Borrison commited on
Commit
a4d67ec
·
verified ·
1 Parent(s): 633604f

Add MAPO checkpoint-500 model files

Browse files
.gitattributes CHANGED
@@ -33,3 +33,4 @@ saved_model/**/* filter=lfs diff=lfs merge=lfs -text
33
  *.zip filter=lfs diff=lfs merge=lfs -text
34
  *.zst filter=lfs diff=lfs merge=lfs -text
35
  *tfevents* filter=lfs diff=lfs merge=lfs -text
 
 
33
  *.zip filter=lfs diff=lfs merge=lfs -text
34
  *.zst filter=lfs diff=lfs merge=lfs -text
35
  *tfevents* filter=lfs diff=lfs merge=lfs -text
36
+ tokenizer.json filter=lfs diff=lfs merge=lfs -text
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 +1,784 @@
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
1
+ {
2
+ "use_ray": false,
3
+ "ray_exp_name": null,
4
+ "device_groups": null,
5
+ "model": "/apdcephfs_qy3/share_302401520/user/cihanxiao/cihanxiao/mapo/exp/mapo/Qwen3-Omni-30B-A3B-Thinking/v35-20260403-165326/checkpoint-900",
6
+ "model_type": "qwen3_omni_moe",
7
+ "model_revision": null,
8
+ "task_type": "causal_lm",
9
+ "torch_dtype": "bfloat16",
10
+ "attn_impl": null,
11
+ "experts_impl": null,
12
+ "new_special_tokens": [],
13
+ "num_labels": null,
14
+ "problem_type": null,
15
+ "rope_scaling": {
16
+ "interleaved": true,
17
+ "mrope_interleaved": true,
18
+ "mrope_section": [
19
+ 24,
20
+ 20,
21
+ 20
22
+ ],
23
+ "rope_type": "default",
24
+ "type": "default"
25
+ },
26
+ "device_map": null,
27
+ "max_memory": {},
28
+ "max_model_len": null,
29
+ "local_repo_path": null,
30
+ "init_strategy": null,
31
+ "template": "qwen3_omni",
32
+ "system": "",
33
+ "max_length": 32768,
34
+ "truncation_strategy": "left",
35
+ "max_pixels": null,
36
+ "agent_template": null,
37
+ "norm_bbox": null,
38
+ "use_chat_template": true,
39
+ "padding_side": "right",
40
+ "padding_free": true,
41
+ "loss_scale": "last_round",
42
+ "sequence_parallel_size": 1,
43
+ "template_backend": "swift",
44
+ "response_prefix": null,
45
+ "enable_thinking": null,
46
+ "add_non_thinking_prefix": true,
47
+ "dataset": [
48
+ "/apdcephfs_qy3/share_302401520/user/cihanxiao/cihanxiao/mapo/data/raw/stage2/20260123_merged_shuffled_filtered_qy.jsonl"
49
+ ],
50
+ "val_dataset": [],
51
+ "cached_dataset": [],
52
+ "cached_val_dataset": [],
53
+ "split_dataset_ratio": 0.0,
54
+ "data_seed": 42,
55
+ "dataset_num_proc": 2,
56
+ "load_from_cache_file": true,
57
+ "dataset_shuffle": true,
58
+ "val_dataset_shuffle": false,
59
+ "streaming": false,
60
+ "interleave_prob": null,
61
+ "stopping_strategy": "first_exhausted",
62
+ "shuffle_buffer_size": 1000,
63
+ "download_mode": "reuse_dataset_if_exists",
64
+ "columns": {},
65
+ "strict": false,
66
+ "remove_unused_columns": false,
67
+ "model_name": null,
68
+ "model_author": null,
69
+ "custom_dataset_info": [],
70
+ "quant_method": null,
71
+ "quant_bits": null,
72
+ "hqq_axis": null,
73
+ "bnb_4bit_compute_dtype": "bfloat16",
74
+ "bnb_4bit_quant_type": "nf4",
75
+ "bnb_4bit_use_double_quant": true,
76
+ "bnb_4bit_quant_storage": null,
77
+ "max_new_tokens": null,
78
+ "temperature": 1.2,
79
+ "top_k": 50,
80
+ "top_p": 0.99,
81
+ "repetition_penalty": 1.0,
82
+ "num_beams": 1,
83
+ "stream": false,
84
+ "stop_words": [],
85
+ "logprobs": false,
86
+ "top_logprobs": null,
87
+ "structured_outputs_regex": null,
88
+ "tuner_backend": "peft",
89
+ "tuner_type": "full",
90
+ "train_type": null,
91
+ "adapters": [],
92
+ "external_plugins": [
93
+ "/apdcephfs_qy3/share_302401520/user/cihanxiao/cihanxiao/mapo/src/examples/multi_node/grpo/plugin_think_plain.py"
94
+ ],
95
+ "custom_register_path": [],
96
+ "seed": 42,
97
+ "model_kwargs": {},
98
+ "load_args": false,
99
+ "load_data_args": false,
100
+ "packing": false,
101
+ "packing_length": null,
102
+ "packing_num_proc": 1,
103
+ "lazy_tokenize": true,
104
+ "use_hf": false,
105
+ "hub_token": null,
106
+ "ddp_timeout": 18000000,
107
+ "ddp_backend": null,
108
+ "ignore_args_error": false,
109
+ "use_swift_lora": false,
110
+ "freeze_llm": false,
111
+ "freeze_vit": false,
112
+ "freeze_aligner": false,
113
+ "freeze_parameters": [],
114
+ "freeze_parameters_regex": null,
115
+ "freeze_parameters_ratio": 0.0,
116
+ "trainable_parameters": [],
117
+ "trainable_parameters_regex": null,
118
+ "adapter_load": null,
119
+ "target_modules": [
120
+ "all-linear"
121
+ ],
122
+ "target_regex": null,
123
+ "modules_to_save": [],
124
+ "lora_rank": 128,
125
+ "lora_alpha": 512,
126
+ "lora_dropout": 0.05,
127
+ "lora_bias": "none",
128
+ "lora_dtype": null,
129
+ "use_rslora": false,
130
+ "rlhf_type": "mapo",
131
+ "ref_load": null,
132
+ "ref_adapter_load": null,
133
+ "beta": 0.001,
134
+ "rpo_alpha": null,
135
+ "reference_free": false,
136
+ "label_smoothing": 0.0,
137
+ "f_divergence_type": "reverse_kl",
138
+ "desirable_weight": 1.0,
139
+ "undesirable_weight": 1.0,
140
+ "calculate_KL": null,
141
+ "center_rewards_coefficient": null,
142
+ "teacher_model": null,
143
+ "teacher_model_type": null,
144
+ "teacher_model_revision": null,
145
+ "lmbda": 0.5,
146
+ "seq_kd": false,
147
+ "offload_teacher_model": false,
148
+ "sft_alpha": 0.0,
149
+ "generation_batch_size": 768,
150
+ "steps_per_generation": 4,
151
+ "num_generations": 8,
152
+ "num_generations_eval": null,
153
+ "max_completion_length": 1024,
154
+ "importance_sampling_level": "token",
155
+ "tau_pos": 1.0,
156
+ "tau_neg": 1.05,
157
+ "epsilon": 0.2,
158
+ "epsilon_high": null,
159
+ "delta": null,
160
+ "use_vllm": true,
161
+ "vllm_mode": "server",
162
+ "vllm_enable_prefix_caching": true,
163
+ "vllm_gpu_memory_utilization": 0.9,
164
+ "vllm_tensor_parallel_size": 1,
165
+ "vllm_max_model_len": null,
166
+ "vllm_enforce_eager": false,
167
+ "vllm_limit_mm_per_prompt": null,
168
+ "vllm_disable_cascade_attn": false,
169
+ "vllm_max_num_seqs": null,
170
+ "vllm_mm_processor_cache_gb": null,
171
+ "vllm_engine_kwargs": {},
172
+ "sleep_level": 0,
173
+ "offload_optimizer": false,
174
+ "offload_model": true,
175
+ "offload_bridge": false,
176
+ "vllm_server_base_url": null,
177
+ "vllm_server_host": [
178
+ "29.164.184.102"
179
+ ],
180
+ "vllm_server_port": [
181
+ 11086
182
+ ],
183
+ "vllm_server_timeout": 240.0,
184
+ "vllm_server_group_port": [
185
+ 51313
186
+ ],
187
+ "reward_funcs": [
188
+ "external_mcqa_accuracy",
189
+ "external_format",
190
+ "external_consistency"
191
+ ],
192
+ "reward_weights": [
193
+ 2.0,
194
+ 1.0,
195
+ 1.0
196
+ ],
197
+ "cosine_min_len_value_wrong": -0.5,
198
+ "cosine_max_len_value_wrong": 0.0,
199
+ "cosine_min_len_value_correct": 1.0,
200
+ "cosine_max_len_value_correct": 0.5,
201
+ "cosine_max_len": null,
202
+ "repetition_n_grams": 3,
203
+ "repetition_max_penalty": -1.0,
204
+ "soft_max_length": null,
205
+ "soft_cache_length": null,
206
+ "dynamic_sample": false,
207
+ "max_resample_times": 3,
208
+ "overlong_filter": true,
209
+ "scale_rewards": "group",
210
+ "advantage_estimator": "grpo",
211
+ "kl_in_reward": false,
212
+ "wandb_log_unique_prompts": null,
213
+ "log_completions": true,
214
+ "rollout_importance_sampling_mode": null,
215
+ "rollout_importance_sampling_threshold": 2.0,
216
+ "log_rollout_offpolicy_metrics": false,
217
+ "off_policy_sequence_mask_delta": null,
218
+ "log_entropy": false,
219
+ "top_entropy_quantile": 1.0,
220
+ "eta": 0.5,
221
+ "mapo_advantage_floor_eps": 0.1,
222
+ "mapo_task_fail_gate_floor": 0.0,
223
+ "mapo_mask_temperature": 1.0,
224
+ "mapo_temporal_kappa": 1.0,
225
+ "mapo_attention_layers": "43,44,45,46,47",
226
+ "mapo_attention_head_reduce": "max",
227
+ "mapo_attention_layer_reduce": "mean",
228
+ "mapo_failure_reward_name": "MCQAAccuracy",
229
+ "mapo_failure_threshold": 0.0,
230
+ "mapo_pos_tags": "NOUN,VERB,ADJ,ADV,NUM,X",
231
+ "mapo_attention_only": false,
232
+ "mapo_debug_attn_grad_probe": false,
233
+ "mapo_debug_attn_grad_probe_interval": 0,
234
+ "diff_objective": "softplus_margin",
235
+ "diff_kl_type": "forward",
236
+ "diff_margin_gamma": 2.0,
237
+ "diff_margin_beta": 4.0,
238
+ "entropy_mask_quantile": 0.25,
239
+ "entropy_mask_scope": "sequence",
240
+ "entropy_mask_type": "hard",
241
+ "text_only_modality_scope": "audio",
242
+ "text_ref_model": null,
243
+ "text_ref_load": null,
244
+ "reward_model": null,
245
+ "reward_model_plugin": null,
246
+ "sync_ref_model": false,
247
+ "ref_model_sync_steps": 512,
248
+ "ref_model_mixup_alpha": 0.6,
249
+ "async_generate": false,
250
+ "move_model_batches": null,
251
+ "multi_turn_scheduler": null,
252
+ "max_turns": null,
253
+ "completion_length_limit_scope": "per_round",
254
+ "vllm_server_pass_dataset": false,
255
+ "num_iterations": 1,
256
+ "loss_type": "grpo",
257
+ "check_model": true,
258
+ "padded_vocab_size": 152064,
259
+ "initialize_embedding": false,
260
+ "mlp_padding_free": false,
261
+ "load_safetensors": true,
262
+ "save_safetensors": true,
263
+ "ref_model": null,
264
+ "ref_adapters": [],
265
+ "merge_lora": false,
266
+ "max_shard_size": "5GB",
267
+ "train_dataloader_shuffle": true,
268
+ "dataloader_pin_memory": true,
269
+ "dataloader_persistent_workers": true,
270
+ "dataloader_prefetch_factor": 2,
271
+ "group_by_length": false,
272
+ "hf_model_type": "qwen3_omni_moe",
273
+ "llm_model_type": "qwen3_omni_moe_text",
274
+ "max_epochs": null,
275
+ "enable_dft_loss": false,
276
+ "enable_channel_loss": false,
277
+ "save_strategy": "steps",
278
+ "original_max_position_embeddings": null,
279
+ "partial_rotary_factor": null,
280
+ "use_shared_expert_gate": false,
281
+ "report_to": null,
282
+ "vit_gradient_checkpointing": true,
283
+ "vit_lr": null,
284
+ "aligner_lr": null,
285
+ "gradient_checkpointing_kwargs": null,
286
+ "linear_num_value_heads": null,
287
+ "linear_num_key_heads": null,
288
+ "linear_key_head_dim": null,
289
+ "linear_value_head_dim": null,
290
+ "linear_conv_kernel_dim": null,
291
+ "layer_types": null,
292
+ "apply_wd_to_qk_layernorm": false,
293
+ "mrope_interleaved": true,
294
+ "micro_batch_size": 4,
295
+ "global_batch_size": 192,
296
+ "recompute_granularity": "selective",
297
+ "recompute_method": null,
298
+ "recompute_num_layers": null,
299
+ "recompute_modules": [
300
+ "core_attn"
301
+ ],
302
+ "use_cpu_initialization": false,
303
+ "deterministic_mode": false,
304
+ "train_iters": 1000,
305
+ "log_interval": 1,
306
+ "tensorboard_dir": "/apdcephfs_qy3/share_302401520/user/cihanxiao/cihanxiao/mapo/exp/mapo/checkpoint-900/v5-20260406-174200/runs",
307
+ "no_masked_softmax_fusion": false,
308
+ "no_bias_dropout_fusion": false,
309
+ "no_bias_swiglu_fusion": false,
310
+ "no_rope_fusion": false,
311
+ "no_gradient_accumulation_fusion": true,
312
+ "cross_entropy_loss_fusion": false,
313
+ "cross_entropy_fusion_impl": "native",
314
+ "calculate_per_token_loss": false,
315
+ "use_flash_attn": false,
316
+ "attention_backend": "flash",
317
+ "optimizer": "adam",
318
+ "optimizer_cpu_offload": true,
319
+ "optimizer_offload_fraction": 1.0,
320
+ "use_precision_aware_optimizer": true,
321
+ "main_grads_dtype": "fp32",
322
+ "main_params_dtype": "fp32",
323
+ "exp_avg_dtype": "fp32",
324
+ "exp_avg_sq_dtype": "fp32",
325
+ "dataloader_type": "cyclic",
326
+ "manual_gc": false,
327
+ "manual_gc_interval": 0,
328
+ "lr": 1e-06,
329
+ "lr_decay_style": "cosine",
330
+ "lr_decay_iters": null,
331
+ "lr_warmup_iters": 0,
332
+ "lr_warmup_fraction": 0.05,
333
+ "min_lr": 0,
334
+ "weight_decay": 0.1,
335
+ "clip_grad": 1.0,
336
+ "adam_beta1": 0.9,
337
+ "adam_beta2": 0.95,
338
+ "adam_eps": 1e-08,
339
+ "sgd_momentum": 0.9,
340
+ "save": "/apdcephfs_qy3/share_302401520/user/cihanxiao/cihanxiao/mapo/exp/mapo/checkpoint-900/v5-20260406-174200",
341
+ "save_interval": 100,
342
+ "save_retain_interval": null,
343
+ "no_save_optim": true,
344
+ "no_save_rng": true,
345
+ "load": null,
346
+ "no_load_optim": false,
347
+ "no_load_rng": false,
348
+ "finetune": true,
349
+ "ckpt_format": "torch_dist",
350
+ "no_initialization": true,
351
+ "auto_detect_ckpt_format": true,
352
+ "exit_on_missing_checkpoint": true,
353
+ "async_save": false,
354
+ "use_persistent_ckpt_worker": false,
355
+ "ckpt_fully_parallel_load": false,
356
+ "ckpt_assume_constant_structure": false,
357
+ "distributed_backend": "nccl",
358
+ "local_rank": 0,
359
+ "use_distributed_optimizer": true,
360
+ "tensor_model_parallel_size": 4,
361
+ "pipeline_model_parallel_size": 2,
362
+ "decoder_first_pipeline_num_layers": null,
363
+ "decoder_last_pipeline_num_layers": null,
364
+ "account_for_embedding_in_pipeline_split": false,
365
+ "account_for_loss_in_pipeline_split": false,
366
+ "sequence_parallel": true,
367
+ "context_parallel_size": 1,
368
+ "tp_comm_overlap": false,
369
+ "overlap_grad_reduce": false,
370
+ "overlap_param_gather": false,
371
+ "distributed_timeout_minutes": 300000,
372
+ "num_layers_per_virtual_pipeline_stage": null,
373
+ "num_virtual_stages_per_pipeline_rank": null,
374
+ "microbatch_group_size_per_virtual_pipeline_stage": null,
375
+ "pipeline_model_parallel_layout": null,
376
+ "num_layers": 48,
377
+ "hidden_size": 2048,
378
+ "ffn_hidden_size": null,
379
+ "num_attention_heads": 32,
380
+ "group_query_attention": true,
381
+ "num_query_groups": 4,
382
+ "softmax_type": "vanilla",
383
+ "window_size": null,
384
+ "window_attn_skip_freq": null,
385
+ "max_position_embeddings": 65536,
386
+ "position_embedding_type": "mrope",
387
+ "mrope_section": [
388
+ 24,
389
+ 20,
390
+ 20
391
+ ],
392
+ "rotary_base": 1000000,
393
+ "rotary_percent": 1.0,
394
+ "rotary_interleaved": false,
395
+ "normalization": "RMSNorm",
396
+ "norm_epsilon": 1e-06,
397
+ "swiglu": true,
398
+ "quick_geglu": false,
399
+ "activation_func_clamp_value": null,
400
+ "glu_linear_offset": 0.0,
401
+ "untie_embeddings_and_output_weights": true,
402
+ "disable_bias_linear": true,
403
+ "add_qkv_bias": false,
404
+ "attention_dropout": 0.0,
405
+ "hidden_dropout": 0.0,
406
+ "kv_channels": 128,
407
+ "qk_layernorm": true,
408
+ "qk_l2_norm": null,
409
+ "no_rope_freq": null,
410
+ "moe_apply_probs_on_input": null,
411
+ "transformer_impl": "transformer_engine",
412
+ "num_experts": 128,
413
+ "moe_layer_freq": 1,
414
+ "moe_ffn_hidden_size": 768,
415
+ "moe_shared_expert_intermediate_size": null,
416
+ "moe_router_topk": 8,
417
+ "moe_router_num_groups": null,
418
+ "moe_router_group_topk": null,
419
+ "moe_router_pre_softmax": false,
420
+ "moe_router_dtype": "fp32",
421
+ "moe_router_score_function": "softmax",
422
+ "moe_router_bias_update_rate": null,
423
+ "moe_router_enable_expert_bias": false,
424
+ "moe_router_topk_scaling_factor": null,
425
+ "moe_router_load_balancing_type": "aux_loss",
426
+ "expert_model_parallel_size": 4,
427
+ "expert_tensor_parallel_size": 1,
428
+ "moe_token_dispatcher_type": null,
429
+ "moe_enable_deepep": false,
430
+ "moe_grouped_gemm": true,
431
+ "moe_permute_fusion": true,
432
+ "moe_aux_loss_coeff": 0.0001,
433
+ "moe_z_loss_coeff": null,
434
+ "moe_shared_expert_overlap": true,
435
+ "moe_layer_recompute": false,
436
+ "moe_expert_capacity_factor": null,
437
+ "moe_pad_expert_input_to_capacity": false,
438
+ "moe_token_drop_policy": null,
439
+ "multi_latent_attention": false,
440
+ "q_lora_rank": null,
441
+ "kv_lora_rank": 32,
442
+ "qk_head_dim": 128,
443
+ "qk_pos_emb_head_dim": 64,
444
+ "v_head_dim": 128,
445
+ "mtp_num_layers": null,
446
+ "mtp_loss_scaling_factor": 0.1,
447
+ "fp8_format": null,
448
+ "fp8_recipe": "delayed",
449
+ "fp8_amax_history_len": 1024,
450
+ "fp8_amax_compute_algo": "max",
451
+ "fp8_param_gather": false,
452
+ "fp16": false,
453
+ "bf16": true,
454
+ "apply_query_key_layer_scaling": false,
455
+ "attention_softmax_in_fp32": true,
456
+ "log_params_norm": false,
457
+ "log_throughput": false,
458
+ "tensorboard_log_interval": 1,
459
+ "tensorboard_queue_size": 50,
460
+ "log_timers_to_tensorboard": true,
461
+ "no_log_learning_rate_to_tensorboard": false,
462
+ "log_validation_ppl_to_tensorboard": true,
463
+ "log_memory_to_tensorboard": true,
464
+ "logging_level": null,
465
+ "wandb_project": "megatron-swift",
466
+ "wandb_exp_name": "/apdcephfs_qy3/share_302401520/user/cihanxiao/cihanxiao/mapo/exp/mapo/checkpoint-900/v5-20260406-174200",
467
+ "wandb_save_dir": null,
468
+ "eval_iters": -1,
469
+ "eval_interval": 200,
470
+ "seq_length": 32768,
471
+ "num_workers": 2,
472
+ "no_data_sharding": false,
473
+ "megatron_extra_kwargs": {},
474
+ "add_version": true,
475
+ "swift_version": "4.0.0.dev0",
476
+ "ckpt_dir": "/apdcephfs_qy3/share_302401520/user/cihanxiao/cihanxiao/mapo/exp/mapo/Qwen3-Omni-30B-A3B-Thinking/v35-20260403-165326/checkpoint-900",
477
+ "rank": 0,
478
+ "global_world_size": 24,
479
+ "local_world_size": 8,
480
+ "model_suffix": "checkpoint-900",
481
+ "model_info": "ModelInfo(model_type='qwen3_omni_moe', model_dir='/apdcephfs_qy3/share_302401520/user/cihanxiao/cihanxiao/mapo/exp/mapo/Qwen3-Omni-30B-A3B-Thinking/v35-20260403-165326/checkpoint-900', torch_dtype=torch.bfloat16, max_model_len=32768, quant_method=None, quant_bits=None, rope_scaling=None, is_moe_model=True, is_multimodal=True, config=None, task_type='causal_lm', num_labels=None)",
482
+ "model_meta": "ModelMeta(model_type='qwen3_omni_moe', 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)], template='qwen3_omni', ignore_patterns=None, requires=None, tags=[])], loader=<class 'swift.model.models.qwen.Qwen3OmniLoader'>, template=None, 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, task_type=None, ignore_patterns=None, requires=['transformers>=4.57.dev0', 'soundfile', 'decord', 'qwen_omni_utils'], tags=['vision', 'video', 'audio'])",
483
+ "model_dir": "/apdcephfs_qy3/share_302401520/user/cihanxiao/cihanxiao/mapo/exp/mapo/Qwen3-Omni-30B-A3B-Thinking/v35-20260403-165326/checkpoint-900",
484
+ "template_meta": "QwenTemplateMeta(template_type='qwen3_omni', prefix=[], prompt=['<|im_start|>user\\n{{QUERY}}<|im_end|>\\n<|im_start|>assistant\\n'], chat_sep=['<|im_end|>\\n'], suffix=['<|im_end|>\\n'], template_cls=<class 'swift.template.templates.qwen.Qwen3OmniTemplate'>, system_prefix=['<|im_start|>system\\n{{SYSTEM}}<|im_end|>\\n'], default_system=None, auto_add_bos=False, stop_words=['<|endoftext|>'], agent_template='hermes', is_thinking=False, thinking_prefix='<think>\\n', non_thinking_prefix='', history_thinking_prefix='')",
485
+ "_val_dataset_exists": [],
486
+ "hub": "<class 'swift.hub.hub.MSHub'>",
487
+ "megatron_model_meta": "MegatronModelMeta(megatron_model_type='qwen3_omni', model_types=['qwen3_omni_moe'], is_multimodal=True, bridge_cls=<class 'swift.megatron.model.mm_gpts.qwen3_vl.Qwen3OmniBridge'>, model_cls=<class 'swift.megatron.model.mm_gpts.qwen3_vl.Qwen3VLGPTModel'>, get_transformer_layer_spec=None, model_provider=<function model_provider at 0x7fee5e3642c0>, visual_cls=<class 'swift.megatron.model.mm_gpts.qwen3_vl.Qwen3Omni_Vit'>, get_mtp_block_spec=None, extra_args_provider=None)",
488
+ "per_device_generation_batch_size": 32,
489
+ "extra_args": {
490
+ "model_dir": "/apdcephfs_qy3/share_302401520/user/cihanxiao/cihanxiao/mapo/exp/mapo/Qwen3-Omni-30B-A3B-Thinking/v35-20260403-165326/checkpoint-900",
491
+ "is_multimodal": true,
492
+ "hf_model_type": "qwen3_omni_moe",
493
+ "use_ray": false,
494
+ "ray_exp_name": null,
495
+ "device_groups": null,
496
+ "model": "/apdcephfs_qy3/share_302401520/user/cihanxiao/cihanxiao/mapo/exp/mapo/Qwen3-Omni-30B-A3B-Thinking/v35-20260403-165326/checkpoint-900",
497
+ "model_type": "qwen3_omni_moe",
498
+ "model_revision": null,
499
+ "task_type": "causal_lm",
500
+ "torch_dtype": "bfloat16",
501
+ "attn_impl": null,
502
+ "experts_impl": null,
503
+ "new_special_tokens": [],
504
+ "num_labels": null,
505
+ "problem_type": null,
506
+ "rope_scaling": {
507
+ "interleaved": true,
508
+ "mrope_interleaved": true,
509
+ "mrope_section": [
510
+ 24,
511
+ 20,
512
+ 20
513
+ ],
514
+ "rope_type": "default",
515
+ "type": "default"
516
+ },
517
+ "device_map": null,
518
+ "max_memory": {},
519
+ "max_model_len": null,
520
+ "local_repo_path": null,
521
+ "init_strategy": null,
522
+ "template": "qwen3_omni",
523
+ "system": "",
524
+ "max_length": 32768,
525
+ "truncation_strategy": "left",
526
+ "max_pixels": null,
527
+ "agent_template": null,
528
+ "norm_bbox": null,
529
+ "use_chat_template": true,
530
+ "padding_side": "right",
531
+ "padding_free": true,
532
+ "sequence_parallel_size": 1,
533
+ "template_backend": "swift",
534
+ "response_prefix": null,
535
+ "enable_thinking": null,
536
+ "add_non_thinking_prefix": true,
537
+ "dataset": [
538
+ "/apdcephfs_qy3/share_302401520/user/cihanxiao/cihanxiao/mapo/data/raw/stage2/20260123_merged_shuffled_filtered_qy.jsonl"
539
+ ],
540
+ "val_dataset": [],
541
+ "cached_dataset": [],
542
+ "cached_val_dataset": [],
543
+ "split_dataset_ratio": 0.0,
544
+ "data_seed": 42,
545
+ "dataset_num_proc": 2,
546
+ "load_from_cache_file": true,
547
+ "dataset_shuffle": true,
548
+ "val_dataset_shuffle": false,
549
+ "streaming": false,
550
+ "interleave_prob": null,
551
+ "stopping_strategy": "first_exhausted",
552
+ "shuffle_buffer_size": 1000,
553
+ "download_mode": "reuse_dataset_if_exists",
554
+ "columns": {},
555
+ "strict": false,
556
+ "remove_unused_columns": false,
557
+ "model_name": null,
558
+ "model_author": null,
559
+ "custom_dataset_info": [],
560
+ "quant_method": null,
561
+ "quant_bits": null,
562
+ "hqq_axis": null,
563
+ "bnb_4bit_compute_dtype": "bfloat16",
564
+ "bnb_4bit_quant_type": "nf4",
565
+ "bnb_4bit_use_double_quant": true,
566
+ "bnb_4bit_quant_storage": null,
567
+ "max_new_tokens": null,
568
+ "temperature": 1.2,
569
+ "top_k": 50,
570
+ "top_p": 0.99,
571
+ "repetition_penalty": 1.0,
572
+ "num_beams": 1,
573
+ "stream": false,
574
+ "stop_words": [],
575
+ "logprobs": false,
576
+ "top_logprobs": null,
577
+ "structured_outputs_regex": null,
578
+ "tuner_backend": "peft",
579
+ "tuner_type": "full",
580
+ "train_type": null,
581
+ "adapters": [],
582
+ "external_plugins": [
583
+ "/apdcephfs_qy3/share_302401520/user/cihanxiao/cihanxiao/mapo/src/examples/multi_node/grpo/plugin_think_plain.py"
584
+ ],
585
+ "custom_register_path": [],
586
+ "model_kwargs": {},
587
+ "load_args": false,
588
+ "load_data_args": false,
589
+ "packing": false,
590
+ "packing_length": null,
591
+ "packing_num_proc": 1,
592
+ "lazy_tokenize": true,
593
+ "use_hf": false,
594
+ "hub_token": null,
595
+ "ddp_timeout": 18000000,
596
+ "ddp_backend": null,
597
+ "ignore_args_error": false,
598
+ "use_swift_lora": false,
599
+ "freeze_llm": false,
600
+ "freeze_vit": false,
601
+ "freeze_aligner": false,
602
+ "freeze_parameters": [],
603
+ "freeze_parameters_regex": null,
604
+ "freeze_parameters_ratio": 0.0,
605
+ "trainable_parameters": [],
606
+ "trainable_parameters_regex": null,
607
+ "adapter_load": null,
608
+ "target_modules": [
609
+ "all-linear"
610
+ ],
611
+ "target_regex": null,
612
+ "modules_to_save": [],
613
+ "lora_rank": 128,
614
+ "lora_alpha": 512,
615
+ "lora_dropout": 0.05,
616
+ "lora_bias": "none",
617
+ "lora_dtype": null,
618
+ "use_rslora": false,
619
+ "rlhf_type": "mapo",
620
+ "ref_load": null,
621
+ "ref_adapter_load": null,
622
+ "beta": 0.001,
623
+ "rpo_alpha": null,
624
+ "reference_free": false,
625
+ "label_smoothing": 0.0,
626
+ "f_divergence_type": "reverse_kl",
627
+ "desirable_weight": 1.0,
628
+ "undesirable_weight": 1.0,
629
+ "calculate_KL": null,
630
+ "center_rewards_coefficient": null,
631
+ "teacher_model": null,
632
+ "teacher_model_type": null,
633
+ "teacher_model_revision": null,
634
+ "lmbda": 0.5,
635
+ "seq_kd": false,
636
+ "offload_teacher_model": false,
637
+ "sft_alpha": 0.0,
638
+ "generation_batch_size": 768,
639
+ "steps_per_generation": 4,
640
+ "num_generations": 8,
641
+ "num_generations_eval": null,
642
+ "max_completion_length": 1024,
643
+ "importance_sampling_level": "token",
644
+ "tau_pos": 1.0,
645
+ "tau_neg": 1.05,
646
+ "epsilon": 0.2,
647
+ "epsilon_high": null,
648
+ "delta": null,
649
+ "use_vllm": true,
650
+ "vllm_mode": "server",
651
+ "vllm_enable_prefix_caching": true,
652
+ "vllm_gpu_memory_utilization": 0.9,
653
+ "vllm_tensor_parallel_size": 1,
654
+ "vllm_max_model_len": null,
655
+ "vllm_enforce_eager": false,
656
+ "vllm_limit_mm_per_prompt": null,
657
+ "vllm_disable_cascade_attn": false,
658
+ "vllm_max_num_seqs": null,
659
+ "vllm_mm_processor_cache_gb": null,
660
+ "vllm_engine_kwargs": {},
661
+ "sleep_level": 0,
662
+ "offload_optimizer": false,
663
+ "offload_model": true,
664
+ "offload_bridge": false,
665
+ "vllm_server_base_url": null,
666
+ "vllm_server_host": [
667
+ "29.164.184.102"
668
+ ],
669
+ "vllm_server_port": [
670
+ 11086
671
+ ],
672
+ "vllm_server_timeout": 240.0,
673
+ "vllm_server_group_port": [
674
+ 51313
675
+ ],
676
+ "reward_funcs": [
677
+ "external_mcqa_accuracy",
678
+ "external_format",
679
+ "external_consistency"
680
+ ],
681
+ "reward_weights": [
682
+ 2.0,
683
+ 1.0,
684
+ 1.0
685
+ ],
686
+ "cosine_min_len_value_wrong": -0.5,
687
+ "cosine_max_len_value_wrong": 0.0,
688
+ "cosine_min_len_value_correct": 1.0,
689
+ "cosine_max_len_value_correct": 0.5,
690
+ "cosine_max_len": null,
691
+ "repetition_n_grams": 3,
692
+ "repetition_max_penalty": -1.0,
693
+ "soft_max_length": null,
694
+ "soft_cache_length": null,
695
+ "dynamic_sample": false,
696
+ "max_resample_times": 3,
697
+ "overlong_filter": true,
698
+ "scale_rewards": "group",
699
+ "advantage_estimator": "grpo",
700
+ "kl_in_reward": false,
701
+ "wandb_log_unique_prompts": null,
702
+ "log_completions": true,
703
+ "rollout_importance_sampling_mode": null,
704
+ "rollout_importance_sampling_threshold": 2.0,
705
+ "log_rollout_offpolicy_metrics": false,
706
+ "off_policy_sequence_mask_delta": null,
707
+ "log_entropy": false,
708
+ "top_entropy_quantile": 1.0,
709
+ "eta": 0.5,
710
+ "mapo_advantage_floor_eps": 0.1,
711
+ "mapo_task_fail_gate_floor": 0.0,
712
+ "mapo_mask_temperature": 1.0,
713
+ "mapo_temporal_kappa": 1.0,
714
+ "mapo_attention_layers": "43,44,45,46,47",
715
+ "mapo_attention_head_reduce": "max",
716
+ "mapo_attention_layer_reduce": "mean",
717
+ "mapo_failure_reward_name": "MCQAAccuracy",
718
+ "mapo_failure_threshold": 0.0,
719
+ "mapo_pos_tags": "NOUN,VERB,ADJ,ADV,NUM,X",
720
+ "mapo_attention_only": false,
721
+ "mapo_debug_attn_grad_probe": false,
722
+ "mapo_debug_attn_grad_probe_interval": 0,
723
+ "diff_objective": "softplus_margin",
724
+ "diff_kl_type": "forward",
725
+ "diff_margin_gamma": 2.0,
726
+ "diff_margin_beta": 4.0,
727
+ "entropy_mask_quantile": 0.25,
728
+ "entropy_mask_scope": "sequence",
729
+ "entropy_mask_type": "hard",
730
+ "text_only_modality_scope": "audio",
731
+ "text_ref_model": null,
732
+ "text_ref_load": null,
733
+ "reward_model": null,
734
+ "reward_model_plugin": null,
735
+ "sync_ref_model": false,
736
+ "ref_model_sync_steps": 512,
737
+ "ref_model_mixup_alpha": 0.6,
738
+ "async_generate": false,
739
+ "move_model_batches": null,
740
+ "multi_turn_scheduler": null,
741
+ "max_turns": null,
742
+ "completion_length_limit_scope": "per_round",
743
+ "vllm_server_pass_dataset": false,
744
+ "num_iterations": 1,
745
+ "loss_type": "grpo",
746
+ "check_model": true,
747
+ "padded_vocab_size": 152064,
748
+ "initialize_embedding": false,
749
+ "mlp_padding_free": false,
750
+ "load_safetensors": true,
751
+ "save_safetensors": true,
752
+ "ref_model": null,
753
+ "ref_adapters": [],
754
+ "merge_lora": false,
755
+ "max_shard_size": "5GB",
756
+ "train_dataloader_shuffle": true,
757
+ "dataloader_pin_memory": true,
758
+ "dataloader_persistent_workers": true,
759
+ "dataloader_prefetch_factor": 2,
760
+ "group_by_length": false,
761
+ "llm_model_type": "qwen3_omni_moe_text",
762
+ "max_epochs": null,
763
+ "enable_dft_loss": false,
764
+ "enable_channel_loss": false,
765
+ "save_strategy": "steps",
766
+ "original_max_position_embeddings": null,
767
+ "partial_rotary_factor": null,
768
+ "use_shared_expert_gate": false,
769
+ "report_to": null,
770
+ "vit_gradient_checkpointing": true,
771
+ "vit_lr": null,
772
+ "aligner_lr": null,
773
+ "gradient_checkpointing_kwargs": null,
774
+ "linear_num_value_heads": null,
775
+ "linear_num_key_heads": null,
776
+ "linear_key_head_dim": null,
777
+ "linear_value_head_dim": null,
778
+ "linear_conv_kernel_dim": null,
779
+ "layer_types": null,
780
+ "apply_wd_to_qk_layernorm": false,
781
+ "mrope_interleaved": true,
782
+ "add_version": true
783
+ }
784
+ }
chat_template.jinja ADDED
@@ -0,0 +1,105 @@
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
1
+ {%- if tools %}
2
+ {{- '<|im_start|>system\n' }}
3
+ {%- if messages[0].role == 'system' %}{{- messages[0].content + '\n\n' }}{%- endif %}
4
+ {{- "# 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>" }}
5
+ {%- for tool in tools %}
6
+ {{- "\n" }}
7
+ {{- tool | tojson }}
8
+ {%- endfor %}
9
+ {{- "\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" }}
10
+ {%- else %}
11
+ {%- if messages[0].role == 'system' %}
12
+ {%- if messages[0].content is string %}
13
+ {{- '<|im_start|>system\n' + messages[0].content + '<|im_end|>\n' }}
14
+ {%- else %}
15
+ {%- for content in messages[0].content %}
16
+ {%- if content.type == 'image' or 'image' in content or 'image_url' in content %}
17
+ {{- '<|im_start|>system\n' +"<|vision_start|><|image_pad|><|vision_end|>"+ '<|im_end|>\n' }}
18
+ {%- elif content.type == 'audio' or 'audio' in content or 'audio_url' in content %}
19
+ {{- '<|im_start|>system\n' +"<|audio_start|><|audio_pad|><|audio_end|>"+ '<|im_end|>\n' }}
20
+ {%- elif content.type == 'video' or 'video' in content %}
21
+ {{- '<|im_start|>system\n' +"<|vision_start|><|video_pad|><|vision_end|>"+ '<|im_end|>\n' }}
22
+ {%- elif content.type == 'text' %}
23
+ {{- '<|im_start|>system\n' +content.text+ '<|im_end|>\n' }}
24
+ {%- endif %}
25
+ {%- endfor %}
26
+ {%- endif %}
27
+ {%- endif %}
28
+ {%- endif %}
29
+ {%- set ns = namespace(multi_step_tool=true, last_query_index=messages|length - 1) %}
30
+ {%- for message in messages[::-1] %}
31
+ {%- set index = (messages|length - 1) - loop.index0 %}
32
+ {%- 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>')) %}
33
+ {%- set ns.multi_step_tool = false %}
34
+ {%- set ns.last_query_index = index %}
35
+ {%- endif %}
36
+ {%- endfor %}
37
+ {%- for message in messages %}
38
+ {%- if message.content is string %}
39
+ {%- set content = message.content %}
40
+ {%- else %}
41
+ {%- set content = namespace(text="") %}
42
+ {%- for mcontent in message.content %}
43
+ {%- if mcontent.type == 'image' or 'image' in mcontent or 'image_url' in mcontent %}
44
+ {%- set content.text = content.text~"<|vision_start|><|image_pad|><|vision_end|>" %}
45
+ {%- elif mcontent.type == 'audio' or 'audio' in mcontent or 'audio_url' in mcontent %}
46
+ {%- set content.text = content.text~"<|audio_start|><|audio_pad|><|audio_end|>" %}
47
+ {%- elif mcontent.type == 'video' or 'video' in mcontent %}
48
+ {%- set content.text = content.text~"<|vision_start|><|video_pad|><|vision_end|>" %}
49
+ {%- elif mcontent.type == 'text' %}
50
+ {%- set content.text = content.text~mcontent.text %}
51
+ {%- endif %}
52
+ {%- endfor %}
53
+ {%- set content = content.text %}
54
+ {%- endif %}
55
+ {%- if (message.role == "user") or (message.role == "system" and not loop.first) %}
56
+ {{- '<|im_start|>' + message.role + '\n' + content + '<|im_end|>' + '\n' }}
57
+ {%- elif message.role == "assistant" %}
58
+ {%- set reasoning_content = "" %}
59
+ {%- if message.reasoning_content is string %}
60
+ {%- set reasoning_content = message.reasoning_content %}
61
+ {%- else %}
62
+ {%- if '</think>' in content %}
63
+ {%- set reasoning_content = content.split('</think>')[0].rstrip('\n').split('<think>')[-1].lstrip('\n') %}
64
+ {%- set content = content.split('</think>')[-1].lstrip('\n') %}
65
+ {%- endif %}
66
+ {%- endif %}
67
+ {%- if loop.index0 > ns.last_query_index %}
68
+ {%- if loop.last or (not loop.last and reasoning_content) %}
69
+ {{- '<|im_start|>' + message.role + '\n<think>\n' + reasoning_content.strip("\n") + '\n</think>\n\n' + content.lstrip('\n') }}
70
+ {%- else %}
71
+ {{- '<|im_start|>' + message.role + '\n' + content }}
72
+ {%- endif %}
73
+ {%- else %}
74
+ {{- '<|im_start|>' + message.role + '\n' + content }}
75
+ {%- endif %}
76
+ {%- if message.tool_calls %}
77
+ {%- for tool_call in message.tool_calls %}
78
+ {%- if (loop.first and content) or (not loop.first) %}{{- '\n' }}{%- endif %}
79
+ {%- if tool_call.function %}
80
+ {%- set tool_call = tool_call.function %}
81
+ {%- endif %}
82
+ {{- '<tool_call>\n{"name": "' }}
83
+ {{- tool_call.name }}
84
+ {{- '", "arguments": ' }}
85
+ {%- if tool_call.arguments is string %}
86
+ {{- tool_call.arguments }}
87
+ {%- else %}
88
+ {{- tool_call.arguments | tojson }}
89
+ {%- endif %}
90
+ {{- '}\n</tool_call>' }}
91
+ {%- endfor %}
92
+ {%- endif %}
93
+ {{- '<|im_end|>\n' }}
94
+ {%- elif message.role == "tool" %}
95
+ {%- if loop.first or (messages[loop.index0 - 1].role != "tool") %}{{- '<|im_start|>user' }}{%- endif %}
96
+ {{- '\n<tool_response>\n' }}
97
+ {{- content }}
98
+ {{- '\n</tool_response>' }}
99
+ {%- if loop.last or (messages[loop.index0 + 1].role != "tool") %}{{- '<|im_end|>\n' }}{%- endif %}
100
+ {%- endif %}
101
+ {%- endfor %}
102
+ {%- if add_generation_prompt %}
103
+ {{- '<|im_start|>assistant\n' }}
104
+ {%- if enable_thinking is defined and enable_thinking is false %}{{- '<think>\n\n</think>\n\n' }}{%- endif %}
105
+ {%- endif %}
config.json ADDED
@@ -0,0 +1,527 @@
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
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_size": 2048,
10
+ "decoder_dim": 1536,
11
+ "hidden_act": "silu",
12
+ "hidden_size": 1024,
13
+ "intermediate_size": 3072,
14
+ "layer_scale_initial_scale": 0.01,
15
+ "max_position_embeddings": 8000,
16
+ "model_type": "",
17
+ "num_attention_heads": 16,
18
+ "num_hidden_layers": 8,
19
+ "num_key_value_heads": 16,
20
+ "num_quantizers": 16,
21
+ "rms_norm_eps": 1e-05,
22
+ "rope_theta": 10000,
23
+ "sliding_window": 72,
24
+ "upsample_rates": [
25
+ 8,
26
+ 5,
27
+ 4,
28
+ 3
29
+ ],
30
+ "upsampling_ratios": [
31
+ 2,
32
+ 2
33
+ ]
34
+ },
35
+ "dtype": "bfloat16",
36
+ "enable_audio_output": false,
37
+ "hidden_size": 1024,
38
+ "im_end_token_id": 151645,
39
+ "im_start_token_id": 151644,
40
+ "keys_to_ignore_at_inference": [
41
+ "past_key_values"
42
+ ],
43
+ "model_type": "qwen3_omni_moe",
44
+ "system_token_id": 8948,
45
+ "talker_config": {
46
+ "accept_hidden_layer": 18,
47
+ "audio_start_token_id": 151669,
48
+ "audio_token_id": 151646,
49
+ "code_predictor_config": {
50
+ "_name_or_path": "",
51
+ "add_cross_attention": false,
52
+ "architectures": null,
53
+ "attention_bias": false,
54
+ "attention_dropout": 0,
55
+ "bad_words_ids": null,
56
+ "begin_suppress_tokens": null,
57
+ "bos_token_id": null,
58
+ "chunk_size_feed_forward": 0,
59
+ "cross_attention_hidden_size": null,
60
+ "decoder_start_token_id": null,
61
+ "diversity_penalty": 0.0,
62
+ "do_sample": false,
63
+ "dtype": null,
64
+ "early_stopping": false,
65
+ "encoder_no_repeat_ngram_size": 0,
66
+ "eos_token_id": null,
67
+ "exponential_decay_length_penalty": null,
68
+ "finetuning_task": null,
69
+ "forced_bos_token_id": null,
70
+ "forced_eos_token_id": null,
71
+ "head_dim": 128,
72
+ "hidden_act": "silu",
73
+ "hidden_size": 1024,
74
+ "id2label": {
75
+ "0": "LABEL_0",
76
+ "1": "LABEL_1"
77
+ },
78
+ "initializer_range": 0.02,
79
+ "intermediate_size": 3072,
80
+ "is_decoder": false,
81
+ "is_encoder_decoder": false,
82
+ "label2id": {
83
+ "LABEL_0": 0,
84
+ "LABEL_1": 1
85
+ },
86
+ "layer_types": [
87
+ "full_attention",
88
+ "full_attention",
89
+ "full_attention",
90
+ "full_attention",
91
+ "full_attention"
92
+ ],
93
+ "length_penalty": 1.0,
94
+ "max_length": 20,
95
+ "max_position_embeddings": 32768,
96
+ "min_length": 0,
97
+ "model_type": "qwen3_omni_moe_talker_code_predictor",
98
+ "no_repeat_ngram_size": 0,
99
+ "num_attention_heads": 16,
100
+ "num_beam_groups": 1,
101
+ "num_beams": 1,
102
+ "num_code_groups": 32,
103
+ "num_hidden_layers": 5,
104
+ "num_key_value_heads": 8,
105
+ "num_return_sequences": 1,
106
+ "output_attentions": false,
107
+ "output_hidden_states": false,
108
+ "output_scores": false,
109
+ "pad_token_id": null,
110
+ "prefix": null,
111
+ "problem_type": null,
112
+ "pruned_heads": {},
113
+ "remove_invalid_values": false,
114
+ "repetition_penalty": 1.0,
115
+ "return_dict": true,
116
+ "return_dict_in_generate": false,
117
+ "rms_norm_eps": 1e-06,
118
+ "rope_scaling": null,
119
+ "rope_theta": 10000,
120
+ "sep_token_id": null,
121
+ "sliding_window": null,
122
+ "suppress_tokens": null,
123
+ "task_specific_params": null,
124
+ "temperature": 1.0,
125
+ "tf_legacy_loss": false,
126
+ "tie_encoder_decoder": false,
127
+ "tie_word_embeddings": false,
128
+ "tokenizer_class": null,
129
+ "top_k": 50,
130
+ "top_p": 1.0,
131
+ "torchscript": false,
132
+ "typical_p": 1.0,
133
+ "use_bfloat16": false,
134
+ "use_cache": true,
135
+ "vocab_size": 2048
136
+ },
137
+ "codec_bos_id": 4197,
138
+ "codec_eos_token_id": 4198,
139
+ "codec_nothink_id": 4203,
140
+ "codec_pad_id": 4196,
141
+ "codec_think_bos_id": 4204,
142
+ "codec_think_eos_id": 4205,
143
+ "image_token_id": 151655,
144
+ "model_type": "",
145
+ "num_code_groups": 32,
146
+ "position_id_per_seconds": 25,
147
+ "speaker_id": null,
148
+ "text_config": {
149
+ "_name_or_path": "",
150
+ "add_cross_attention": false,
151
+ "architectures": null,
152
+ "attention_bias": false,
153
+ "attention_dropout": 0,
154
+ "bad_words_ids": null,
155
+ "begin_suppress_tokens": null,
156
+ "bos_token_id": null,
157
+ "chunk_size_feed_forward": 0,
158
+ "cross_attention_hidden_size": null,
159
+ "decoder_sparse_step": 1,
160
+ "decoder_start_token_id": null,
161
+ "diversity_penalty": 0.0,
162
+ "do_sample": false,
163
+ "dtype": "bfloat16",
164
+ "early_stopping": false,
165
+ "encoder_no_repeat_ngram_size": 0,
166
+ "eos_token_id": null,
167
+ "exponential_decay_length_penalty": null,
168
+ "finetuning_task": null,
169
+ "forced_bos_token_id": null,
170
+ "forced_eos_token_id": null,
171
+ "hidden_act": "silu",
172
+ "hidden_size": 1024,
173
+ "id2label": {
174
+ "0": "LABEL_0",
175
+ "1": "LABEL_1"
176
+ },
177
+ "initializer_range": 0.02,
178
+ "intermediate_size": 2048,
179
+ "is_decoder": false,
180
+ "is_encoder_decoder": false,
181
+ "label2id": {
182
+ "LABEL_0": 0,
183
+ "LABEL_1": 1
184
+ },
185
+ "length_penalty": 1.0,
186
+ "max_length": 20,
187
+ "max_position_embeddings": 32768,
188
+ "min_length": 0,
189
+ "mlp_only_layers": [],
190
+ "model_type": "qwen3_omni_moe_talker_text",
191
+ "moe_intermediate_size": 384,
192
+ "no_repeat_ngram_size": 0,
193
+ "norm_topk_prob": false,
194
+ "num_attention_heads": 16,
195
+ "num_beam_groups": 1,
196
+ "num_beams": 1,
197
+ "num_experts": 128,
198
+ "num_experts_per_tok": 8,
199
+ "num_hidden_layers": 20,
200
+ "num_key_value_heads": 2,
201
+ "num_return_sequences": 1,
202
+ "output_attentions": false,
203
+ "output_hidden_states": false,
204
+ "output_router_logits": false,
205
+ "output_scores": false,
206
+ "pad_token_id": null,
207
+ "prefix": null,
208
+ "problem_type": null,
209
+ "pruned_heads": {},
210
+ "remove_invalid_values": false,
211
+ "repetition_penalty": 1.0,
212
+ "return_dict": true,
213
+ "return_dict_in_generate": false,
214
+ "rms_norm_eps": 1e-06,
215
+ "rope_scaling": null,
216
+ "rope_theta": 10000,
217
+ "router_aux_loss_coef": 0.001,
218
+ "sep_token_id": null,
219
+ "sliding_window": null,
220
+ "suppress_tokens": null,
221
+ "task_specific_params": null,
222
+ "temperature": 1.0,
223
+ "tf_legacy_loss": false,
224
+ "tie_encoder_decoder": false,
225
+ "tie_word_embeddings": false,
226
+ "tokenizer_class": null,
227
+ "top_k": 50,
228
+ "top_p": 1.0,
229
+ "torchscript": false,
230
+ "typical_p": 1.0,
231
+ "use_bfloat16": false,
232
+ "use_cache": true,
233
+ "vocab_size": 3072
234
+ },
235
+ "thinker_hidden_size": 2048,
236
+ "video_token_id": 151656,
237
+ "vision_start_token_id": 151652
238
+ },
239
+ "thinker_config": {
240
+ "audio_config": {
241
+ "_name_or_path": "",
242
+ "activation_dropout": 0,
243
+ "activation_function": "gelu",
244
+ "add_cross_attention": false,
245
+ "architectures": null,
246
+ "attention_dropout": 0,
247
+ "bad_words_ids": null,
248
+ "begin_suppress_tokens": null,
249
+ "bos_token_id": null,
250
+ "chunk_size_feed_forward": 0,
251
+ "conv_chunksize": 500,
252
+ "cross_attention_hidden_size": null,
253
+ "d_model": 1280,
254
+ "decoder_start_token_id": null,
255
+ "diversity_penalty": 0.0,
256
+ "do_sample": false,
257
+ "downsample_hidden_size": 480,
258
+ "dropout": 0,
259
+ "dtype": "bfloat16",
260
+ "early_stopping": false,
261
+ "encoder_attention_heads": 20,
262
+ "encoder_ffn_dim": 5120,
263
+ "encoder_layers": 32,
264
+ "encoder_no_repeat_ngram_size": 0,
265
+ "eos_token_id": null,
266
+ "exponential_decay_length_penalty": null,
267
+ "finetuning_task": null,
268
+ "forced_bos_token_id": null,
269
+ "forced_eos_token_id": null,
270
+ "id2label": {
271
+ "0": "LABEL_0",
272
+ "1": "LABEL_1"
273
+ },
274
+ "initializer_range": 0.02,
275
+ "is_decoder": false,
276
+ "is_encoder_decoder": false,
277
+ "label2id": {
278
+ "LABEL_0": 0,
279
+ "LABEL_1": 1
280
+ },
281
+ "length_penalty": 1.0,
282
+ "max_length": 20,
283
+ "max_source_positions": 1500,
284
+ "min_length": 0,
285
+ "model_type": "qwen3_omni_moe_audio_encoder",
286
+ "n_window": 50,
287
+ "n_window_infer": 800,
288
+ "no_repeat_ngram_size": 0,
289
+ "num_beam_groups": 1,
290
+ "num_beams": 1,
291
+ "num_hidden_layers": 32,
292
+ "num_mel_bins": 128,
293
+ "num_return_sequences": 1,
294
+ "output_attentions": false,
295
+ "output_dim": 2048,
296
+ "output_hidden_states": false,
297
+ "output_scores": false,
298
+ "pad_token_id": null,
299
+ "prefix": null,
300
+ "problem_type": null,
301
+ "pruned_heads": {},
302
+ "remove_invalid_values": false,
303
+ "repetition_penalty": 1.0,
304
+ "return_dict": true,
305
+ "return_dict_in_generate": false,
306
+ "scale_embedding": false,
307
+ "sep_token_id": null,
308
+ "suppress_tokens": null,
309
+ "task_specific_params": null,
310
+ "temperature": 1.0,
311
+ "tf_legacy_loss": false,
312
+ "tie_encoder_decoder": false,
313
+ "tie_word_embeddings": true,
314
+ "tokenizer_class": null,
315
+ "top_k": 50,
316
+ "top_p": 1.0,
317
+ "torchscript": false,
318
+ "typical_p": 1.0,
319
+ "use_bfloat16": false
320
+ },
321
+ "audio_end_token_id": 151670,
322
+ "audio_start_token_id": 151669,
323
+ "audio_token_id": 151675,
324
+ "dtype": "bfloat16",
325
+ "image_token_id": 151655,
326
+ "initializer_range": 0.02,
327
+ "model_type": "qwen3_omni_moe_thinker",
328
+ "position_id_per_seconds": 13,
329
+ "seconds_per_chunk": 2,
330
+ "text_config": {
331
+ "_name_or_path": "",
332
+ "add_cross_attention": false,
333
+ "architectures": null,
334
+ "attention_bias": false,
335
+ "attention_dropout": 0.0,
336
+ "bad_words_ids": null,
337
+ "begin_suppress_tokens": null,
338
+ "bos_token_id": null,
339
+ "chunk_size_feed_forward": 0,
340
+ "cross_attention_hidden_size": null,
341
+ "decoder_sparse_step": 1,
342
+ "decoder_start_token_id": null,
343
+ "diversity_penalty": 0.0,
344
+ "do_sample": false,
345
+ "dtype": "bfloat16",
346
+ "early_stopping": false,
347
+ "encoder_no_repeat_ngram_size": 0,
348
+ "eos_token_id": null,
349
+ "exponential_decay_length_penalty": null,
350
+ "finetuning_task": null,
351
+ "forced_bos_token_id": null,
352
+ "forced_eos_token_id": null,
353
+ "head_dim": 128,
354
+ "hidden_act": "silu",
355
+ "hidden_size": 2048,
356
+ "id2label": {
357
+ "0": "LABEL_0",
358
+ "1": "LABEL_1"
359
+ },
360
+ "initializer_range": 0.02,
361
+ "intermediate_size": 768,
362
+ "is_decoder": false,
363
+ "is_encoder_decoder": false,
364
+ "label2id": {
365
+ "LABEL_0": 0,
366
+ "LABEL_1": 1
367
+ },
368
+ "length_penalty": 1.0,
369
+ "max_length": 20,
370
+ "max_position_embeddings": 65536,
371
+ "min_length": 0,
372
+ "mlp_only_layers": [],
373
+ "model_type": "qwen3_omni_moe_text",
374
+ "moe_intermediate_size": 768,
375
+ "no_repeat_ngram_size": 0,
376
+ "norm_topk_prob": true,
377
+ "num_attention_heads": 32,
378
+ "num_beam_groups": 1,
379
+ "num_beams": 1,
380
+ "num_experts": 128,
381
+ "num_experts_per_tok": 8,
382
+ "num_hidden_layers": 48,
383
+ "num_key_value_heads": 4,
384
+ "num_return_sequences": 1,
385
+ "output_attentions": false,
386
+ "output_hidden_states": false,
387
+ "output_router_logits": false,
388
+ "output_scores": false,
389
+ "pad_token_id": null,
390
+ "prefix": null,
391
+ "problem_type": null,
392
+ "pruned_heads": {},
393
+ "remove_invalid_values": false,
394
+ "repetition_penalty": 1.0,
395
+ "return_dict": true,
396
+ "return_dict_in_generate": false,
397
+ "rms_norm_eps": 1e-06,
398
+ "rope_scaling": {
399
+ "interleaved": true,
400
+ "mrope_interleaved": true,
401
+ "mrope_section": [
402
+ 24,
403
+ 20,
404
+ 20
405
+ ],
406
+ "rope_type": "default",
407
+ "type": "default"
408
+ },
409
+ "rope_theta": 1000000,
410
+ "router_aux_loss_coef": 0.001,
411
+ "sep_token_id": null,
412
+ "shared_expert_intermediate_size": 0,
413
+ "sliding_window": null,
414
+ "suppress_tokens": null,
415
+ "task_specific_params": null,
416
+ "temperature": 1.0,
417
+ "tf_legacy_loss": false,
418
+ "tie_encoder_decoder": false,
419
+ "tie_word_embeddings": false,
420
+ "tokenizer_class": null,
421
+ "top_k": 50,
422
+ "top_p": 1.0,
423
+ "torchscript": false,
424
+ "typical_p": 1.0,
425
+ "use_bfloat16": false,
426
+ "use_cache": true,
427
+ "use_qk_norm": true,
428
+ "use_sliding_window": false,
429
+ "vocab_size": 152064
430
+ },
431
+ "user_token_id": 872,
432
+ "video_token_id": 151656,
433
+ "vision_config": {
434
+ "_name_or_path": "",
435
+ "add_cross_attention": false,
436
+ "apply_vit_abs_pos_embed": true,
437
+ "architectures": null,
438
+ "bad_words_ids": null,
439
+ "begin_suppress_tokens": null,
440
+ "bos_token_id": null,
441
+ "chunk_size_feed_forward": 0,
442
+ "cross_attention_hidden_size": null,
443
+ "decoder_start_token_id": null,
444
+ "deepstack_visual_indexes": [
445
+ 8,
446
+ 16,
447
+ 24
448
+ ],
449
+ "depth": 27,
450
+ "diversity_penalty": 0.0,
451
+ "do_sample": false,
452
+ "dtype": "bfloat16",
453
+ "early_stopping": false,
454
+ "encoder_no_repeat_ngram_size": 0,
455
+ "eos_token_id": null,
456
+ "exponential_decay_length_penalty": null,
457
+ "finetuning_task": null,
458
+ "forced_bos_token_id": null,
459
+ "forced_eos_token_id": null,
460
+ "hidden_act": "gelu_pytorch_tanh",
461
+ "hidden_size": 1152,
462
+ "id2label": {
463
+ "0": "LABEL_0",
464
+ "1": "LABEL_1"
465
+ },
466
+ "image_size": 768,
467
+ "in_channels": 3,
468
+ "in_chans": 3,
469
+ "initializer_range": 0.02,
470
+ "intermediate_size": 4304,
471
+ "is_decoder": false,
472
+ "is_encoder_decoder": false,
473
+ "label2id": {
474
+ "LABEL_0": 0,
475
+ "LABEL_1": 1
476
+ },
477
+ "length_penalty": 1.0,
478
+ "max_length": 20,
479
+ "min_length": 0,
480
+ "model_type": "qwen3_omni_moe_vision_encoder",
481
+ "no_repeat_ngram_size": 0,
482
+ "num_beam_groups": 1,
483
+ "num_beams": 1,
484
+ "num_heads": 16,
485
+ "num_position_embeddings": 2304,
486
+ "num_return_sequences": 1,
487
+ "out_hidden_size": 2048,
488
+ "output_attentions": false,
489
+ "output_hidden_states": false,
490
+ "output_scores": false,
491
+ "pad_token_id": null,
492
+ "patch_size": 16,
493
+ "prefix": null,
494
+ "problem_type": null,
495
+ "pruned_heads": {},
496
+ "remove_invalid_values": false,
497
+ "repetition_penalty": 1.0,
498
+ "return_dict": true,
499
+ "return_dict_in_generate": false,
500
+ "sep_token_id": null,
501
+ "spatial_merge_size": 2,
502
+ "spatial_patch_size": 16,
503
+ "suppress_tokens": null,
504
+ "task_specific_params": null,
505
+ "temperature": 1.0,
506
+ "temporal_patch_size": 2,
507
+ "tf_legacy_loss": false,
508
+ "tie_encoder_decoder": false,
509
+ "tie_word_embeddings": true,
510
+ "tokenizer_class": null,
511
+ "tokens_per_second": 2,
512
+ "top_k": 50,
513
+ "top_p": 1.0,
514
+ "torchscript": false,
515
+ "typical_p": 1.0,
516
+ "use_bfloat16": false
517
+ },
518
+ "vision_end_token_id": 151653,
519
+ "vision_start_token_id": 151652
520
+ },
521
+ "transformers_version": "4.57.6",
522
+ "tts_bos_token_id": 151672,
523
+ "tts_eos_token_id": 151673,
524
+ "tts_pad_token_id": 151671,
525
+ "user_token_id": 872,
526
+ "vocab_size": 152064
527
+ }
iter_0000500/.metadata ADDED
Binary file (481 Bytes). View file
 
iter_0000500/common.pt ADDED
@@ -0,0 +1,3 @@
 
 
 
 
1
+ version https://git-lfs.github.com/spec/v1
2
+ oid sha256:9f91eeaec7e224881f874ae6453705a1ec9e4565b530f45dc583cf4f3b747478
3
+ size 29799
iter_0000500/metadata.json ADDED
@@ -0,0 +1 @@
 
 
1
+ {"sharded_backend": "torch_dist", "sharded_backend_version": 1, "common_backend": "torch", "common_backend_version": 1}
latest_checkpointed_iteration.txt ADDED
@@ -0,0 +1 @@
 
 
1
+ 500
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:e29c99cc6727c0cd4ece09993ac7f094c1a8da902acfce0f2772d84cdefeb18b
3
+ size 4997903800
model-00002-of-00013.safetensors ADDED
@@ -0,0 +1,3 @@
 
 
 
 
1
+ version https://git-lfs.github.com/spec/v1
2
+ oid sha256:2e5ce6122f3f133aa1f6b3b5877edddfae17df1f4ce57c8f423b9dd41aa6670a
3
+ size 4997754192
model-00003-of-00013.safetensors ADDED
@@ -0,0 +1,3 @@
 
 
 
 
1
+ version https://git-lfs.github.com/spec/v1
2
+ oid sha256:97bfe08bfa52373ba84a0b53e809b2edf490ef9c1efb87f3df7b7e18f7e143d0
3
+ size 4997754192
model-00004-of-00013.safetensors ADDED
@@ -0,0 +1,3 @@
 
 
 
 
1
+ version https://git-lfs.github.com/spec/v1
2
+ oid sha256:56da8f23f118d6de71adc90e72248e6f1129556df8e278ef41f6333b9d3f1abb
3
+ size 4997755616
model-00005-of-00013.safetensors ADDED
@@ -0,0 +1,3 @@
 
 
 
 
1
+ version https://git-lfs.github.com/spec/v1
2
+ oid sha256:925bc0a976154b23412928bd92b28f987edfcd5a170a7cdd54da21983d224356
3
+ size 4997755768
model-00006-of-00013.safetensors ADDED
@@ -0,0 +1,3 @@
 
 
 
 
1
+ version https://git-lfs.github.com/spec/v1
2
+ oid sha256:76cea60da3d567312d3e581dbcd2172c5e782e8be711fcb400d38ee2ba8d0a5e
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:bef1caefd112503fed7b781013cab996cc2563b9f363d21ac833f5629813d98c
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:e4372a5c9584c00ebffe5d3be2c4a2ab08522a701b27b38610a26e34a18d1077
3
+ size 4997755768
model-00009-of-00013.safetensors ADDED
@@ -0,0 +1,3 @@
 
 
 
 
1
+ version https://git-lfs.github.com/spec/v1
2
+ oid sha256:56b5df236f30ac3ea369e6fc24174784d6a4f581e8f195842a2399d9b20c0a51
3
+ size 4997755776
model-00010-of-00013.safetensors ADDED
@@ -0,0 +1,3 @@
 
 
 
 
1
+ version https://git-lfs.github.com/spec/v1
2
+ oid sha256:7fb4a5f64c9f8ea861358b290c28e79b5838c6eab95107c186b2cd75d265765d
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:eea0401ca39050aa6ba81f7b6213b288fd753d57ddd95cf245a4aa00c0e0a095
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:ffbead75686dea9d66a5066ec51b4359f1810cf5b4d19518a87768de2069edca
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:fff5146e415ff8e1a5f05ef1176182f0c952dfc25eab4ecd4eb54db6758f24c7
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