mazesmazes/tiny-audio-next-plus
05
1{2 "architectures": [3 "ASRModel"4 ],5 "attn_implementation": "sdpa",6 "audio_config": {7 "_name_or_path": "zai-org/GLM-ASR-Nano-2512",8 "architectures": [9 "GlmAsrForConditionalGeneration"10 ],11 "audio_config": {12 "_name_or_path": "",13 "architectures": null,14 "attention_dropout": 0.0,15 "chunk_size_feed_forward": 0,16 "dtype": null,17 "head_dim": 64,18 "hidden_act": "gelu",19 "hidden_size": 1280,20 "id2label": {21 "0": "LABEL_0",22 "1": "LABEL_1"23 },24 "initializer_range": 0.02,25 "intermediate_size": 5120,26 "is_encoder_decoder": false,27 "label2id": {28 "LABEL_0": 0,29 "LABEL_1": 130 },31 "max_position_embeddings": 1500,32 "model_type": "glmasr_encoder",33 "num_attention_heads": 20,34 "num_hidden_layers": 32,35 "num_key_value_heads": 20,36 "num_mel_bins": 128,37 "output_attentions": false,38 "output_hidden_states": false,39 "partial_rotary_factor": 0.5,40 "problem_type": null,41 "return_dict": true,42 "rope_parameters": {43 "partial_rotary_factor": 0.5,44 "rope_theta": 10000.0,45 "rope_type": "default"46 }47 },48 "audio_token_id": 59260,49 "dtype": "float32",50 "hidden_size": 2048,51 "model_type": "glmasr",52 "num_mel_bins": 128,53 "projector_hidden_act": "gelu",54 "text_config": {55 "_name_or_path": "",56 "architectures": null,57 "attention_bias": false,58 "attention_dropout": 0.0,59 "bos_token_id": 1,60 "chunk_size_feed_forward": 0,61 "dtype": null,62 "eos_token_id": [63 59246,64 59253,65 5925566 ],67 "head_dim": 128,68 "hidden_act": "silu",69 "hidden_size": 2048,70 "id2label": {71 "0": "LABEL_0",72 "1": "LABEL_1"73 },74 "initializer_range": 0.02,75 "intermediate_size": 6144,76 "is_encoder_decoder": false,77 "label2id": {78 "LABEL_0": 0,79 "LABEL_1": 180 },81 "max_position_embeddings": 8192,82 "mlp_bias": false,83 "model_type": "llama",84 "num_attention_heads": 16,85 "num_hidden_layers": 28,86 "num_key_value_heads": 4,87 "output_attentions": false,88 "output_hidden_states": false,89 "pad_token_id": null,90 "pretraining_tp": 1,91 "problem_type": null,92 "return_dict": true,93 "rms_norm_eps": 1e-05,94 "rope_parameters": {95 "rope_theta": 10000.0,96 "rope_type": "default"97 },98 "tie_word_embeddings": false,99 "use_cache": true,100 "vocab_size": 59264101 },102 "vocab_size": 59264103 },104 "audio_model_id": "zai-org/GLM-ASR-Nano-2512",105 "audio_sample_rate": 16000,106 "auto_map": {107 "AutoConfig": "asr_config.ASRConfig",108 "AutoModel": "asr_modeling.ASRModel",109 "AutoModelForSpeechSeq2Seq": "asr_modeling.ASRModel",110 "AutoProcessor": "asr_processing.ASRProcessor"111 },112 "bos_token_id": null,113 "custom_pipelines": {114 "automatic-speech-recognition": {115 "impl": "asr_pipeline.ASRPipeline",116 "pt": [117 "AutoModelForSpeechSeq2Seq"118 ],119 "tf": [],120 "type": "audio"121 }122 },123 "do_sample": false,124 "downsample_rate": 5,125 "dtype": "float32",126 "encoder": {127 "_name_or_path": "zai-org/GLM-ASR-Nano-2512",128 "architectures": [129 "GlmAsrForConditionalGeneration"130 ],131 "audio_config": {132 "_name_or_path": "",133 "architectures": null,134 "attention_dropout": 0.0,135 "chunk_size_feed_forward": 0,136 "dtype": null,137 "head_dim": 64,138 "hidden_act": "gelu",139 "hidden_size": 1280,140 "id2label": {141 "0": "LABEL_0",142 "1": "LABEL_1"143 },144 "initializer_range": 0.02,145 "intermediate_size": 5120,146 "is_encoder_decoder": false,147 "label2id": {148 "LABEL_0": 0,149 "LABEL_1": 1150 },151 "max_position_embeddings": 1500,152 "model_type": "glmasr_encoder",153 "num_attention_heads": 20,154 "num_hidden_layers": 32,155 "num_key_value_heads": 20,156 "num_mel_bins": 128,157 "output_attentions": false,158 "output_hidden_states": false,159 "partial_rotary_factor": 0.5,160 "problem_type": null,161 "return_dict": true,162 "rope_parameters": {163 "partial_rotary_factor": 0.5,164 "rope_theta": 10000.0,165 "rope_type": "default"166 }167 },168 "audio_token_id": 59260,169 "dtype": "float32",170 "hidden_size": 2048,171 "model_type": "glmasr",172 "num_mel_bins": 128,173 "projector_hidden_act": "gelu",174 "text_config": {175 "_name_or_path": "",176 "architectures": null,177 "attention_bias": false,178 "attention_dropout": 0.0,179 "bos_token_id": 1,180 "chunk_size_feed_forward": 0,181 "dtype": null,182 "eos_token_id": [183 59246,184 59253,185 59255186 ],187 "head_dim": 128,188 "hidden_act": "silu",189 "hidden_size": 2048,190 "id2label": {191 "0": "LABEL_0",192 "1": "LABEL_1"193 },194 "initializer_range": 0.02,195 "intermediate_size": 6144,196 "is_encoder_decoder": false,197 "label2id": {198 "LABEL_0": 0,199 "LABEL_1": 1200 },201 "max_position_embeddings": 8192,202 "mlp_bias": false,203 "model_type": "llama",204 "num_attention_heads": 16,205 "num_hidden_layers": 28,206 "num_key_value_heads": 4,207 "output_attentions": false,208 "output_hidden_states": false,209 "pad_token_id": null,210 "pretraining_tp": 1,211 "problem_type": null,212 "return_dict": true,213 "rms_norm_eps": 1e-05,214 "rope_parameters": {215 "rope_theta": 10000.0,216 "rope_type": "default"217 },218 "tie_word_embeddings": false,219 "use_cache": true,220 "vocab_size": 59264221 },222 "vocab_size": 59264223 },224 "encoder_conv_layers": [225 [226 1,227 3,228 1229 ],230 [231 1,232 3,233 2234 ]235 ],236 "encoder_dim": 1280,237 "eos_token_id": 151645,238 "freeze_language_model": false,239 "freeze_projector": false,240 "length_penalty": 1.0,241 "llm_dim": 1024,242 "lora_alpha": 32,243 "lora_dropout": 0.0,244 "lora_rank": 8,245 "lora_target_modules": [246 "q_proj",247 "k_proj",248 "v_proj",249 "o_proj",250 "gate_proj",251 "up_proj",252 "down_proj"253 ],254 "max_new_tokens": 256,255 "min_new_tokens": 0,256 "model_dtype": "float32",257 "model_type": "asr_model",258 "no_repeat_ngram_size": 0,259 "num_beams": 1,260 "num_experts": 4,261 "num_experts_per_tok": 2,262 "pad_token_id": 151643,263 "pipeline_tag": "automatic-speech-recognition",264 "pretrained_model_path": "mazesmazes/tiny-audio-next-plus",265 "projector_hidden_dim": 1024,266 "projector_pool_stride": 4,267 "projector_type": "mlp",268 "qformer_hidden_size": null,269 "qformer_intermediate_size": null,270 "qformer_num_heads": 16,271 "qformer_num_layers": 2,272 "qformer_window_size": 15,273 "repetition_penalty": 1.0,274 "router_aux_loss_coef": 0.01,275 "system_prompt": "",276 "temperature": null,277 "text_config": {278 "_name_or_path": "Qwen/Qwen3-0.6B",279 "architectures": [280 "Qwen3ForCausalLM"281 ],282 "attention_bias": false,283 "attention_dropout": 0.0,284 "bos_token_id": null,285 "dtype": "float32",286 "eos_token_id": 151645,287 "head_dim": 128,288 "hidden_act": "silu",289 "hidden_size": 1024,290 "initializer_range": 0.02,291 "intermediate_size": 3072,292 "layer_types": [293 "full_attention",294 "full_attention",295 "full_attention",296 "full_attention",297 "full_attention",298 "full_attention",299 "full_attention",300 "full_attention",301 "full_attention",302 "full_attention",303 "full_attention",304 "full_attention",305 "full_attention",306 "full_attention",307 "full_attention",308 "full_attention",309 "full_attention",310 "full_attention",311 "full_attention",312 "full_attention",313 "full_attention",314 "full_attention",315 "full_attention",316 "full_attention",317 "full_attention",318 "full_attention",319 "full_attention",320 "full_attention"321 ],322 "max_position_embeddings": 40960,323 "max_window_layers": 28,324 "model_type": "qwen3",325 "num_attention_heads": 16,326 "num_hidden_layers": 28,327 "num_key_value_heads": 8,328 "pad_token_id": 151643,329 "rms_norm_eps": 1e-06,330 "rope_parameters": {331 "rope_theta": 1000000,332 "rope_type": "default"333 },334 "sliding_window": null,335 "tie_word_embeddings": true,336 "use_cache": true,337 "use_sliding_window": false,338 "vocab_size": 151670339 },340 "text_model_id": "Qwen/Qwen3-0.6B",341 "top_k": null,342 "top_p": null,343 "transformers_version": "5.7.0",344 "use_cache": false,345 "use_lora": false,346 "vocab_size": 151670347}348