{ "access": { "embargo": { "active": false, "reason": null }, "files": "public", "record": "public", "status": "open" }, "created": "2020-12-21T03:07:08.672426+00:00", "custom_fields": {}, "deletion_status": { "is_deleted": false, "status": "P" }, "files": { "count": 1, "enabled": true, "entries": { "tts_train_fastspeech2_tacotron2_teacher_raw_phn_jaconv_pyopenjtalk_accent_train.loss.ave.zip": { "checksum": "md5:bad2b35f2ef5aa087a304dbe407b51d3", "ext": "zip", "id": "c215bac4-78ba-4703-95f6-4e7f774293a7", "key": "tts_train_fastspeech2_tacotron2_teacher_raw_phn_jaconv_pyopenjtalk_accent_train.loss.ave.zip", "metadata": null, "mimetype": "application/zip", "size": 149542429 } }, "order": [], "total_bytes": 149542429 }, "id": "4381100", "is_draft": false, "is_published": true, "links": { "access": "https://zenodo.org/api/records/4381100/access", "access_links": "https://zenodo.org/api/records/4381100/access/links", "access_request": "https://zenodo.org/api/records/4381100/access/request", "access_users": "https://zenodo.org/api/records/4381100/access/users", "archive": "https://zenodo.org/api/records/4381100/files-archive", "archive_media": "https://zenodo.org/api/records/4381100/media-files-archive", "communities": "https://zenodo.org/api/records/4381100/communities", "communities-suggestions": "https://zenodo.org/api/records/4381100/communities-suggestions", "doi": "https://doi.org/10.5281/zenodo.4381100", "draft": "https://zenodo.org/api/records/4381100/draft", "files": "https://zenodo.org/api/records/4381100/files", "latest": "https://zenodo.org/api/records/4381100/versions/latest", "latest_html": "https://zenodo.org/records/4381100/latest", "media_files": "https://zenodo.org/api/records/4381100/media-files", "parent": "https://zenodo.org/api/records/4381099", "parent_doi": "https://zenodo.org/doi/10.5281/zenodo.4381099", "parent_html": "https://zenodo.org/records/4381099", "requests": "https://zenodo.org/api/records/4381100/requests", "reserve_doi": "https://zenodo.org/api/records/4381100/draft/pids/doi", "self": "https://zenodo.org/api/records/4381100", "self_doi": "https://zenodo.org/doi/10.5281/zenodo.4381100", "self_html": "https://zenodo.org/records/4381100", "self_iiif_manifest": "https://zenodo.org/api/iiif/record:4381100/manifest", "self_iiif_sequence": "https://zenodo.org/api/iiif/record:4381100/sequence/default", "versions": "https://zenodo.org/api/records/4381100/versions" }, "media_files": { "count": 0, "enabled": false, "entries": {}, "order": [], "total_bytes": 0 }, "metadata": { "creators": [ { "person_or_org": { "family_name": "kan-bayashi", "name": "kan-bayashi", "type": "personal" } } ], "description": "
This model was trained by kan-bayashi using jsut/tts1 recipe in espnet.
\n\n\n\n
See https://github.com/espnet/espnet_model_zoo
\n\tgit clone https://github.com/espnet/espnet\ncd espnet\ngit checkout acd69577240687cc1c6c9d7ca024462aa87fcf89\npip install -e .\ncd egs2/jsut/tts1\n# Download the model file here\n./run.sh --skip_data_prep false --skip_train true --download_model kan-bayashi/jsut_tts_train_fastspeech2_tacotron2_teacher_raw_phn_jaconv_pyopenjtalk_accent_train.loss.ave
\n
\n\tconfig: conf/tuning/train_fastspeech2.yaml\nprint_config: false\nlog_level: INFO\ndry_run: false\niterator_type: sequence\noutput_dir: exp/tts_train_fastspeech2_tacotron2_teacher_raw_phn_jaconv_pyopenjtalk_accent\nngpu: 1\nseed: 0\nnum_workers: 1\nnum_att_plot: 3\ndist_backend: nccl\ndist_init_method: env://\ndist_world_size: 4\ndist_rank: 0\nlocal_rank: 0\ndist_master_addr: localhost\ndist_master_port: 55835\ndist_launcher: null\nmultiprocessing_distributed: true\ncudnn_enabled: true\ncudnn_benchmark: false\ncudnn_deterministic: true\ncollect_stats: false\nwrite_collected_feats: false\nmax_epoch: 200\npatience: null\nval_scheduler_criterion:\n- valid\n- loss\nearly_stopping_criterion:\n- valid\n- loss\n- min\nbest_model_criterion:\n- - valid\n - loss\n - min\n- - train\n - loss\n - min\nkeep_nbest_models: 5\ngrad_clip: 1.0\ngrad_clip_type: 2.0\ngrad_noise: false\naccum_grad: 1\nno_forward_run: false\nresume: true\ntrain_dtype: float32\nuse_amp: false\nlog_interval: null\nunused_parameters: false\nuse_tensorboard: true\nuse_wandb: false\nwandb_project: null\nwandb_id: null\npretrain_path: null\ninit_param: []\nfreeze_param: []\nnum_iters_per_epoch: 500\nbatch_size: 20\nvalid_batch_size: null\nbatch_bins: 18000000\nvalid_batch_bins: null\ntrain_shape_file:\n- exp/tts_train_tacotron2_raw_phn_jaconv_pyopenjtalk_accent/decode_use_teacher_forcingtrue_train.loss.ave/stats/train/text_shape.phn\n- exp/tts_train_tacotron2_raw_phn_jaconv_pyopenjtalk_accent/decode_use_teacher_forcingtrue_train.loss.ave/stats/train/speech_shape\nvalid_shape_file:\n- exp/tts_train_tacotron2_raw_phn_jaconv_pyopenjtalk_accent/decode_use_teacher_forcingtrue_train.loss.ave/stats/valid/text_shape.phn\n- exp/tts_train_tacotron2_raw_phn_jaconv_pyopenjtalk_accent/decode_use_teacher_forcingtrue_train.loss.ave/stats/valid/speech_shape\nbatch_type: numel\nvalid_batch_type: null\nfold_length:\n- 150\n- 240000\nsort_in_batch: descending\nsort_batch: descending\nmultiple_iterator: false\nchunk_length: 500\nchunk_shift_ratio: 0.5\nnum_cache_chunks: 1024\ntrain_data_path_and_name_and_type:\n- - dump/raw/tr_no_dev/text\n - text\n - text\n- - exp/tts_train_tacotron2_raw_phn_jaconv_pyopenjtalk_accent/decode_use_teacher_forcingtrue_train.loss.ave/tr_no_dev/durations\n - durations\n - text_int\n- - dump/raw/tr_no_dev/wav.scp\n - speech\n - sound\n- - exp/tts_train_tacotron2_raw_phn_jaconv_pyopenjtalk_accent/decode_use_teacher_forcingtrue_train.loss.ave/stats/train/collect_feats/pitch.scp\n - pitch\n - npy\n- - exp/tts_train_tacotron2_raw_phn_jaconv_pyopenjtalk_accent/decode_use_teacher_forcingtrue_train.loss.ave/stats/train/collect_feats/energy.scp\n - energy\n - npy\nvalid_data_path_and_name_and_type:\n- - dump/raw/dev/text\n - text\n - text\n- - exp/tts_train_tacotron2_raw_phn_jaconv_pyopenjtalk_accent/decode_use_teacher_forcingtrue_train.loss.ave/dev/durations\n - durations\n - text_int\n- - dump/raw/dev/wav.scp\n - speech\n - sound\n- - exp/tts_train_tacotron2_raw_phn_jaconv_pyopenjtalk_accent/decode_use_teacher_forcingtrue_train.loss.ave/stats/valid/collect_feats/pitch.scp\n - pitch\n - npy\n- - exp/tts_train_tacotron2_raw_phn_jaconv_pyopenjtalk_accent/decode_use_teacher_forcingtrue_train.loss.ave/stats/valid/collect_feats/energy.scp\n - energy\n - npy\nallow_variable_data_keys: false\nmax_cache_size: 0.0\nmax_cache_fd: 32\nvalid_max_cache_size: null\noptim: adam\noptim_conf:\n lr: 1.0\nscheduler: noamlr\nscheduler_conf:\n model_size: 384\n warmup_steps: 4000\ntoken_list:\n- \n- \n- '1'\n- '2'\n- '0'\n- '3'\n- '4'\n- '-1'\n- '5'\n- a\n- o\n- '-2'\n- i\n- '-3'\n- u\n- e\n- k\n- n\n- t\n- '6'\n- r\n- '-4'\n- s\n- N\n- m\n- '7'\n- sh\n- d\n- g\n- w\n- '8'\n- U\n- '-5'\n- I\n- cl\n- h\n- y\n- b\n- '9'\n- j\n- ts\n- ch\n- '-6'\n- z\n- p\n- '-7'\n- f\n- ky\n- ry\n- '-8'\n- gy\n- '-9'\n- hy\n- ny\n- '-10'\n- by\n- my\n- '-11'\n- '-12'\n- '-13'\n- py\n- '-14'\n- '-15'\n- v\n- '10'\n- '-16'\n- '-17'\n- '11'\n- '-21'\n- '-20'\n- '12'\n- '-19'\n- '13'\n- '-18'\n- '14'\n- dy\n- '15'\n- ty\n- '-22'\n- '16'\n- '18'\n- '19'\n- '17'\n- \nodim: null\nmodel_conf: {}\nuse_preprocessor: true\ntoken_type: phn\nbpemodel: null\nnon_linguistic_symbols: null\ncleaner: jaconv\ng2p: pyopenjtalk_accent\nfeats_extract: fbank\nfeats_extract_conf:\n fs: 24000\n fmin: 80\n fmax: 7600\n n_mels: 80\n hop_length: 300\n n_fft: 2048\n win_length: 1200\nnormalize: global_mvn\nnormalize_conf:\n stats_file: exp/tts_train_tacotron2_raw_phn_jaconv_pyopenjtalk_accent/decode_use_teacher_forcingtrue_train.loss.ave/stats/train/feats_stats.npz\ntts: fastspeech2\ntts_conf:\n adim: 384\n aheads: 2\n elayers: 4\n eunits: 1536\n dlayers: 4\n dunits: 1536\n positionwise_layer_type: conv1d\n positionwise_conv_kernel_size: 3\n duration_predictor_layers: 2\n duration_predictor_chans: 256\n duration_predictor_kernel_size: 3\n postnet_layers: 5\n postnet_filts: 5\n postnet_chans: 256\n use_masking: true\n use_scaled_pos_enc: true\n encoder_normalize_before: true\n decoder_normalize_before: true\n reduction_factor: 1\n init_type: xavier_uniform\n init_enc_alpha: 1.0\n init_dec_alpha: 1.0\n transformer_enc_dropout_rate: 0.2\n transformer_enc_positional_dropout_rate: 0.2\n transformer_enc_attn_dropout_rate: 0.2\n transformer_dec_dropout_rate: 0.2\n transformer_dec_positional_dropout_rate: 0.2\n transformer_dec_attn_dropout_rate: 0.2\n pitch_predictor_layers: 5\n pitch_predictor_chans: 256\n pitch_predictor_kernel_size: 5\n pitch_predictor_dropout: 0.5\n pitch_embed_kernel_size: 1\n pitch_embed_dropout: 0.0\n stop_gradient_from_pitch_predictor: true\n energy_predictor_layers: 2\n energy_predictor_chans: 256\n energy_predictor_kernel_size: 3\n energy_predictor_dropout: 0.5\n energy_embed_kernel_size: 1\n energy_embed_dropout: 0.0\n stop_gradient_from_energy_predictor: false\npitch_extract: dio\npitch_extract_conf:\n fs: 24000\n n_fft: 2048\n hop_length: 300\n f0max: 400\n f0min: 80\n reduction_factor: 1\npitch_normalize: global_mvn\npitch_normalize_conf:\n stats_file: exp/tts_train_tacotron2_raw_phn_jaconv_pyopenjtalk_accent/decode_use_teacher_forcingtrue_train.loss.ave/stats/train/pitch_stats.npz\nenergy_extract: energy\nenergy_extract_conf:\n fs: 24000\n n_fft: 2048\n hop_length: 300\n win_length: 1200\n reduction_factor: 1\nenergy_normalize: global_mvn\nenergy_normalize_conf:\n stats_file: exp/tts_train_tacotron2_raw_phn_jaconv_pyopenjtalk_accent/decode_use_teacher_forcingtrue_train.loss.ave/stats/train/energy_stats.npz\nrequired:\n- output_dir\n- token_list\ndistributed: true
\n\t