LiquidAI/LFM2-Audio-1.5B

1{2  "architectures": [3    "Lfm2AudioForConditionalGeneration"4  ],5  "codebooks": 8,6  "tie_audio_embeddings": false,7  "semantic_codebook_factor": 100,8  "codebook_weight": "log",9  "interleaved_n_text": 6,10  "interleaved_n_audio": 12,11  "preprocessor": {12    "sample_rate": 16000,13    "normalize": "per_feature",14    "window_size": 0.025,15    "window_stride": 0.01,16    "window": "hann",17    "features": 128,18    "n_fft": 512,19    "log": true,20    "frame_splicing": 1,21    "dither": 1.0e-05,22    "pad_to": 0,23    "pad_value": 0.024  },25  "encoder": {26    "feat_in": 128,27    "feat_out": -1,28    "n_layers": 17,29    "d_model": 512,30    "subsampling": "dw_striding",31    "subsampling_factor": 8,32    "subsampling_conv_channels": 256,33    "causal_downsampling": false,34    "reduction": null,35    "reduction_position": null,36    "reduction_factor": 1,37    "ff_expansion_factor": 4,38    "self_attention_model": "rel_pos",39    "n_heads": 8,40    "att_context_size": [41      -1,42      -143    ],44    "xscaling": false,45    "untie_biases": true,46    "pos_emb_max_len": 5000,47    "conv_kernel_size": 9,48    "conv_norm_type": "batch_norm",49    "conv_context_size": null,50    "dropout": 0.1,51    "dropout_pre_encoder": 0.1,52    "dropout_emb": 0,53    "dropout_att": 0.154  },55  "lfm": {56    "_name_or_path": "LiquidAI/LFM2-1.2B",57    "architectures": [58      "Lfm2ForCausalLM"59    ],60    "block_auto_adjust_ff_dim": true,61    "block_dim": 2048,62    "block_ff_dim": 12288,63    "block_ffn_dim_multiplier": 1.0,64    "block_mlp_init_scale": 1,65    "block_multiple_of": 256,66    "block_norm_eps": 1e-05,67    "block_out_init_scale": 1,68    "block_use_swiglu": true,69    "block_use_xavier_init": true,70    "conv_L_cache": 3,71    "conv_bias": false,72    "conv_dim": 2048,73    "conv_dim_out": 2048,74    "conv_use_xavier_init": true,75    "eos_token_id": 7,76    "hidden_size": 2048,77    "initializer_range": 0.02,78    "intermediate_size": 12288,79    "layer_types": [80      "conv",81      "conv",82      "full_attention",83      "conv",84      "conv",85      "full_attention",86      "conv",87      "conv",88      "full_attention",89      "conv",90      "full_attention",91      "conv",92      "full_attention",93      "conv",94      "full_attention",95      "conv"96    ],97    "max_position_embeddings": 128000,98    "model_type": "lfm2",99    "norm_eps": 1e-05,100    "num_attention_heads": 32,101    "num_heads": 32,102    "num_hidden_layers": 16,103    "num_key_value_heads": 8,104    "rope_theta": 1000000,105    "torch_dtype": "bfloat16",106    "use_cache": true,107    "use_pos_enc": true,108    "vocab_size": 65536109  },110  "depthformer": {111    "layers": 6,112    "dim": 1024,113    "tie": true114  }115}116