iNLP-Lab/Myna-Hokkien
1473
1{2 "architectures": [3 "Qwen3OmniMoeForConditionalGeneration"4 ],5 "assistant_token_id": 77091,6 "code2wav_config": {7 "attention_bias": false,8 "attention_dropout": 0.0,9 "codebook_dim": 512,10 "codebook_size": 2048,11 "decoder_dim": 1536,12 "dtype": "bfloat16",13 "hidden_act": "silu",14 "hidden_size": 1024,15 "intermediate_size": 3072,16 "layer_scale_initial_scale": 0.01,17 "max_position_embeddings": 8000,18 "model_type": "",19 "num_attention_heads": 16,20 "num_hidden_layers": 8,21 "num_key_value_heads": 16,22 "num_quantizers": 16,23 "num_semantic_quantizers": 1,24 "rms_norm_eps": 1e-05,25 "rope_theta": 10000,26 "semantic_codebook_size": 4096,27 "sliding_window": 72,28 "upsample_rates": [29 8,30 5,31 4,32 333 ],34 "upsampling_ratios": [35 2,36 237 ],38 "vector_quantization_hidden_dimension": 51239 },40 "dtype": "bfloat16",41 "enable_audio_output": true,42 "im_end_token_id": 151645,43 "im_start_token_id": 151644,44 "model_type": "qwen3_omni_moe",45 "system_token_id": 8948,46 "talker_config": {47 "accept_hidden_layer": 24,48 "audio_end_token_id": 151670,49 "audio_start_token_id": 151669,50 "audio_token_id": 151675,51 "code_predictor_config": {52 "_name_or_path": "",53 "add_cross_attention": false,54 "architectures": null,55 "attention_bias": false,56 "attention_dropout": 0,57 "bad_words_ids": null,58 "begin_suppress_tokens": null,59 "bos_token_id": null,60 "chunk_size_feed_forward": 0,61 "cross_attention_hidden_size": null,62 "decoder_start_token_id": null,63 "diversity_penalty": 0.0,64 "do_sample": false,65 "dtype": null,66 "early_stopping": false,67 "encoder_no_repeat_ngram_size": 0,68 "eos_token_id": null,69 "exponential_decay_length_penalty": null,70 "finetuning_task": null,71 "forced_bos_token_id": null,72 "forced_eos_token_id": null,73 "head_dim": 128,74 "hidden_act": "silu",75 "hidden_size": 1024,76 "id2label": {77 "0": "LABEL_0",78 "1": "LABEL_1"79 },80 "initializer_range": 0.02,81 "intermediate_size": 3072,82 "is_decoder": false,83 "is_encoder_decoder": false,84 "label2id": {85 "LABEL_0": 0,86 "LABEL_1": 187 },88 "layer_types": [89 "full_attention",90 "full_attention",91 "full_attention",92 "full_attention",93 "full_attention"94 ],95 "length_penalty": 1.0,96 "max_length": 20,97 "max_position_embeddings": 32768,98 "max_window_layers": 28,99 "min_length": 0,100 "model_type": "qwen3_omni_moe_talker_code_predictor",101 "no_repeat_ngram_size": 0,102 "num_attention_heads": 16,103 "num_beam_groups": 1,104 "num_beams": 1,105 "num_code_groups": 16,106 "num_hidden_layers": 5,107 "num_key_value_heads": 8,108 "num_return_sequences": 1,109 "output_attentions": false,110 "output_hidden_states": false,111 "output_scores": false,112 "pad_token_id": null,113 "prefix": null,114 "problem_type": null,115 "pruned_heads": {},116 "remove_invalid_values": false,117 "repetition_penalty": 1.0,118 "return_dict": true,119 "return_dict_in_generate": false,120 "rms_norm_eps": 1e-06,121 "rope_scaling": null,122 "rope_theta": 1000000,123 "sep_token_id": null,124 "sliding_window": null,125 "suppress_tokens": null,126 "task_specific_params": null,127 "temperature": 1.0,128 "tf_legacy_loss": false,129 "tie_encoder_decoder": false,130 "tie_word_embeddings": false,131 "tokenizer_class": null,132 "top_k": 50,133 "top_p": 1.0,134 "torchscript": false,135 "typical_p": 1.0,136 "use_bfloat16": false,137 "use_cache": true,138 "use_sliding_window": false,139 "vocab_size": 2048140 },141 "codec_bos_id": 2149,142 "codec_eos_token_id": 2150,143 "codec_nothink_id": 2155,144 "codec_pad_id": 2148,145 "codec_think_bos_id": 2156,146 "codec_think_eos_id": 2157,147 "dtype": "bfloat16",148 "image_token_id": 151655,149 "model_type": "",150 "num_code_groups": 16,151 "output_router_logits": false,152 "position_id_per_seconds": 13,153 "seconds_per_chunk": 2,154 "spatial_merge_size": 2,155 "speaker_id": {156 "aiden": 2303,157 "chelsie": 2301,158 "ethan": 2302159 },160 "text_config": {161 "_name_or_path": "",162 "add_cross_attention": false,163 "architectures": null,164 "attention_bias": false,165 "attention_dropout": 0,166 "bad_words_ids": null,167 "begin_suppress_tokens": null,168 "bos_token_id": null,169 "chunk_size_feed_forward": 0,170 "cross_attention_hidden_size": null,171 "decoder_sparse_step": 1,172 "decoder_start_token_id": null,173 "diversity_penalty": 0.0,174 "do_sample": false,175 "dtype": null,176 "early_stopping": false,177 "encoder_no_repeat_ngram_size": 0,178 "eos_token_id": null,179 "exponential_decay_length_penalty": null,180 "finetuning_task": null,181 "forced_bos_token_id": null,182 "forced_eos_token_id": null,183 "head_dim": 128,184 "hidden_act": "silu",185 "hidden_size": 1024,186 "id2label": {187 "0": "LABEL_0",188 "1": "LABEL_1"189 },190 "initializer_range": 0.02,191 "intermediate_size": 2048,192 "is_decoder": false,193 "is_encoder_decoder": false,194 "label2id": {195 "LABEL_0": 0,196 "LABEL_1": 1197 },198 "length_penalty": 1.0,199 "max_length": 20,200 "max_position_embeddings": 65536,201 "min_length": 0,202 "mlp_only_layers": [],203 "model_type": "qwen3_omni_moe_talker_text",204 "moe_intermediate_size": 384,205 "no_repeat_ngram_size": 0,206 "norm_topk_prob": true,207 "num_attention_heads": 16,208 "num_beam_groups": 1,209 "num_beams": 1,210 "num_experts": 128,211 "num_experts_per_tok": 6,212 "num_hidden_layers": 20,213 "num_key_value_heads": 2,214 "num_return_sequences": 1,215 "output_attentions": false,216 "output_hidden_states": false,217 "output_router_logits": false,218 "output_scores": false,219 "pad_token_id": null,220 "prefix": null,221 "problem_type": null,222 "pruned_heads": {},223 "remove_invalid_values": false,224 "repetition_penalty": 1.0,225 "return_dict": true,226 "return_dict_in_generate": false,227 "rms_norm_eps": 1e-06,228 "rope_scaling": {229 "interleaved": true,230 "mrope_section": [231 24,232 20,233 20234 ],235 "rope_type": "default",236 "type": "default"237 },238 "rope_theta": 1000000,239 "router_aux_loss_coef": 0.001,240 "sep_token_id": null,241 "shared_expert_intermediate_size": 768,242 "sliding_window": null,243 "suppress_tokens": null,244 "task_specific_params": null,245 "temperature": 1.0,246 "tf_legacy_loss": false,247 "tie_encoder_decoder": false,248 "tie_word_embeddings": false,249 "tokenizer_class": null,250 "top_k": 50,251 "top_p": 1.0,252 "torchscript": false,253 "typical_p": 1.0,254 "use_bfloat16": false,255 "use_cache": true,256 "use_sliding_window": false,257 "vocab_size": 3072258 },259 "thinker_hidden_size": 2048,260 "video_token_id": 151656,261 "vision_start_token_id": 151652262 },263 "thinker_config": {264 "audio_config": {265 "_name_or_path": "",266 "activation_dropout": 0,267 "activation_function": "gelu",268 "add_cross_attention": false,269 "architectures": null,270 "attention_dropout": 0,271 "bad_words_ids": null,272 "begin_suppress_tokens": null,273 "bos_token_id": null,274 "chunk_size_feed_forward": 0,275 "conv_chunksize": 500,276 "cross_attention_hidden_size": null,277 "d_model": 1280,278 "decoder_start_token_id": null,279 "diversity_penalty": 0.0,280 "do_sample": false,281 "downsample_hidden_size": 480,282 "dropout": 0,283 "dtype": null,284 "early_stopping": false,285 "encoder_attention_heads": 20,286 "encoder_ffn_dim": 5120,287 "encoder_layers": 32,288 "encoder_no_repeat_ngram_size": 0,289 "eos_token_id": null,290 "exponential_decay_length_penalty": null,291 "finetuning_task": null,292 "forced_bos_token_id": null,293 "forced_eos_token_id": null,294 "id2label": {295 "0": "LABEL_0",296 "1": "LABEL_1"297 },298 "initializer_range": 0.02,299 "is_decoder": false,300 "is_encoder_decoder": false,301 "label2id": {302 "LABEL_0": 0,303 "LABEL_1": 1304 },305 "length_penalty": 1.0,306 "max_length": 20,307 "max_source_positions": 1500,308 "min_length": 0,309 "model_type": "qwen3_omni_moe_audio_encoder",310 "n_window": 50,311 "n_window_infer": 800,312 "no_repeat_ngram_size": 0,313 "num_beam_groups": 1,314 "num_beams": 1,315 "num_hidden_layers": 32,316 "num_mel_bins": 128,317 "num_return_sequences": 1,318 "output_attentions": false,319 "output_dim": 2048,320 "output_hidden_states": false,321 "output_scores": false,322 "pad_token_id": null,323 "prefix": null,324 "problem_type": null,325 "pruned_heads": {},326 "remove_invalid_values": false,327 "repetition_penalty": 1.0,328 "return_dict": true,329 "return_dict_in_generate": false,330 "scale_embedding": false,331 "sep_token_id": null,332 "suppress_tokens": null,333 "task_specific_params": null,334 "temperature": 1.0,335 "tf_legacy_loss": false,336 "tie_encoder_decoder": false,337 "tie_word_embeddings": true,338 "tokenizer_class": null,339 "top_k": 50,340 "top_p": 1.0,341 "torchscript": false,342 "typical_p": 1.0,343 "use_bfloat16": false344 },345 "audio_end_token_id": 151670,346 "audio_start_token_id": 151669,347 "audio_token_id": 151675,348 "dtype": "bfloat16",349 "image_token_id": 151655,350 "initializer_range": 0.02,351 "model_type": "qwen3_omni_moe_thinker",352 "position_id_per_seconds": 13,353 "seconds_per_chunk": 2,354 "text_config": {355 "_name_or_path": "",356 "add_cross_attention": false,357 "architectures": null,358 "attention_bias": false,359 "attention_dropout": 0.0,360 "bad_words_ids": null,361 "begin_suppress_tokens": null,362 "bos_token_id": null,363 "chunk_size_feed_forward": 0,364 "cross_attention_hidden_size": null,365 "decoder_sparse_step": 1,366 "decoder_start_token_id": null,367 "diversity_penalty": 0.0,368 "do_sample": false,369 "dtype": null,370 "early_stopping": false,371 "encoder_no_repeat_ngram_size": 0,372 "eos_token_id": null,373 "exponential_decay_length_penalty": null,374 "finetuning_task": null,375 "forced_bos_token_id": null,376 "forced_eos_token_id": null,377 "head_dim": 128,378 "hidden_act": "silu",379 "hidden_size": 2048,380 "id2label": {381 "0": "LABEL_0",382 "1": "LABEL_1"383 },384 "initializer_range": 0.02,385 "intermediate_size": 768,386 "is_decoder": false,387 "is_encoder_decoder": false,388 "label2id": {389 "LABEL_0": 0,390 "LABEL_1": 1391 },392 "length_penalty": 1.0,393 "max_length": 20,394 "max_position_embeddings": 65536,395 "min_length": 0,396 "mlp_only_layers": [],397 "model_type": "qwen3_omni_moe_text",398 "moe_intermediate_size": 768,399 "no_repeat_ngram_size": 0,400 "norm_topk_prob": true,401 "num_attention_heads": 32,402 "num_beam_groups": 1,403 "num_beams": 1,404 "num_experts": 128,405 "num_experts_per_tok": 8,406 "num_hidden_layers": 48,407 "num_key_value_heads": 4,408 "num_return_sequences": 1,409 "output_attentions": false,410 "output_hidden_states": false,411 "output_router_logits": false,412 "output_scores": false,413 "pad_token_id": null,414 "prefix": null,415 "problem_type": null,416 "pruned_heads": {},417 "remove_invalid_values": false,418 "repetition_penalty": 1.0,419 "return_dict": true,420 "return_dict_in_generate": false,421 "rms_norm_eps": 1e-06,422 "rope_scaling": {423 "interleaved": true,424 "mrope_interleaved": true,425 "mrope_section": [426 24,427 20,428 20429 ],430 "rope_type": "default",431 "type": "default"432 },433 "rope_theta": 1000000,434 "router_aux_loss_coef": 0.001,435 "sep_token_id": null,436 "shared_expert_intermediate_size": 0,437 "sliding_window": null,438 "suppress_tokens": null,439 "task_specific_params": null,440 "temperature": 1.0,441 "tf_legacy_loss": false,442 "tie_encoder_decoder": false,443 "tie_word_embeddings": false,444 "tokenizer_class": null,445 "top_k": 50,446 "top_p": 1.0,447 "torchscript": false,448 "typical_p": 1.0,449 "use_bfloat16": false,450 "use_cache": true,451 "use_qk_norm": true,452 "use_sliding_window": false,453 "vocab_size": 152064454 },455 "user_token_id": 872,456 "video_token_id": 151656,457 "vision_config": {458 "_name_or_path": "",459 "add_cross_attention": false,460 "apply_vit_abs_pos_embed": true,461 "architectures": null,462 "bad_words_ids": null,463 "begin_suppress_tokens": null,464 "bos_token_id": null,465 "chunk_size_feed_forward": 0,466 "cross_attention_hidden_size": null,467 "decoder_start_token_id": null,468 "deepstack_visual_indexes": [469 8,470 16,471 24472 ],473 "depth": 27,474 "diversity_penalty": 0.0,475 "do_sample": false,476 "dtype": null,477 "early_stopping": false,478 "encoder_no_repeat_ngram_size": 0,479 "eos_token_id": null,480 "exponential_decay_length_penalty": null,481 "finetuning_task": null,482 "forced_bos_token_id": null,483 "forced_eos_token_id": null,484 "hidden_act": "gelu_pytorch_tanh",485 "hidden_size": 1152,486 "id2label": {487 "0": "LABEL_0",488 "1": "LABEL_1"489 },490 "image_size": 768,491 "in_channels": 3,492 "in_chans": 3,493 "initializer_range": 0.02,494 "intermediate_size": 4304,495 "is_decoder": false,496 "is_encoder_decoder": false,497 "label2id": {498 "LABEL_0": 0,499 "LABEL_1": 1500 },501 "length_penalty": 1.0,502 "max_length": 20,503 "min_length": 0,504 "model_type": "qwen3_omni_moe_vision_encoder",505 "no_repeat_ngram_size": 0,506 "num_beam_groups": 1,507 "num_beams": 1,508 "num_heads": 16,509 "num_position_embeddings": 2304,510 "num_return_sequences": 1,511 "out_hidden_size": 2048,512 "output_attentions": false,513 "output_hidden_states": false,514 "output_scores": false,515 "pad_token_id": null,516 "patch_size": 16,517 "prefix": null,518 "problem_type": null,519 "pruned_heads": {},520 "remove_invalid_values": false,521 "repetition_penalty": 1.0,522 "return_dict": true,523 "return_dict_in_generate": false,524 "sep_token_id": null,525 "spatial_merge_size": 2,526 "spatial_patch_size": 16,527 "suppress_tokens": null,528 "task_specific_params": null,529 "temperature": 1.0,530 "temporal_patch_size": 2,531 "tf_legacy_loss": false,532 "tie_encoder_decoder": false,533 "tie_word_embeddings": true,534 "tokenizer_class": null,535 "tokens_per_second": 2,536 "top_k": 50,537 "top_p": 1.0,538 "torchscript": false,539 "typical_p": 1.0,540 "use_bfloat16": false541 },542 "vision_end_token_id": 151653,543 "vision_start_token_id": 151652544 },545 "transformers_version": "4.57.1",546 "tts_bos_token_id": 151672,547 "tts_eos_token_id": 151673,548 "tts_pad_token_id": 151671,549 "user_token_id": 872550}551