cucl2/AnyAudio-Judge-30B
0152
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 "dtype": "bfloat16",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 "rms_norm_eps": 1e-05,23 "rope_theta": 10000,24 "sliding_window": 72,25 "upsample_rates": [26 8,27 5,28 4,29 330 ],31 "upsampling_ratios": [32 2,33 234 ]35 },36 "dtype": "bfloat16",37 "enable_audio_output": false,38 "eos_token_id": 151645,39 "hidden_size": 1024,40 "im_end_token_id": 151645,41 "im_start_token_id": 151644,42 "keys_to_ignore_at_inference": [43 "past_key_values",44 "hidden_states",45 "attention_mask"46 ],47 "model_type": "qwen3_omni_moe",48 "pad_token_id": 151643,49 "system_token_id": 8948,50 "talker_config": {51 "accept_hidden_layer": 18,52 "audio_start_token_id": 151669,53 "audio_token_id": 151646,54 "code_predictor_config": {55 "_name_or_path": "",56 "add_cross_attention": false,57 "architectures": null,58 "attention_bias": false,59 "attention_dropout": 0,60 "bad_words_ids": null,61 "begin_suppress_tokens": null,62 "bos_token_id": null,63 "chunk_size_feed_forward": 0,64 "cross_attention_hidden_size": null,65 "decoder_start_token_id": null,66 "diversity_penalty": 0.0,67 "do_sample": false,68 "dtype": null,69 "early_stopping": false,70 "encoder_no_repeat_ngram_size": 0,71 "eos_token_id": null,72 "exponential_decay_length_penalty": null,73 "finetuning_task": null,74 "forced_bos_token_id": null,75 "forced_eos_token_id": null,76 "head_dim": 128,77 "hidden_act": "silu",78 "hidden_size": 1024,79 "id2label": {80 "0": "LABEL_0",81 "1": "LABEL_1"82 },83 "initializer_range": 0.02,84 "intermediate_size": 3072,85 "is_decoder": false,86 "is_encoder_decoder": false,87 "label2id": {88 "LABEL_0": 0,89 "LABEL_1": 190 },91 "layer_types": [92 "full_attention",93 "full_attention",94 "full_attention",95 "full_attention",96 "full_attention"97 ],98 "length_penalty": 1.0,99 "max_length": 20,100 "max_position_embeddings": 32768,101 "min_length": 0,102 "model_type": "qwen3_omni_moe_talker_code_predictor",103 "no_repeat_ngram_size": 0,104 "num_attention_heads": 16,105 "num_beam_groups": 1,106 "num_beams": 1,107 "num_code_groups": 32,108 "num_hidden_layers": 5,109 "num_key_value_heads": 8,110 "num_return_sequences": 1,111 "output_attentions": false,112 "output_hidden_states": false,113 "output_scores": false,114 "pad_token_id": null,115 "prefix": null,116 "problem_type": null,117 "pruned_heads": {},118 "remove_invalid_values": false,119 "repetition_penalty": 1.0,120 "return_dict": true,121 "return_dict_in_generate": false,122 "rms_norm_eps": 1e-06,123 "rope_scaling": null,124 "rope_theta": 10000,125 "sep_token_id": null,126 "sliding_window": null,127 "suppress_tokens": null,128 "task_specific_params": null,129 "temperature": 1.0,130 "tie_encoder_decoder": false,131 "tie_word_embeddings": false,132 "tokenizer_class": null,133 "top_k": 50,134 "top_p": 1.0,135 "torchscript": false,136 "typical_p": 1.0,137 "use_cache": true,138 "vocab_size": 2048139 },140 "codec_bos_id": 4197,141 "codec_eos_token_id": 4198,142 "codec_nothink_id": 4203,143 "codec_pad_id": 4196,144 "codec_think_bos_id": 4204,145 "codec_think_eos_id": 4205,146 "dtype": "bfloat16",147 "image_token_id": 151655,148 "model_type": "",149 "num_code_groups": 32,150 "position_id_per_seconds": 25,151 "speaker_id": null,152 "text_config": {153 "_name_or_path": "",154 "add_cross_attention": false,155 "architectures": null,156 "attention_bias": false,157 "attention_dropout": 0,158 "bad_words_ids": null,159 "begin_suppress_tokens": null,160 "bos_token_id": null,161 "chunk_size_feed_forward": 0,162 "cross_attention_hidden_size": null,163 "decoder_sparse_step": 1,164 "decoder_start_token_id": null,165 "diversity_penalty": 0.0,166 "do_sample": false,167 "dtype": "bfloat16",168 "early_stopping": false,169 "encoder_no_repeat_ngram_size": 0,170 "eos_token_id": null,171 "exponential_decay_length_penalty": null,172 "finetuning_task": null,173 "forced_bos_token_id": null,174 "forced_eos_token_id": null,175 "hidden_act": "silu",176 "hidden_size": 1024,177 "id2label": {178 "0": "LABEL_0",179 "1": "LABEL_1"180 },181 "initializer_range": 0.02,182 "intermediate_size": 2048,183 "is_decoder": false,184 "is_encoder_decoder": false,185 "label2id": {186 "LABEL_0": 0,187 "LABEL_1": 1188 },189 "length_penalty": 1.0,190 "max_length": 20,191 "max_position_embeddings": 32768,192 "min_length": 0,193 "mlp_only_layers": [],194 "model_type": "qwen3_omni_moe_talker_text",195 "moe_intermediate_size": 384,196 "no_repeat_ngram_size": 0,197 "norm_topk_prob": false,198 "num_attention_heads": 16,199 "num_beam_groups": 1,200 "num_beams": 1,201 "num_experts": 128,202 "num_experts_per_tok": 8,203 "num_hidden_layers": 20,204 "num_key_value_heads": 2,205 "num_return_sequences": 1,206 "output_attentions": false,207 "output_hidden_states": false,208 "output_router_logits": false,209 "output_scores": false,210 "pad_token_id": 151643,211 "prefix": null,212 "problem_type": null,213 "pruned_heads": {},214 "remove_invalid_values": false,215 "repetition_penalty": 1.0,216 "return_dict": true,217 "return_dict_in_generate": false,218 "rms_norm_eps": 1e-06,219 "rope_scaling": null,220 "rope_theta": 10000,221 "router_aux_loss_coef": 0.0,222 "sep_token_id": null,223 "sliding_window": null,224 "suppress_tokens": null,225 "task_specific_params": null,226 "temperature": 1.0,227 "tie_encoder_decoder": false,228 "tie_word_embeddings": false,229 "tokenizer_class": null,230 "top_k": 50,231 "top_p": 1.0,232 "torchscript": false,233 "typical_p": 1.0,234 "use_cache": false,235 "vocab_size": 3072236 },237 "thinker_hidden_size": 2048,238 "video_token_id": 151656,239 "vision_start_token_id": 151652240 },241 "thinker_config": {242 "audio_config": {243 "_name_or_path": "",244 "activation_dropout": 0,245 "activation_function": "gelu",246 "add_cross_attention": false,247 "architectures": null,248 "attention_dropout": 0,249 "bad_words_ids": null,250 "begin_suppress_tokens": null,251 "bos_token_id": null,252 "chunk_size_feed_forward": 0,253 "conv_chunksize": 500,254 "cross_attention_hidden_size": null,255 "d_model": 1280,256 "decoder_start_token_id": null,257 "diversity_penalty": 0.0,258 "do_sample": false,259 "downsample_hidden_size": 480,260 "dropout": 0,261 "dtype": "bfloat16",262 "early_stopping": false,263 "encoder_attention_heads": 20,264 "encoder_ffn_dim": 5120,265 "encoder_layers": 32,266 "encoder_no_repeat_ngram_size": 0,267 "eos_token_id": null,268 "exponential_decay_length_penalty": null,269 "finetuning_task": null,270 "forced_bos_token_id": null,271 "forced_eos_token_id": null,272 "id2label": {273 "0": "LABEL_0",274 "1": "LABEL_1"275 },276 "initializer_range": 0.02,277 "is_decoder": false,278 "is_encoder_decoder": false,279 "label2id": {280 "LABEL_0": 0,281 "LABEL_1": 1282 },283 "length_penalty": 1.0,284 "max_length": 20,285 "max_source_positions": 1500,286 "min_length": 0,287 "model_type": "qwen3_omni_moe_audio_encoder",288 "n_window": 50,289 "n_window_infer": 800,290 "no_repeat_ngram_size": 0,291 "num_beam_groups": 1,292 "num_beams": 1,293 "num_hidden_layers": 32,294 "num_mel_bins": 128,295 "num_return_sequences": 1,296 "output_attentions": false,297 "output_dim": 2048,298 "output_hidden_states": false,299 "output_scores": false,300 "pad_token_id": null,301 "prefix": null,302 "problem_type": null,303 "pruned_heads": {},304 "remove_invalid_values": false,305 "repetition_penalty": 1.0,306 "return_dict": true,307 "return_dict_in_generate": false,308 "scale_embedding": false,309 "sep_token_id": null,310 "suppress_tokens": null,311 "task_specific_params": null,312 "temperature": 1.0,313 "tf_legacy_loss": false,314 "tie_encoder_decoder": false,315 "tie_word_embeddings": true,316 "tokenizer_class": null,317 "top_k": 50,318 "top_p": 1.0,319 "torchscript": false,320 "typical_p": 1.0,321 "use_bfloat16": false322 },323 "audio_end_token_id": 151670,324 "audio_start_token_id": 151669,325 "dtype": "bfloat16",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": 1367 },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": 151643,390 "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 20405 ],406 "rope_type": "default",407 "type": "default"408 },409 "rope_theta": 1000000,410 "router_aux_loss_coef": 0.0,411 "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": false,427 "use_qk_norm": true,428 "use_sliding_window": false,429 "vocab_size": 152064430 },431 "user_token_id": 872,432 "vision_config": {433 "_name_or_path": "",434 "add_cross_attention": false,435 "apply_vit_abs_pos_embed": true,436 "architectures": null,437 "bad_words_ids": null,438 "begin_suppress_tokens": null,439 "bos_token_id": null,440 "chunk_size_feed_forward": 0,441 "cross_attention_hidden_size": null,442 "decoder_start_token_id": null,443 "deepstack_visual_indexes": [444 8,445 16,446 24447 ],448 "depth": 27,449 "diversity_penalty": 0.0,450 "do_sample": false,451 "dtype": "bfloat16",452 "early_stopping": false,453 "encoder_no_repeat_ngram_size": 0,454 "eos_token_id": null,455 "exponential_decay_length_penalty": null,456 "finetuning_task": null,457 "forced_bos_token_id": null,458 "forced_eos_token_id": null,459 "hidden_act": "gelu_pytorch_tanh",460 "hidden_size": 1152,461 "id2label": {462 "0": "LABEL_0",463 "1": "LABEL_1"464 },465 "image_size": 768,466 "in_channels": 3,467 "in_chans": 3,468 "initializer_range": 0.02,469 "intermediate_size": 4304,470 "is_decoder": false,471 "is_encoder_decoder": false,472 "label2id": {473 "LABEL_0": 0,474 "LABEL_1": 1475 },476 "length_penalty": 1.0,477 "max_length": 20,478 "min_length": 0,479 "model_type": "qwen3_omni_moe_vision_encoder",480 "no_repeat_ngram_size": 0,481 "num_beam_groups": 1,482 "num_beams": 1,483 "num_heads": 16,484 "num_position_embeddings": 2304,485 "num_return_sequences": 1,486 "out_hidden_size": 2048,487 "output_attentions": false,488 "output_hidden_states": false,489 "output_scores": false,490 "pad_token_id": null,491 "patch_size": 16,492 "prefix": null,493 "problem_type": null,494 "pruned_heads": {},495 "remove_invalid_values": false,496 "repetition_penalty": 1.0,497 "return_dict": true,498 "return_dict_in_generate": false,499 "sep_token_id": null,500 "spatial_merge_size": 2,501 "spatial_patch_size": 16,502 "suppress_tokens": null,503 "task_specific_params": null,504 "temperature": 1.0,505 "temporal_patch_size": 2,506 "tf_legacy_loss": false,507 "tie_encoder_decoder": false,508 "tie_word_embeddings": true,509 "tokenizer_class": null,510 "tokens_per_second": 2,511 "top_k": 50,512 "top_p": 1.0,513 "torchscript": false,514 "typical_p": 1.0,515 "use_bfloat16": false516 },517 "vision_end_token_id": 151653,518 "vision_start_token_id": 151652,519 "audio_token_id": 151675,520 "image_token_id": 151655,521 "video_token_id": 151656522 },523 "transformers_version": "4.57.0.dev0",524 "tts_bos_token_id": 151672,525 "tts_eos_token_id": 151673,526 "tts_pad_token_id": 151671,527 "user_token_id": 872528}