{ "architectures": [ "Qwen2AudioModel" ], "audio_config": { "activation_dropout": 0.0, "activation_function": "gelu", "attention_dropout": 0.0, "d_model": 32, "dropout": 0.0, "encoder_attention_heads": 2, "encoder_ffn_dim": 32, "encoder_layerdrop": 0.0, "encoder_layers": 2, "initializer_range": 0.02, "max_source_positions": 30, "model_type": "qwen2_audio_encoder", "num_mel_bins": 80, "scale_embedding": false }, "audio_token_index": 0, "dtype": "float32", "model_type": "qwen2_audio", "pad_token_id": 151643, "text_config": { "attention_dropout": 0.0, "bos_token_id": 1, "eos_token_id": 151645, "hidden_act": "gelu", "hidden_size": 32, "initializer_range": 0.02, "intermediate_size": 32, "layer_types": [ "full_attention", "full_attention" ], "max_position_embeddings": 512, "max_window_layers": 28, "model_type": "qwen2", "num_attention_heads": 2, "num_hidden_layers": 2, "num_key_value_heads": 2, "pad_token_id": 151643, "rms_norm_eps": 1e-06, "rope_parameters": { "rope_theta": 10000.0, "rope_type": "default" }, "sliding_window": null, "tie_word_embeddings": false, "use_cache": true, "use_sliding_window": false, "vocab_size": 154931 }, "transformers_version": "5.16.0.dev0" }