FreedomIntelligence/EchoX-8B
1076
1{2 "input_wavs_dir": "/private/home/adampolyak/datasets/LJ/LJSpeech-1.1/wavs_16khz_padded",3 "input_training_file": "/large_experiments/ust/annl/datasets/tts/LJSpeech/filelist/mhubert_vp_en_es_fr_it3_400k/lj_train_layer11_hubert1000_filelist.txt",4 "input_validation_file": "/large_experiments/ust/annl/datasets/tts/LJSpeech/filelist/mhubert_vp_en_es_fr_it3_400k/lj_dev_layer11_hubert1000_filelist.txt",5 6 "resblock": "1",7 "num_gpus": 0,8 "batch_size": 16,9 "learning_rate": 0.0002,10 "adam_b1": 0.8,11 "adam_b2": 0.99,12 "lr_decay": 0.999,13 "seed": 1234,14 15 "upsample_rates": [5,4,4,2,2],16 "upsample_kernel_sizes": [11,8,8,4,4],17 "upsample_initial_channel": 512,18 "resblock_kernel_sizes": [3,7,11],19 "resblock_dilation_sizes": [[1,3,5], [1,3,5], [1,3,5]],20 "num_embeddings": 1000,21 "embedding_dim": 128,22 "model_in_dim": 128,23 24 "segment_size": 8960,25 "code_hop_size": 320,26 "f0": false,27 "num_mels": 80,28 "num_freq": 1025,29 "n_fft": 1024,30 "hop_size": 256,31 "win_size": 1024,32 33 "dur_prediction_weight": 1.0,34 "dur_predictor_params": {35 "encoder_embed_dim": 128,36 "var_pred_hidden_dim": 128,37 "var_pred_kernel_size": 3,38 "var_pred_dropout": 0.539 },40 41 "sampling_rate": 16000,42 43 "fmin": 0,44 "fmax": 8000,45 "fmax_for_loss": null,46 47 "num_workers": 4,48 49 "dist_config": {50 "dist_backend": "nccl",51 "dist_url": "env://"52 }53}54 