diff options
| -rw-r--r-- | become_yukarin/config.py | 2 | ||||
| -rw-r--r-- | become_yukarin/dataset/dataset.py | 54 | ||||
| -rw-r--r-- | become_yukarin/model.py | 20 |
3 files changed, 61 insertions, 15 deletions
diff --git a/become_yukarin/config.py b/become_yukarin/config.py index 424e598..864de4e 100644 --- a/become_yukarin/config.py +++ b/become_yukarin/config.py @@ -16,6 +16,7 @@ class DatasetConfig(NamedTuple): target_mean_path: Path target_var_path: Path features: List[str] + train_crop_size: int seed: int num_test: int @@ -82,6 +83,7 @@ def create_from_json(s: Union[str, Path]): target_mean_path=Path(d['dataset']['target_mean_path']).expanduser(), target_var_path=Path(d['dataset']['target_var_path']).expanduser(), features=d['dataset']['features'], + train_crop_size=d['dataset']['train_crop_size'], seed=d['dataset']['seed'], num_test=d['dataset']['num_test'], ), diff --git a/become_yukarin/dataset/dataset.py b/become_yukarin/dataset/dataset.py index 6328a1c..a6550b7 100644 --- a/become_yukarin/dataset/dataset.py +++ b/become_yukarin/dataset/dataset.py @@ -1,3 +1,4 @@ +import copy import typing from abc import ABCMeta, abstractmethod from collections import defaultdict @@ -41,13 +42,16 @@ class DictKeyReplaceProcess(BaseDataProcess): class ChainProcess(BaseDataProcess): def __init__(self, process: typing.Iterable[BaseDataProcess]): - self._process = process + self._process = list(process) def __call__(self, data, test): for p in self._process: data = p(data, test) return data + def append(self, process: BaseDataProcess): + self._process.append(process) + class SplitProcess(BaseDataProcess): def __init__(self, process: typing.Dict[str, typing.Optional[BaseDataProcess]]): @@ -248,6 +252,24 @@ class ShapeAlignProcess(BaseDataProcess): return data +class CropProcess(BaseDataProcess): + def __init__(self, crop_size: int, time_axis: int = 1): + self._crop_size = crop_size + self._time_axis = time_axis + + def __call__(self, datas: Dict[str, any], test=True): + assert not test + + data, seed = datas['data'], datas['seed'] + random = numpy.random.RandomState(seed) + + len_time = data.shape[self._time_axis] + assert len_time >= self._crop_size + + start = random.randint(len_time - self._crop_size + 1) + return numpy.split(data, [start, start + self._crop_size], axis=self._time_axis)[1] + + class DataProcessDataset(chainer.dataset.DatasetMixin): def __init__(self, data: typing.List, data_process: BaseDataProcess): self._data = data @@ -273,7 +295,7 @@ def create(config: DatasetConfig): target_var = acoustic_feature_load_process(config.target_var_path, test=True) # {input_path, target_path} - data_process = ChainProcess([ + data_process_base = ChainProcess([ SplitProcess(dict( input=ChainProcess([ LambdaProcess(lambda d, test: d['input_path']), @@ -300,6 +322,28 @@ def create(config: DatasetConfig): ShapeAlignProcess(), ]) + data_process_train = copy.deepcopy(data_process_base) + if config.train_crop_size is not None: + data_process_train.append(ChainProcess([ + LambdaProcess(lambda d, test: dict(seed=numpy.random.randint(2 ** 32), **d)), + SplitProcess(dict( + input=ChainProcess([ + LambdaProcess(lambda d, test: dict(data=d['input'], seed=d['seed'])), + CropProcess(crop_size=config.train_crop_size), + ]), + target=ChainProcess([ + LambdaProcess(lambda d, test: dict(data=d['target'], seed=d['seed'])), + CropProcess(crop_size=config.train_crop_size), + ]), + mask=ChainProcess([ + LambdaProcess(lambda d, test: dict(data=d['mask'], seed=d['seed'])), + CropProcess(crop_size=config.train_crop_size), + ]), + )), + ])) + + data_process_test = data_process_base + num_test = config.num_test pairs = [ dict(input_path=input_path, target_path=target_path) @@ -311,7 +355,7 @@ def create(config: DatasetConfig): train_for_evaluate_paths = train_paths[:num_test] return { - 'train': DataProcessDataset(train_paths, data_process), - 'test': DataProcessDataset(test_paths, data_process), - 'train_eval': DataProcessDataset(train_for_evaluate_paths, data_process), + 'train': DataProcessDataset(train_paths, data_process_train), + 'test': DataProcessDataset(test_paths, data_process_test), + 'train_eval': DataProcessDataset(train_for_evaluate_paths, data_process_test), } diff --git a/become_yukarin/model.py b/become_yukarin/model.py index 5f46876..9d69378 100644 --- a/become_yukarin/model.py +++ b/become_yukarin/model.py @@ -112,7 +112,7 @@ class CBHG(chainer.link.Chain): mode='constant', ) self.max_pooling = chainer.functions.MaxPoolingND(1, max_pooling_k, 1, cover_all=False) - self.out_size = out_channels + self.out_size = out_channels * 2 with self.init_scope(): self.conv_bank = Conv1DBank( @@ -128,12 +128,12 @@ class CBHG(chainer.link.Chain): self.highways = chainer.link.ChainList( *([ConvHighway(out_channels) for _ in range(highway_layers)]) ) - # self.gru = chainer.links.NStepBiGRU( - # n_layers=1, - # in_size=out_channels, - # out_size=out_channels, - # dropout=0.0, - # ) + self.gru = chainer.links.NStepBiGRU( + n_layers=1, + in_size=out_channels, + out_size=out_channels, + dropout=0.0, + ) def __call__(self, x): h = x @@ -144,9 +144,9 @@ class CBHG(chainer.link.Chain): for highway in self.highways: h = highway(h) - # h = chainer.functions.separate(chainer.functions.transpose(h, axes=(0, 2, 1))) - # _, h = self.gru(None, h) - # h = chainer.functions.transpose(chainer.functions.stack(h), axes=(0, 2, 1)) + h = chainer.functions.separate(chainer.functions.transpose(h, axes=(0, 2, 1))) + _, h = self.gru(None, h) + h = chainer.functions.transpose(chainer.functions.stack(h), axes=(0, 2, 1)) return h |