| [ | |
| 43, | |
| 80, | |
| { | |
| "accum_grad": 1, | |
| "adim": 384, | |
| "aheads": 4, | |
| "backend": "pytorch", | |
| "batch_bins": 4554000, | |
| "batch_count": "auto", | |
| "batch_frames_in": 0, | |
| "batch_frames_inout": 0, | |
| "batch_frames_out": 0, | |
| "batch_size": 0, | |
| "batch_sort_key": "output", | |
| "bce_pos_weight": 5.0, | |
| "config": "conf/train_pytorch_transformer.yaml", | |
| "config2": null, | |
| "config3": null, | |
| "debugmode": 1, | |
| "dec_init": null, | |
| "dec_init_mods": [ | |
| "dec." | |
| ], | |
| "decoder_concat_after": false, | |
| "decoder_normalize_before": false, | |
| "dlayers": 3, | |
| "dprenet_dropout_rate": 0.5, | |
| "dprenet_layers": 2, | |
| "dprenet_units": 256, | |
| "dunits": 1536, | |
| "early_stop_criterion": "validation/main/loss", | |
| "elayers": 3, | |
| "embed_dim": 0, | |
| "enc_init": null, | |
| "enc_init_mods": [ | |
| "enc." | |
| ], | |
| "encoder_concat_after": false, | |
| "encoder_normalize_before": false, | |
| "epochs": 1000, | |
| "eprenet_conv_chans": 0, | |
| "eprenet_conv_filts": 0, | |
| "eprenet_conv_layers": 0, | |
| "eprenet_dropout_rate": 0.0, | |
| "eps": 1e-06, | |
| "eunits": 1536, | |
| "eval_interval_epochs": 1, | |
| "freeze_mods": null, | |
| "grad_clip": 1.0, | |
| "guided_attn_loss_lambda": 1.0, | |
| "guided_attn_loss_sigma": 0.4, | |
| "initial_decoder_alpha": 1.0, | |
| "initial_encoder_alpha": 1.0, | |
| "keep_all_data_on_mem": false, | |
| "loss_type": "L1", | |
| "lr": 0.001, | |
| "maxlen_in": 100, | |
| "maxlen_out": 200, | |
| "minibatches": 0, | |
| "model_module": "espnet.nets.pytorch_backend.e2e_tts_transformer:Transformer", | |
| "modules_applied_guided_attn": [ | |
| "encoder-decoder" | |
| ], | |
| "ngpu": 3, | |
| "num_heads_applied_guided_attn": 2, | |
| "num_iter_processes": 0, | |
| "num_layers_applied_guided_attn": 2, | |
| "num_save_attention": 5, | |
| "opt": "noam", | |
| "outdir": "exp/phn_train_no_dev_pytorch_train_pytorch_transformer/results", | |
| "patience": 0, | |
| "positionwise_conv_kernel_size": 1, | |
| "positionwise_layer_type": "linear", | |
| "postnet_chans": 256, | |
| "postnet_dropout_rate": 0.5, | |
| "postnet_filts": 5, | |
| "postnet_layers": 5, | |
| "preprocess_conf": null, | |
| "pretrained_model": null, | |
| "reduction_factor": 3, | |
| "report_interval_iters": 100, | |
| "resume": null, | |
| "save_interval_epochs": 10, | |
| "seed": 1, | |
| "sortagrad": 0, | |
| "spc_dim": null, | |
| "spk_embed_dim": null, | |
| "spk_embed_integration_type": "add", | |
| "tensorboard_dir": "tensorboard/phn_train_no_dev_pytorch_train_pytorch_transformer", | |
| "train_json": "dump/phn_train_no_dev/data.json", | |
| "transformer_dec_attn_dropout_rate": 0.1, | |
| "transformer_dec_dropout_rate": 0.1, | |
| "transformer_dec_positional_dropout_rate": 0.1, | |
| "transformer_enc_attn_dropout_rate": 0.1, | |
| "transformer_enc_dec_attn_dropout_rate": 0.1, | |
| "transformer_enc_dropout_rate": 0.1, | |
| "transformer_enc_positional_dropout_rate": 0.1, | |
| "transformer_init": "pytorch", | |
| "transformer_lr": 1.0, | |
| "transformer_warmup_steps": 4000, | |
| "use_batch_norm": true, | |
| "use_guided_attn_loss": true, | |
| "use_masking": true, | |
| "use_scaled_pos_enc": true, | |
| "use_second_target": false, | |
| "use_speaker_embedding": false, | |
| "use_weighted_masking": false, | |
| "valid_json": "dump/phn_dev/data.json", | |
| "verbose": 1, | |
| "version": "202209", | |
| "weight_decay": 0.0 | |
| } | |
| ] |