nnsvs.gen.predict_timing

nnsvs.gen.predict_timing(device, labels, binary_dict, numeric_dict, timelag_model, timelag_config, timelag_in_scaler, timelag_out_scaler, duration_model, duration_config, duration_in_scaler, duration_out_scaler, log_f0_conditioning=True, allowed_range=None, allowed_range_rest=None, force_clip_input_features=True, frame_period=5)[source]

Predict timinigs from HTS labels

This is equivalent to predict_timelag + predict_duration + postprocess_duration.

Parameters:

  • device (torch.device) – device to run the model on

  • labels (nnmnkwii.io.hts.HTSLabelFile) – labels

  • binary_dict (dict) – binary feature dictionary

  • numeric_dict (dict) – numeric feature dictionary

  • timelag_model (nn.Module) – timelag model

  • timelag_config (dict) – timelag config

  • timelag_in_scaler (sklearn.preprocessing.MinMaxScaler) – timelag input scaler

  • timelag_out_scaler (sklearn.preprocessing.MinMaxScaler) – timelag output scaler

  • duration_model (nn.Module) – duration model

  • duration_config (dict) – duration config

  • duration_in_scaler (sklearn.preprocessing.MinMaxScaler) – duration input scaler

  • duration_out_scaler (sklearn.preprocessing.MinMaxScaler) – duration output scaler

  • log_f0_conditioning (bool) – whether to condition on log f0

  • allowed_range (list) – allowed range of time-lag

  • allowed_range_rest (list) – allowed range of time-lag for rest

  • force_clip_input_features (bool) – whether to clip input features

  • frame_period (int) – frame period in milliseconds

Returns:

duration modified labels

Return type:

nnmnkwii.io.hts.HTSLabelFile