{ "architectures": [ "CsmForConditionalGeneration" ], "attention_bias": false, "attention_dropout": 0.0, "audio_eos_token_id": 128003, "audio_token_id": 128002, "bos_token_id": 128000, "codebook_eos_token_id": 3, "codebook_pad_token_id": 2, "codec_config": { "_frame_rate": null, "attention_bias": false, "attention_dropout": 0.0, "audio_channels": 1, "chunk_in_sec": null, "codebook_dim": 64, "codebook_size": 64, "compress": 2, "dilation_growth_rate": 2, "head_dim": 16, "hidden_act": "gelu", "hidden_size": 32, "initializer_range": 0.02, "intermediate_size": 2048, "kernel_size": 7, "last_kernel_size": 3, "layer_scale_initial_scale": 0.01, "max_position_embeddings": 8000, "model_type": "mimi", "norm_eps": 1e-05, "num_attention_heads": 2, "num_filters": 8, "num_hidden_layers": 2, "num_key_value_heads": 2, "num_quantizers": 32, "num_residual_layers": 1, "num_semantic_quantizers": 1, "pad_mode": "constant", "residual_kernel_size": 3, "rope_parameters": { "rope_theta": 10000.0, "rope_type": "default" }, "sampling_rate": 24000, "sliding_window": 4, "tie_word_embeddings": true, "trim_right_ratio": 1.0, "upsample_groups": 32, "upsampling_ratios": [ 8, 4 ], "use_cache": false, "use_causal_conv": true, "use_conv_shortcut": false, "use_streaming": false, "vector_quantization_hidden_dimension": 64 }, "depth_decoder_config": { "attention_bias": false, "attention_dropout": 0.0, "backbone_hidden_size": 64, "bos_token_id": null, "eos_token_id": null, "head_dim": 16, "hidden_act": "silu", "hidden_size": 64, "initializer_range": 0.02, "intermediate_size": 128, "max_position_embeddings": 10, "mlp_bias": false, "model_type": "csm_depth_decoder_model", "num_attention_heads": 4, "num_codebooks": 10, "num_hidden_layers": 2, "num_key_value_heads": 2, "pad_token_id": null, "rms_norm_eps": 1e-05, "rope_parameters": { "rope_theta": 500000.0, "rope_type": "default" }, "use_cache": true, "vocab_size": 6 }, "dtype": "float32", "eos_token_id": 128001, "head_dim": 16, "hidden_act": "silu", "hidden_size": 64, "initializer_range": 0.02, "intermediate_size": 64, "max_position_embeddings": 200, "mlp_bias": false, "model_type": "csm", "num_attention_heads": 4, "num_codebooks": 10, "num_hidden_layers": 2, "num_key_value_heads": 2, "pad_token_id": 128001, "rms_norm_eps": 1e-05, "rope_parameters": { "rope_theta": 500000.0, "rope_type": "default" }, "text_vocab_size": 99, "tie_codebooks_embeddings": true, "tie_word_embeddings": false, "transformers_version": "5.16.0.dev0", "use_cache": true, "vocab_size": 128256 }