ "0": "LABEL_0",
388
+ "1": "LABEL_1"
389
+ },
390
+ "initializer_range": 0.02,
391
+ "intermediate_size": 768,
392
+ "is_decoder": false,
393
+ "is_encoder_decoder": false,
394
+ "label2id": {
395
+ "LABEL_0": 0,
396
+ "LABEL_1": 1
397
+ },
398
+ "length_penalty": 1.0,
399
+ "max_length": 20,
400
+ "max_position_embeddings": 65536,
401
+ "min_length": 0,
402
+ "mlp_only_layers": [],
403
+ "model_type": "qwen3_omni_moe_text",
404
+ "moe_intermediate_size": 768,
405
+ "no_repeat_ngram_size": 0,
406
+ "norm_topk_prob": true,
407
+ "num_attention_heads": 32,
408
+ "num_beam_groups": 1,
409
+ "num_beams": 1,
410
+ "num_experts": 128,
411
+ "num_experts_per_tok": 8,
412
+ "num_hidden_layers": 48,
413
+ "num_key_value_heads": 4,
414
+ "num_return_sequences": 1,
415
+ "output_attentions": false,
416
+ "output_hidden_states": false,
417
+ "output_router_logits": false,
418
+ "output_scores": false,
419
+ "pad_token_id": 151643,
420
+ "prefix": null,
421
+ "problem_type": null,
422
+ "pruned_heads": {},
423
+ "remove_invalid_values": false,
424
+ "repetition_penalty": 1.0,
425
+ "return_dict": true,
426
+ "return_dict_in_generate": false,
427
+ "rms_norm_eps": 1e-06,
428
+ "rope_scaling": {
429
+ "interleaved": true,
430
+ "mrope_interleaved": true,
431
+ "mrope_section": [
432
+ 24,
433
+ 20,
434
+ 20
435
+ ],
436
+ "rope_type": "default",
437
+ "type": "default"
438
+ },
439
+ "rope_theta": 1000000,
440
+ "router_aux_loss_coef": 0.001,
441
+ "sep_token_id": null,
442
+ "shared_expert_intermediate_size": 0,
443
+ "sliding_window": null,
444
+ "suppress_tokens": null,
445
+ "task_specific_params": null,
446
+ "temperature": 1.0,
447
+ "tf_legacy_loss": false,
448
+ "tie_encoder_decoder": false,
449
+ "tie_word_embeddings": false,
450
+ "tokenizer_class": null,
451
+ "top_k": 50,
452
+ "top_p": 1.0,
453
+ "torchscript": false,
454
+ "typical_p": 1.0,
455
+ "use_bfloat16": false,
456
+ "use_cache": true,
457
+ "use_qk_norm": true,
458
+ "use_sliding_window": false,
459
+ "vocab_size": 152064
460
+ },
461
+ "user_token_id": 872,
462
+ "video_token_index": 151656,
463
+ "vision_config": {
464
+ "_name_or_path": "",
465
+ "add_cross_attention": false,
466
+ "apply_vit_abs_pos_embed": true,
467
+ "architectures": null,
468
+ "bad_words_ids": null,
469
+ "begin_suppress_tokens": null,
470
+ "bos_token_id": null,
471
+ "chunk_size_feed_forward": 0,
472
+ "cross_attention_hidden_size": null,
473
+ "decoder_start_token_id": null,
474
+ "deepstack_visual_indexes": [
475
+ 8,
476
+ 16,
477
+ 24
478
+ ],
479
+ "depth": 27,
480
+ "diversity_penalty": 0.0,
481
+ "do_sample": false,
482
+ "dtype": "bfloat16",
483
+ "early_stopping": false,
484
+ "encoder_no_repeat_ngram_size": 0,
485
+ "eos_token_id": null,
486
+ "exponential_decay_length_penalty": null,
487
+ "finetuning_task": null,
488
+ "forced_bos_token_id": null,
489
+ "forced_eos_token_id": null,
490
+ "hidden_act": "gelu_pytorch_tanh",
491
+ "hidden_size": 1152,
492
+ "id2label": {
493
+ "0": "LABEL_0",
494
+ "1": "LABEL_1"
495
+ },
496
+ "image_size": 768,
497
+ "in_channels": 3,
498
+ "in_chans": 3,
499
+ "initializer_range": 0.02,
500
+ "intermediate_size": 4304,
501
+ "is_decoder": false,
502
+ "is_encoder_decoder": false,
503
+ "label2id": {
504
+ "LABEL_0": 0,
505
+ "LABEL_1": 1
506
+ },
507
+ "length_penalty": 1.0,
508
+ "max_length": 20,
509
+ "min_length": 0,
510
+ "model_type": "qwen3_omni_moe_vision_encoder",
511
+ "no_repeat_ngram_size": 0,
512
+ "num_beam_groups": 1,
513
+ "num_beams": 1,
514
+ "num_heads": 16,
515
+ "num_position_embeddings": 2304,
516
+ "num_return_sequences": 1,
517
+ "out_hidden_size": 2048,
518
+ "output_attentions": false,
519
+ "output_hidden_states": false,
520
+ "output_scores": false,
521
+ "pad_token_id": 151643,
522
+ "patch_size": 16,
523
+ "prefix": null,
524
+ "problem_type": null,
525
+ "pruned_heads": {},
526
+ "remove_invalid_values": false,
527
+ "repetition_penalty": 1.0,
528
+ "return_dict": true,
529
+ "return_dict_in_generate": false,
530
+ "sep_token_id": null,
531
+ "spatial_merge_size": 2,
532
+ "spatial_patch_size": 16,
533
+ "suppress_tokens": null,
534
+ "task_specific_params": null,
535
+ "temperature": 1.0,
536
+ "temporal_patch_size": 2,
537
+ "tf_legacy_loss": false,
538
+ "tie_encoder_decoder": false,
539
+ "tie_word_embeddings": true,
540
+ "tokenizer_class": null,
541
+ "tokens_per_second": 2,
542
+ "top_k": 50,
543
+ "top_p": 1.0,
544
+ "torchscript": false,
545
+ "typical_p": 1.0,
546
+ "use_bfloat16": false
547
+ },
548
+ "vision_end_token_id": 151653,
549
+ "vision_start_token_id": 151652
550
+ },
551
+ "transformers_version": "4.57.0.dev0",
552
+ "tts_bos_token_id": 151672,
553
+ "tts_eos_token_id": 151673,
554
+ "tts_pad_token_id": 151671,
555
+ "user_token_id": 872,
556
+ "vocab_size": 152064
557
+ }
infer_result/20260202-221152.jsonl ADDED
The diff for this file is too large to render. See raw diff
 
merges.txt ADDED
The diff for this file is too large to render. See raw diff
 
model-00001-of-00013.safetensors ADDED
@@ -0,0 +1,3 @@
 
 
 
 
1
+ version https://git-lfs.github.com/spec/v1
2
+ oid sha256:2eb46c16c137ace71cc66d66d009d0f1b0b3e40e641011ceffe96fdb74f2208f
3
+ size 4997903800
model-00002-of-00013.safetensors ADDED
@@ -0,0 +1,3 @@
 
 
 
 
1
+ version https://git-lfs.github.com/spec/v1
2
+ oid sha256:dfcf281714825c8391698bff74aa08560be523b4fd52eddc5bf48c5fd82feb47
3
+ size 4997754192
model-00003-of-00013.safetensors ADDED
@@ -0,0 +1,3 @@
 
 
 
 
1
+ version https://git-lfs.github.com/spec/v1
2
+ oid sha256:4dd58796eca091886cdcecf65e8525c6d957dbc9b8c475f8696a4b11976c591f
3
+ size 4997754192
model-00004-of-00013.safetensors ADDED
@@ -0,0 +1,3 @@
 
 
 
 
1
+ version https://git-lfs.github.com/spec/v1
2
+ oid sha256:7d167bdd97bae3b2db8f66413ee3805f36df856813af11e86c92be5c1cab7e2a
3
+ size 4997755616
model-00005-of-00013.safetensors ADDED
@@ -0,0 +1,3 @@
 
 
 
 
1
+ version https://git-lfs.github.com/spec/v1
2
+ oid sha256:a7bf76516bdccb2790b71e62139c4f97f05a0a658dc7ba1b6dfcba9fa19a9e2a
3
+ size 4997755768
model-00006-of-00013.safetensors ADDED
@@ -0,0 +1,3 @@
 
 
 
 
1
+ version https://git-lfs.github.com/spec/v1
2
+ oid sha256:c3245b0d5186afcfc1e3fc6c0aabc00f2f99ffdd04dc24628b33ba89f637163f
3
+ size 4997755768
model-00007-of-00013.safetensors ADDED
@@ -0,0 +1,3 @@
 
 
 
 
1
+ version https://git-lfs.github.com/spec/v1
2
+ oid sha256:1d6cda84c8917e25cd6449d1987c66e923e1da74ac7576375d9d70a28b051a30
3
+ size 4997755768
model-00008-of-00013.safetensors ADDED
@@ -0,0 +1,3 @@
 
 
 
 
1
+ version https://git-lfs.github.com/spec/v1
2
+ oid sha256:0901e1faf746a2837999618a8a7ba699849200f8a814b2bdb2944350128b890f
3
+ size 4997755768
model-00009-of-00013.safetensors ADDED
@@ -0,0 +1,3 @@
 
 
 
 
1
+ version https://git-lfs.github.com/spec/v1
2
+ oid sha256:fbbba68e27d2037a7cb9ef984daf8510503eee912267f93bb9da91fe878dfbe4
3
+ size 4997755776
model-00010-of-00013.safetensors ADDED
@@ -0,0 +1,3 @@
 
 
 
 
1
+ version https://git-lfs.github.com/spec/v1
2
+ oid sha256:7afad1a3812919a796bf9b0e0416c0b143f263aec5e437dec8922e573f0ab3e8
3
+ size 4997755776
model-00011-of-00013.safetensors ADDED
@@ -0,0 +1,3 @@
 
 
 
 
1
+ version https://git-lfs.github.com/spec/v1
2
+ oid sha256:61cb4c3d2e03052a209a4286872e92ab6e3595b7cc2de7f573b3f97ebb261f28
3
+ size 4997755776
model-00012-of-00013.safetensors ADDED
@@ -0,0 +1,3 @@
 
 
 
 
1
+ version https://git-lfs.github.com/spec/v1
2
+ oid sha256:5f353ec0c7f4687476931b2725225262b395475140db9df582e162c8078bae4c
3
+ size 4997755776
model-00013-of-00013.safetensors ADDED
@@ -0,0 +1,3 @@
 
 
 
 
1
+ version https://git-lfs.github.com/spec/v1
2
+ oid sha256:69aafe3cdd25148b013f2ff897d027ff6dba7cf9bfe08417e9a0938c0d687963
3
+ size 3467785392
model.safetensors.index.json ADDED
The diff for this file is too large to render. See raw diff
 
preprocessor_config.json ADDED
@@ -0,0 +1,30 @@
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
1
+ {
2
+ "dither": 0.0,
3
+ "feature_extractor_type": "WhisperFeatureExtractor",
4
+ "feature_size": 128,
5
+ "hop_length": 160,
6
+ "image_mean": [
7
+ 0.5,
8
+ 0.5,
9
+ 0.5
10
+ ],
11
+ "image_processor_type": "Qwen2VLImageProcessor",
12
+ "image_std": [
13
+ 0.5,
14
+ 0.5,
15
+ 0.5
16
+ ],
17
+ "max_pixels": 12845056,
18
+ "merge_size": 2,
19
+ "min_pixels": 3136,
20
+ "n_fft": 400,
21
+ "n_samples": 4800000,
22
+ "nb_max_frames": 30000,
23
+ "padding_side": "right",
24
+ "padding_value": 0.0,
25
+ "patch_size": 16,
26
+ "processor_class": "Qwen3OmniMoeProcessor",
27
+ "return_attention_mask": true,
28
+ "sampling_rate": 16000,
29
+ "temporal_patch_size": 2
30
+ }
special_tokens_map.json ADDED
@@ -0,0 +1,44 @@
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
1
+ {
2
+ "additional_special_tokens": [
3
+ "<|im_start|>",
4
+ "<|im_end|>",
5
+ "<|object_ref_start|>",
6
+ "<|object_ref_end|>",
7
+ "<|box_start|>",
8
+ "<|box_end|>",
9
+ "<|quad_start|>",
10
+ "<|quad_end|>",
11
+ "<|vision_start|>",
12
+ "<|vision_end|>",
13
+ "<|vision_pad|>",
14
+ "<|image_pad|>",
15
+ "<|video_pad|>",
16
+ "<|audio_start|>",
17
+ "<|audio_end|>",
18
+ "<tts_pad>",
19
+ "<tts_text_bos>",
20
+ "<tts_text_bos_single>",
21
+ "<|audio_pad|>"
22
+ ],
23
+ "audio_bos_token": "<|audio_start|>",
24
+ "audio_eos_token": "<|audio_end|>",
25
+ "audio_token": "<|audio_pad|>",
26
+ "eos_token": {
27
+ "content": "<|im_end|>",
28
+ "lstrip": false,
29
+ "normalized": false,
30
+ "rstrip": false,
31
+ "single_word": false
32
+ },
33
+ "image_token": "<|image_pad|>",
34
+ "pad_token": {
35
+ "content": "<|endoftext|>",
36
+ "lstrip": false,
37
+ "normalized": false,
38
+ "rstrip": false,
39
+ "single_word": false
40
+ },
41
+ "video_token": "<|video_pad|>",
42
+ "vision_bos_token": "<|vision_start|>",
43
+ "vision_eos_token": "<|vision_end|>"
44
+ }
tokenizer.json ADDED
@@ -0,0 +1,3 @@
 
 
 
 
1
+ version https://git-lfs.github.com/spec/v1
2
+ oid sha256:09267689b8362020b9763b65dd5be7e086b31e28d72e02837a9e781de9a91bc7
3
+ size 11423986
tokenizer_config.json ADDED
@@ -0,0 +1,317 @@
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
1
+ {
2
+ "add_bos_token": false,
3
+ "add_prefix_space": false,
4
+ "added_tokens_decoder": {
5
+ "151643": {
6
+ "content": "<|endoftext|>",
7
+ "lstrip": false,
8
+ "normalized": false,
9
+ "rstrip": false,
10
+ "single_word": false,
11
+ "special": true
12
+ },
13
+ "151644": {
14
+ "content": "<|im_start|>",
15
+ "lstrip": false,
16
+ "normalized": false,
17
+ "rstrip": false,
18
+ "single_word": false,
19
+ "special": true
20
+ },
21
+ "151645": {
22
+ "content": "<|im_end|>",
23
+ "lstrip": false,
24
+ "normalized": false,
25
+ "rstrip": false,
26
+ "single_word": false,
27
+ "special": true
28
+ },
29
+ "151646": {
30
+ "content": "<|object_ref_start|>",
31
+ "lstrip": false,
32
+ "normalized": false,
33
+ "rstrip": false,
34
+ "single_word": false,
35
+ "special": true
36
+ },
37
+ "151647": {
38
+ "content": "<|object_ref_end|>",
39
+ "lstrip": false,
40
+ "normalized": false,
41
+ "rstrip": false,
42
+ "single_word": false,
43
+ "special": true
44
+ },
45
+ "151648": {
46
+ "content": "<|box_start|>",
47
+ "lstrip": false,
48
+ "normalized": false,
49
+ "rstrip": false,
50
+ "single_word": false,
51
+ "special": true
52
+ },
53
+ "151649": {
54
+ "content": "<|box_end|>",
55
+ "lstrip": false,
56
+ "normalized": false,
57
+ "rstrip": false,
58
+ "single_word": false,
59
+ "special": true
60
+ },
61
+ "151650": {
62
+ "content": "<|quad_start|>",
63
+ "lstrip": false,
64
+ "normalized": false,
65
+ "rstrip": false,
66
+ "single_word": false,
67
+ "special": true
68
+ },
69
+ "151651": {
70
+ "content": "<|quad_end|>",
71
+ "lstrip": false,
72
+ "normalized": false,
73
+ "rstrip": false,
74
+ "single_word": false,
75
+ "special": true
76
+ },
77
+ "151652": {
78
+ "content": "<|vision_start|>",
79
+ "lstrip": false,
80
+ "normalized": false,
81
+ "rstrip": false,
82
+ "single_word": false,
83
+ "special": true
84
+ },
85
+ "151653": {
86
+ "content": "<|vision_end|>",
87
+ "lstrip": false,
88
+ "normalized": false,
89
+ "rstrip": false,
90
+ "single_word": false,
91
+ "special": true
92
+ },
93
+ "151654": {
94
+ "content": "<|vision_pad|>",
95
+ "lstrip": false,
96
+ "normalized": false,
97
+ "rstrip": false,
98
+ "single_word": false,
99
+ "special": true
100
+ },
101
+ "151655": {
102
+ "content": "<|image_pad|>",
103
+ "lstrip": false,
104
+ "normalized": false,
105
+ "rstrip": false,
106
+ "single_word": false,
107
+ "special": true
108
+ },
109
+ "151656": {
110
+ "content": "<|video_pad|>",
111
+ "lstrip": false,
112
+ "normalized": false,
113
+ "rstrip": false,
114
+ "single_word": false,
115
+ "special": true
116
+ },
117
+ "151657": {
118
+ "content": "<tool_call>",
119
+ "lstrip": false,
120
+ "normalized": false,
121
+ "rstrip": false,
122
+ "single_word": false,
123
+ "special": false
124
+ },
125
+ "151658": {
126
+ "content": "</tool_call>",
127
+ "lstrip": false,
128
+ "normalized": false,
129
+ "rstrip": false,
130
+ "single_word": false,
131
+ "special": false
132
+ },
133
+ "151659": {
134
+ "content": "<|fim_prefix|>",
135
+ "lstrip": false,
136
+ "normalized": false,
137
+ "rstrip": false,
138
+ "single_word": false,
139
+ "special": false
140
+ },
141
+ "151660": {
142
+ "content": "<|fim_middle|>",
143
+ "lstrip": false,
144
+ "normalized": false,
145
+ "rstrip": false,
146
+ "single_word": false,
147
+ "special": false
148
+ },
149
+ "151661": {
150
+ "content": "<|fim_suffix|>",
151
+ "lstrip": false,
152
+ "normalized": false,
153
+ "rstrip": false,
154
+ "single_word": false,
155
+ "special": false
156
+ },
157
+ "151662": {
158
+ "content": "<|fim_pad|>",
159
+ "lstrip": false,
160
+ "normalized": false,
161
+ "rstrip": false,
162
+ "single_word": false,
163
+ "special": false
164
+ },
165
+ "151663": {
166
+ "content": "<|repo_name|>",
167
+ "lstrip": false,
168
+ "normalized": false,
169
+ "rstrip": false,
170
+ "single_word": false,
171
+ "special": false
172
+ },
173
+ "151664": {
174
+ "content": "<|file_sep|>",
175
+ "lstrip": false,
176
+ "normalized": false,
177
+ "rstrip": false,
178
+ "single_word": false,
179
+ "special": false
180
+ },
181
+ "151665": {
182
+ "content": "<tool_response>",
183
+ "lstrip": false,
184
+ "normalized": false,
185
+ "rstrip": false,
186
+ "single_word": false,
187
+ "special": false
188
+ },
189
+ "151666": {
190
+ "content": "</tool_response>",
191
+ "lstrip": false,
192
+ "normalized": false,
193
+ "rstrip": false,
194
+ "single_word": false,
195
+ "special": false
196
+ },
197
+ "151667": {
198
+ "content": "<think>",
199
+ "lstrip": false,
200
+ "normalized": false,
201
+ "rstrip": false,
202
+ "single_word": false,
203
+ "special": false
204
+ },
205
+ "151668": {
206
+ "content": "</think>",
207
+ "lstrip": false,
208
+ "normalized": false,
209
+ "rstrip": false,
210
+ "single_word": false,
211
+ "special": false
212
+ },
213
+ "151669": {
214
+ "content": "<|audio_start|>",
215
+ "lstrip": false,
216
+ "normalized": false,
217
+ "rstrip": false,
218
+ "single_word": false,
219
+ "special": true
220
+ },
221
+ "151670": {
222
+ "content": "<|audio_end|>",
223
+ "lstrip": false,
224
+ "normalized": false,
225
+ "rstrip": false,
226
+ "single_word": false,
227
+ "special": true
228
+ },
229
+ "151671": {
230
+ "content": "<tts_pad>",
231
+ "lstrip": false,
232
+ "normalized": false,
233
+ "rstrip": false,
234
+ "single_word": false,
235
+ "special": true
236
+ },
237
+ "151672": {
238
+ "content": "<tts_text_bos>",
239
+ "lstrip": false,
240
+ "normalized": false,
241
+ "rstrip": false,
242
+ "single_word": false,
243
+ "special": true
244
+ },
245
+ "151673": {
246
+ "content": "<tts_text_eod>",
247
+ "lstrip": false,
248
+ "normalized": false,
249
+ "rstrip": false,
250
+ "single_word": false,
251
+ "special": true
252
+ },
253
+ "151674": {
254
+ "content": "<tts_text_bos_single>",
255
+ "lstrip": false,
256
+ "normalized": false,
257
+ "rstrip": false,
258
+ "single_word": false,
259
+ "special": true
260
+ },
261
+ "151675": {
262
+ "content": "<|audio_pad|>",
263
+ "lstrip": false,
264
+ "normalized": false,
265
+ "rstrip": false,
266
+ "single_word": false,
267
+ "special": true
268
+ }
269
+ },
270
+ "additional_special_tokens": [
271
+ "<|im_start|>",
272
+ "<|im_end|>",
273
+ "<|object_ref_start|>",
274
+ "<|object_ref_end|>",
275
+ "<|box_start|>",
276
+ "<|box_end|>",
277
+ "<|quad_start|>",
278
+ "<|quad_end|>",
279
+ "<|vision_start|>",
280
+ "<|vision_end|>",
281
+ "<|vision_pad|>",
282
+ "<|image_pad|>",
283
+ "<|video_pad|>",
284
+ "<|audio_start|>",
285
+ "<|audio_end|>",
286
+ "<tts_pad>",
287
+ "<tts_text_bos>",
288
+ "<tts_text_bos_single>",
289
+ "<|audio_pad|>"
290
+ ],
291
+ "audio_bos_token": "<|audio_start|>",
292
+ "audio_eos_token": "<|audio_end|>",
293
+ "audio_token": "<|audio_pad|>",
294
+ "bos_token": null,
295
+ "clean_up_tokenization_spaces": false,
296
+ "eos_token": "<|im_end|>",
297
+ "errors": "replace",
298
+ "extra_special_tokens": {
299
+ "audio_bos_token": "<|audio_start|>",
300
+ "audio_eos_token": "<|audio_end|>",
301
+ "audio_token": "<|audio_pad|>",
302
+ "image_token": "<|image_pad|>",
303
+ "video_token": "<|video_pad|>",
304
+ "vision_bos_token": "<|vision_start|>",
305
+ "vision_eos_token": "<|vision_end|>"
306
+ },
307
+ "image_token": "<|image_pad|>",
308
+ "model_max_length": 131072,
309
+ "pad_token": "<|endoftext|>",
310
+ "processor_class": "Qwen3OmniMoeProcessor",
311
+ "split_special_tokens": false,
312
+ "tokenizer_class": "Qwen2Tokenizer",
313
+ "unk_token": null,
314
+ "video_token": "<|video_pad|>",
315
+ "vision_bos_token": "<|vision_start|>",
316
+ "vision_eos_token": "<|vision_end|>"
317
+ }
video_preprocessor_config.json ADDED
@@ -0,0 +1,54 @@
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
1
+ {
2
+ "crop_size": null,
3
+ "data_format": "channels_first",
4
+ "default_to_square": true,
5
+ "device": null,
6
+ "dither": 0.0,
7
+ "do_center_crop": null,
8
+ "do_convert_rgb": true,
9
+ "do_normalize": true,
10
+ "do_rescale": true,
11
+ "do_resize": true,
12
+ "do_sample_frames": false,
13
+ "feature_extractor_type": "WhisperFeatureExtractor",
14
+ "feature_size": 128,
15
+ "fps": null,
16
+ "hop_length": 160,
17
+ "image_mean": [
18
+ 0.5,
19
+ 0.5,
20
+ 0.5
21
+ ],
22
+ "image_std": [
23
+ 0.5,
24
+ 0.5,
25
+ 0.5
26
+ ],
27
+ "input_data_format": null,
28
+ "max_frames": 768,
29
+ "max_pixels": 12845056,
30
+ "merge_size": 2,
31
+ "min_frames": 4,
32
+ "min_pixels": 3136,
33
+ "n_fft": 400,
34
+ "n_samples": 4800000,
35
+ "nb_max_frames": 30000,
36
+ "num_frames": null,
37
+ "pad_size": null,
38
+ "padding_side": "right",
39
+ "padding_value": 0.0,
40
+ "patch_size": 16,
41
+ "processor_class": "Qwen3OmniMoeProcessor",
42
+ "resample": 3,
43
+ "rescale_factor": 0.00392156862745098,
44
+ "return_attention_mask": true,
45
+ "return_metadata": false,
46
+ "sampling_rate": 16000,
47
+ "size": {
48
+ "longest_edge": 12845056,
49
+ "shortest_edge": 3136
50
+ },
51
+ "temporal_patch_size": 2,
52
+ "video_metadata": null,
53
+ "video_processor_type": "Qwen2VLVideoProcessor"
54
+ }
vocab.json ADDED
The diff for this file is too large to render. See raw diff