Merge remote-tracking branch 'origin/develop' into merge_paddlespeech

README.md

@@ -335,4 +335,4 @@ PaddleSpeech is provided under the [Apache-2.0 License](./LICENSE).
 
 ## Acknowledgement
 
-PaddleSpeech depends on a lot of open source repos. See [references](docs/source/asr/reference.md) for more information.
+PaddleSpeech depends on a lot of open source repos. See [references](docs/source/reference.md) for more information.

audio/paddleaudio/features/core.py

@@ -11,6 +11,7 @@
 # WITHOUT WARRANTIES OR CONDITIONS OF ANY KIND, either express or implied.
 # See the License for the specific language governing permissions and
 # limitations under the License.
+# Modified from librosa(https://github.com/librosa/librosa)
 import warnings
 from typing import List
 from typing import Optional

docs/source/asr/reference.md

@@ -1,8 +0,0 @@
-# Reference
-
-We refer these repos to build `model` and `engine`:
-
-* [delta](https://github.com/Delta-ML/delta.git)
-* [espnet](https://github.com/espnet/espnet.git)
-* [kaldi](https://github.com/kaldi-asr/kaldi.git)
-* [wenet](https://github.com/mobvoi/wenet)

docs/source/reference.md

@@ -0,0 +1,37 @@
+# Reference
+
+We borrowed a lot of code from these repos to build `model` and `engine`, thank for these great work:
+
+* [espnet](https://github.com/espnet/espnet/blob/master/LICENSE)
+- Apache-2.0 License
+- python/shell `utils`
+- kaldi feat preprocessing
+- datapipeline and `transform`
+- a lot of tts model, like `fastspeech2` and GAN-based `vocoder`
+
+* [wenet](https://github.com/wenet-e2e/wenet/blob/main/LICENSE)
+- Apache-2.0 License
+- U2 model
+- Building TLG based Graph
+
+* [kaldi](https://github.com/kaldi-asr/kaldi/blob/master/COPYING)
+- Apache-2.0 License
+- shell/perl/python utils.
+- feature bins.
+- WFST based decoding for LM integration.
+
+* [delta](https://github.com/Delta-ML/delta/blob/master/LICENSE)
+- Apache-2.0 License
+- `engine` arch
+
+* [speechbrain](https://github.com/speechbrain/speechbrain/blob/develop/LICENSE)
+- Apache-2.0 License
+- ECAPA-TDNN SV model
+
+* [chainer](https://github.com/chainer/chainer/blob/master/LICENSE)
+- MIT License
+- Updater, Trainer and more utils.
+
+* [librosa](https://github.com/librosa/librosa/blob/main/LICENSE.md)
+- ISC License
+- Audio feature

examples/aishell3/vc0/local/preprocess.sh

@@ -10,7 +10,7 @@ ge2e_ckpt_path=$4
 
 if [ ${stage} -le 0 ] && [ ${stop_stage} -ge 0 ]; then
     python3 ${BIN_DIR}/../../ge2e/inference.py \
-        --input=${input} \
+        --input=${input}/wav \
         --output=${preprocess_path}/embed \
         --checkpoint_path=${ge2e_ckpt_path}
 fi

paddlespeech/s2t/decoders/README.md

@@ -1,4 +1,5 @@
 # Decoders
+we borrow a lot of code from Espnet Apache 2.0  (http://www.apache.org/licenses/LICENSE-2.0)
 
 ## Reference
 ### CTC Prefix Beam Search

paddlespeech/s2t/decoders/beam_search/batch_beam_search.py

@@ -11,6 +11,7 @@
 # WITHOUT WARRANTIES OR CONDITIONS OF ANY KIND, either express or implied.
 # See the License for the specific language governing permissions and
 # limitations under the License.
+# Reference espnet Apache 2.0 (http://www.apache.org/licenses/LICENSE-2.0)
 
 
 class BatchBeamSearch():

paddlespeech/s2t/decoders/beam_search/beam_search.py

@@ -11,6 +11,7 @@
 # WITHOUT WARRANTIES OR CONDITIONS OF ANY KIND, either express or implied.
 # See the License for the specific language governing permissions and
 # limitations under the License.
+# Modified from espnet(https://github.com/espnet/espnet)
 """Beam search module."""
 from itertools import chain
 from typing import Any

paddlespeech/s2t/decoders/recog.py

@@ -11,6 +11,7 @@
 # WITHOUT WARRANTIES OR CONDITIONS OF ANY KIND, either express or implied.
 # See the License for the specific language governing permissions and
 # limitations under the License.
+# Reference espnet Apache 2.0 (http://www.apache.org/licenses/LICENSE-2.0)
 """V2 backend for `asr_recog.py` using py:class:`decoders.beam_search.BeamSearch`."""
 import jsonlines
 import paddle

paddlespeech/s2t/decoders/recog_bin.py

@@ -11,6 +11,7 @@
 # WITHOUT WARRANTIES OR CONDITIONS OF ANY KIND, either express or implied.
 # See the License for the specific language governing permissions and
 # limitations under the License.
+# Reference espnet Apache 2.0 (http://www.apache.org/licenses/LICENSE-2.0)
 """End-to-end speech recognition model decoding script."""
 import logging
 import os

paddlespeech/s2t/decoders/scorers/ctc.py

@@ -11,6 +11,7 @@
 # WITHOUT WARRANTIES OR CONDITIONS OF ANY KIND, either express or implied.
 # See the License for the specific language governing permissions and
 # limitations under the License.
+# Modified from espnet(https://github.com/espnet/espnet)
 """ScorerInterface implementation for CTC."""
 import numpy as np
 import paddle

paddlespeech/s2t/decoders/scorers/length_bonus.py

@@ -11,6 +11,7 @@
 # WITHOUT WARRANTIES OR CONDITIONS OF ANY KIND, either express or implied.
 # See the License for the specific language governing permissions and
 # limitations under the License.
+# Modified from espnet(https://github.com/espnet/espnet)
 """Length bonus module."""
 from typing import Any
 from typing import List

paddlespeech/s2t/decoders/scorers/ngram.py

@@ -11,6 +11,7 @@
 # WITHOUT WARRANTIES OR CONDITIONS OF ANY KIND, either express or implied.
 # See the License for the specific language governing permissions and
 # limitations under the License.
+# Modified from espnet(https://github.com/espnet/espnet)
 """Ngram lm implement."""
 from abc import ABC
 

paddlespeech/s2t/decoders/scorers/scorer_interface.py

@@ -11,6 +11,7 @@
 # WITHOUT WARRANTIES OR CONDITIONS OF ANY KIND, either express or implied.
 # See the License for the specific language governing permissions and
 # limitations under the License.
+# Modified from espnet(https://github.com/espnet/espnet)
 """Scorer interface module."""
 import warnings
 from typing import Any

paddlespeech/s2t/decoders/utils.py

@@ -11,6 +11,7 @@
 # WITHOUT WARRANTIES OR CONDITIONS OF ANY KIND, either express or implied.
 # See the License for the specific language governing permissions and
 # limitations under the License.
+# Modified from espnet(https://github.com/espnet/espnet)
 import numpy as np
 
 from paddlespeech.s2t.utils.log import Log

paddlespeech/s2t/io/batchfy.py

@@ -11,6 +11,7 @@
 # WITHOUT WARRANTIES OR CONDITIONS OF ANY KIND, either express or implied.
 # See the License for the specific language governing permissions and
 # limitations under the License.
+# Modified from espnet(https://github.com/espnet/espnet)
 import itertools
 
 import numpy as np

paddlespeech/s2t/io/converter.py

@@ -11,6 +11,7 @@
 # WITHOUT WARRANTIES OR CONDITIONS OF ANY KIND, either express or implied.
 # See the License for the specific language governing permissions and
 # limitations under the License.
+# Modified from espnet(https://github.com/espnet/espnet)
 import numpy as np
 
 from paddlespeech.s2t.io.utility import pad_list

paddlespeech/s2t/io/dataset.py

@@ -11,6 +11,8 @@
 # WITHOUT WARRANTIES OR CONDITIONS OF ANY KIND, either express or implied.
 # See the License for the specific language governing permissions and
 # limitations under the License.
+# Modified from espnet(https://github.com/espnet/espnet)
+# Modified from wenet(https://github.com/wenet-e2e/wenet)
 from typing import Optional
 
 from paddle.io import Dataset

paddlespeech/s2t/io/reader.py

@@ -11,6 +11,7 @@
 # WITHOUT WARRANTIES OR CONDITIONS OF ANY KIND, either express or implied.
 # See the License for the specific language governing permissions and
 # limitations under the License.
+# Modified from espnet(https://github.com/espnet/espnet)
 from collections import OrderedDict
 
 import kaldiio

paddlespeech/s2t/models/asr_interface.py

@@ -11,6 +11,7 @@
 # WITHOUT WARRANTIES OR CONDITIONS OF ANY KIND, either express or implied.
 # See the License for the specific language governing permissions and
 # limitations under the License.
+# Modified from espnet(https://github.com/espnet/espnet)
 """ASR Interface module."""
 import argparse
 

paddlespeech/s2t/models/lm/transformer.py

@@ -11,6 +11,7 @@
 # WITHOUT WARRANTIES OR CONDITIONS OF ANY KIND, either express or implied.
 # See the License for the specific language governing permissions and
 # limitations under the License.
+# Modified from espnet(https://github.com/espnet/espnet)
 from typing import Any
 from typing import List
 from typing import Tuple

paddlespeech/s2t/models/lm_interface.py

@@ -11,6 +11,7 @@
 # WITHOUT WARRANTIES OR CONDITIONS OF ANY KIND, either express or implied.
 # See the License for the specific language governing permissions and
 # limitations under the License.
+# Modified from espnet(https://github.com/espnet/espnet)
 """Language model interface."""
 import argparse
 

paddlespeech/s2t/models/st_interface.py

@@ -11,6 +11,7 @@
 # WITHOUT WARRANTIES OR CONDITIONS OF ANY KIND, either express or implied.
 # See the License for the specific language governing permissions and
 # limitations under the License.
+# Modified from espnet(https://github.com/espnet/espnet)
 """ST Interface module."""
 from .asr_interface import ASRInterface
 from paddlespeech.s2t.utils.dynamic_import import dynamic_import

paddlespeech/s2t/modules/attention.py

@@ -11,6 +11,7 @@
 # WITHOUT WARRANTIES OR CONDITIONS OF ANY KIND, either express or implied.
 # See the License for the specific language governing permissions and
 # limitations under the License.
+# Modified from wenet(https://github.com/wenet-e2e/wenet)
 """Multi-Head Attention layer definition."""
 import math
 from typing import Optional

paddlespeech/s2t/modules/cmvn.py

@@ -11,6 +11,7 @@
 # WITHOUT WARRANTIES OR CONDITIONS OF ANY KIND, either express or implied.
 # See the License for the specific language governing permissions and
 # limitations under the License.
+# Modified from wenet(https://github.com/wenet-e2e/wenet)
 import paddle
 from paddle import nn
 

paddlespeech/s2t/modules/conformer_convolution.py

@@ -11,6 +11,7 @@
 # WITHOUT WARRANTIES OR CONDITIONS OF ANY KIND, either express or implied.
 # See the License for the specific language governing permissions and
 # limitations under the License.
+# Modified from wenet(https://github.com/wenet-e2e/wenet)
 """ConvolutionModule definition."""
 from typing import Optional
 from typing import Tuple

paddlespeech/s2t/modules/decoder.py

@@ -11,6 +11,7 @@
 # WITHOUT WARRANTIES OR CONDITIONS OF ANY KIND, either express or implied.
 # See the License for the specific language governing permissions and
 # limitations under the License.
+# Modified from wenet(https://github.com/wenet-e2e/wenet)
 """Decoder definition."""
 from typing import Any
 from typing import List

paddlespeech/s2t/modules/decoder_layer.py

@@ -11,6 +11,7 @@
 # WITHOUT WARRANTIES OR CONDITIONS OF ANY KIND, either express or implied.
 # See the License for the specific language governing permissions and
 # limitations under the License.
+# Modified from wenet(https://github.com/wenet-e2e/wenet)
 """Decoder self-attention layer definition."""
 from typing import Optional
 from typing import Tuple

paddlespeech/s2t/modules/embedding.py

@@ -11,6 +11,7 @@
 # WITHOUT WARRANTIES OR CONDITIONS OF ANY KIND, either express or implied.
 # See the License for the specific language governing permissions and
 # limitations under the License.
+# Modified from wenet(https://github.com/wenet-e2e/wenet)
 """Positonal Encoding Module."""
 import math
 from typing import Tuple

paddlespeech/s2t/modules/encoder.py

@@ -11,6 +11,7 @@
 # WITHOUT WARRANTIES OR CONDITIONS OF ANY KIND, either express or implied.
 # See the License for the specific language governing permissions and
 # limitations under the License.
+# Modified from wenet(https://github.com/wenet-e2e/wenet)
 """Encoder definition."""
 from typing import List
 from typing import Optional

paddlespeech/s2t/modules/encoder_layer.py

@@ -11,6 +11,7 @@
 # WITHOUT WARRANTIES OR CONDITIONS OF ANY KIND, either express or implied.
 # See the License for the specific language governing permissions and
 # limitations under the License.
+# Modified from wenet(https://github.com/wenet-e2e/wenet)
 """Encoder self-attention layer definition."""
 from typing import Optional
 from typing import Tuple

paddlespeech/s2t/modules/loss.py

@@ -11,6 +11,7 @@
 # WITHOUT WARRANTIES OR CONDITIONS OF ANY KIND, either express or implied.
 # See the License for the specific language governing permissions and
 # limitations under the License.
+# Modified from wenet(https://github.com/wenet-e2e/wenet)
 import inspect
 
 import paddle

paddlespeech/s2t/modules/mask.py

@@ -11,6 +11,7 @@
 # WITHOUT WARRANTIES OR CONDITIONS OF ANY KIND, either express or implied.
 # See the License for the specific language governing permissions and
 # limitations under the License.
+# Modified from wenet(https://github.com/wenet-e2e/wenet)
 import paddle
 
 from paddlespeech.s2t.utils.log import Log

paddlespeech/s2t/modules/positionwise_feed_forward.py

@@ -11,6 +11,7 @@
 # WITHOUT WARRANTIES OR CONDITIONS OF ANY KIND, either express or implied.
 # See the License for the specific language governing permissions and
 # limitations under the License.
+# Modified from wenet(https://github.com/wenet-e2e/wenet)
 """Positionwise feed forward layer definition."""
 import paddle
 from paddle import nn

paddlespeech/s2t/modules/subsampling.py

@@ -11,6 +11,7 @@
 # WITHOUT WARRANTIES OR CONDITIONS OF ANY KIND, either express or implied.
 # See the License for the specific language governing permissions and
 # limitations under the License.
+# Modified from wenet(https://github.com/wenet-e2e/wenet)
 """Subsampling layer definition."""
 from typing import Tuple
 

paddlespeech/s2t/training/extensions/evaluator.py

@@ -11,6 +11,7 @@
 # WITHOUT WARRANTIES OR CONDITIONS OF ANY KIND, either express or implied.
 # See the License for the specific language governing permissions and
 # limitations under the License.
+# Modified from chainer(https://github.com/chainer/chainer)
 from typing import Dict
 
 import paddle

paddlespeech/s2t/training/extensions/extension.py

@@ -11,6 +11,7 @@
 # WITHOUT WARRANTIES OR CONDITIONS OF ANY KIND, either express or implied.
 # See the License for the specific language governing permissions and
 # limitations under the License.
+# Modified from chainer(https://github.com/chainer/chainer)
 PRIORITY_WRITER = 300
 PRIORITY_EDITOR = 200
 PRIORITY_READER = 100

paddlespeech/s2t/training/extensions/plot.py

@@ -11,6 +11,7 @@
 # WITHOUT WARRANTIES OR CONDITIONS OF ANY KIND, either express or implied.
 # See the License for the specific language governing permissions and
 # limitations under the License.
+# Modified from chainer(https://github.com/chainer/chainer)
 import copy
 import os
 

paddlespeech/s2t/training/extensions/snapshot.py

@@ -11,6 +11,7 @@
 # WITHOUT WARRANTIES OR CONDITIONS OF ANY KIND, either express or implied.
 # See the License for the specific language governing permissions and
 # limitations under the License.
+# Modified from chainer(https://github.com/chainer/chainer)
 import os
 from datetime import datetime
 from pathlib import Path

paddlespeech/s2t/training/optimizer.py

@@ -11,6 +11,7 @@
 # WITHOUT WARRANTIES OR CONDITIONS OF ANY KIND, either express or implied.
 # See the License for the specific language governing permissions and
 # limitations under the License.
+# Modified from espnet(https://github.com/espnet/espnet)
 from typing import Any
 from typing import Dict
 from typing import Text

paddlespeech/s2t/training/reporter.py

@@ -11,6 +11,7 @@
 # WITHOUT WARRANTIES OR CONDITIONS OF ANY KIND, either express or implied.
 # See the License for the specific language governing permissions and
 # limitations under the License.
+# Modified from chainer(https://github.com/chainer/chainer)
 import contextlib
 import math
 from collections import defaultdict

paddlespeech/s2t/training/scheduler.py

@@ -11,6 +11,7 @@
 # WITHOUT WARRANTIES OR CONDITIONS OF ANY KIND, either express or implied.
 # See the License for the specific language governing permissions and
 # limitations under the License.
+# Modified from espnet(https://github.com/espnet/espnet)
 from typing import Any
 from typing import Dict
 from typing import Text

paddlespeech/s2t/training/triggers/compare_value_trigger.py

@@ -11,6 +11,7 @@
 # WITHOUT WARRANTIES OR CONDITIONS OF ANY KIND, either express or implied.
 # See the License for the specific language governing permissions and
 # limitations under the License.
+# Modified from chainer(https://github.com/chainer/chainer)
 from ..reporter import DictSummary
 from .utils import get_trigger
 

paddlespeech/s2t/training/triggers/interval_trigger.py

@@ -11,6 +11,7 @@
 # WITHOUT WARRANTIES OR CONDITIONS OF ANY KIND, either express or implied.
 # See the License for the specific language governing permissions and
 # limitations under the License.
+# Reference chainer MIT (https://opensource.org/licenses/MIT)
 
 
 class IntervalTrigger():

paddlespeech/s2t/training/triggers/limit_trigger.py

@@ -11,6 +11,7 @@
 # WITHOUT WARRANTIES OR CONDITIONS OF ANY KIND, either express or implied.
 # See the License for the specific language governing permissions and
 # limitations under the License.
+# Reference chainer MIT (https://opensource.org/licenses/MIT)
 
 
 class LimitTrigger():

paddlespeech/s2t/training/triggers/time_trigger.py

@@ -11,6 +11,7 @@
 # WITHOUT WARRANTIES OR CONDITIONS OF ANY KIND, either express or implied.
 # See the License for the specific language governing permissions and
 # limitations under the License.
+# Reference chainer MIT (https://opensource.org/licenses/MIT)
 
 
 class TimeTrigger():

paddlespeech/s2t/training/updaters/standard_updater.py

@@ -11,6 +11,7 @@
 # WITHOUT WARRANTIES OR CONDITIONS OF ANY KIND, either express or implied.
 # See the License for the specific language governing permissions and
 # limitations under the License.
+# Modified from chainer(https://github.com/chainer/chainer)
 from typing import Dict
 from typing import Optional
 

paddlespeech/s2t/training/updaters/trainer.py

@@ -11,6 +11,7 @@
 # WITHOUT WARRANTIES OR CONDITIONS OF ANY KIND, either express or implied.
 # See the License for the specific language governing permissions and
 # limitations under the License.
+# Modified from chainer(https://github.com/chainer/chainer)
 import sys
 import traceback
 from collections import OrderedDict

paddlespeech/s2t/training/updaters/updater.py

@@ -11,6 +11,7 @@
 # WITHOUT WARRANTIES OR CONDITIONS OF ANY KIND, either express or implied.
 # See the License for the specific language governing permissions and
 # limitations under the License.
+# Modified from chainer(https://github.com/chainer/chainer)
 from dataclasses import dataclass
 
 import paddle

paddlespeech/s2t/transform/add_deltas.py

@@ -11,6 +11,7 @@
 # WITHOUT WARRANTIES OR CONDITIONS OF ANY KIND, either express or implied.
 # See the License for the specific language governing permissions and
 # limitations under the License.
+# Modified from espnet(https://github.com/espnet/espnet)
 import numpy as np
 
 

paddlespeech/s2t/transform/channel_selector.py

@@ -11,6 +11,7 @@
 # WITHOUT WARRANTIES OR CONDITIONS OF ANY KIND, either express or implied.
 # See the License for the specific language governing permissions and
 # limitations under the License.
+# Modified from espnet(https://github.com/espnet/espnet)
 import numpy
 
 

paddlespeech/s2t/transform/cmvn.py

@@ -11,6 +11,7 @@
 # WITHOUT WARRANTIES OR CONDITIONS OF ANY KIND, either express or implied.
 # See the License for the specific language governing permissions and
 # limitations under the License.
+# Modified from espnet(https://github.com/espnet/espnet)
 import io
 
 import h5py

paddlespeech/s2t/transform/functional.py

@@ -11,6 +11,7 @@
 # WITHOUT WARRANTIES OR CONDITIONS OF ANY KIND, either express or implied.
 # See the License for the specific language governing permissions and
 # limitations under the License.
+# Modified from espnet(https://github.com/espnet/espnet)
 import inspect
 
 from paddlespeech.s2t.transform.transform_interface import TransformInterface

paddlespeech/s2t/transform/perturb.py

@@ -11,6 +11,7 @@
 # WITHOUT WARRANTIES OR CONDITIONS OF ANY KIND, either express or implied.
 # See the License for the specific language governing permissions and
 # limitations under the License.
+# Modified from espnet(https://github.com/espnet/espnet)
 import librosa
 import numpy
 import scipy

paddlespeech/s2t/transform/spec_augment.py

@@ -11,6 +11,7 @@
 # WITHOUT WARRANTIES OR CONDITIONS OF ANY KIND, either express or implied.
 # See the License for the specific language governing permissions and
 # limitations under the License.
+# Modified from espnet(https://github.com/espnet/espnet)
 """Spec Augment module for preprocessing i.e., data augmentation"""
 import random
 

paddlespeech/s2t/transform/spectrogram.py

@@ -11,6 +11,7 @@
 # WITHOUT WARRANTIES OR CONDITIONS OF ANY KIND, either express or implied.
 # See the License for the specific language governing permissions and
 # limitations under the License.
+# Modified from espnet(https://github.com/espnet/espnet)
 import librosa
 import numpy as np
 

paddlespeech/s2t/transform/transform_interface.py

@@ -11,7 +11,9 @@
 # WITHOUT WARRANTIES OR CONDITIONS OF ANY KIND, either express or implied.
 # See the License for the specific language governing permissions and
 # limitations under the License.
-# TODO(karita): add this to all the transform impl.
+# Modified from espnet(https://github.com/espnet/espnet)
+
+
 class TransformInterface:
     """Transform Interface"""
 

paddlespeech/s2t/transform/transformation.py

@@ -11,6 +11,7 @@
 # WITHOUT WARRANTIES OR CONDITIONS OF ANY KIND, either express or implied.
 # See the License for the specific language governing permissions and
 # limitations under the License.
+# Modified from espnet(https://github.com/espnet/espnet)
 """Transformation module."""
 import copy
 import io
@@ -23,8 +24,6 @@ import yaml
 
 from paddlespeech.s2t.utils.dynamic_import import dynamic_import
 
-# TODO(karita): inherit TransformInterface
-# TODO(karita): register cmd arguments in asr_train.py
 import_alias = dict(
     identity="paddlespeech.s2t.transform.transform_interface:Identity",
     time_warp="paddlespeech.s2t.transform.spec_augment:TimeWarp",

paddlespeech/s2t/transform/wpe.py

@@ -11,6 +11,7 @@
 # WITHOUT WARRANTIES OR CONDITIONS OF ANY KIND, either express or implied.
 # See the License for the specific language governing permissions and
 # limitations under the License.
+# Modified from espnet(https://github.com/espnet/espnet)
 from nara_wpe.wpe import wpe
 
 

paddlespeech/s2t/utils/asr_utils.py

@@ -11,6 +11,7 @@
 # WITHOUT WARRANTIES OR CONDITIONS OF ANY KIND, either express or implied.
 # See the License for the specific language governing permissions and
 # limitations under the License.
+# Reference espnet Apache 2.0 (http://www.apache.org/licenses/LICENSE-2.0)
 import json
 
 import numpy as np
@@ -18,7 +19,6 @@ import numpy as np
 __all__ = ["label_smoothing_dist"]
 
 
-# TODO(takaaki-hori): add different smoothing methods
 def label_smoothing_dist(odim, lsm_type, transcript=None, blank=0):
     """Obtain label distribution for loss smoothing.
 

paddlespeech/s2t/utils/check_kwargs.py

@@ -11,6 +11,7 @@
 # WITHOUT WARRANTIES OR CONDITIONS OF ANY KIND, either express or implied.
 # See the License for the specific language governing permissions and
 # limitations under the License.
+# Modified from espnet(https://github.com/espnet/espnet)
 import inspect
 
 

paddlespeech/s2t/utils/cli_readers.py

@@ -11,6 +11,7 @@
 # WITHOUT WARRANTIES OR CONDITIONS OF ANY KIND, either express or implied.
 # See the License for the specific language governing permissions and
 # limitations under the License.
+# Modified from espnet(https://github.com/espnet/espnet)
 import io
 import logging
 import sys

paddlespeech/s2t/utils/cli_utils.py

@@ -11,6 +11,7 @@
 # WITHOUT WARRANTIES OR CONDITIONS OF ANY KIND, either express or implied.
 # See the License for the specific language governing permissions and
 # limitations under the License.
+# Modified from espnet(https://github.com/espnet/espnet)
 import sys
 from collections.abc import Sequence
 from distutils.util import strtobool as dist_strtobool

paddlespeech/s2t/utils/cli_writers.py

@@ -11,6 +11,7 @@
 # WITHOUT WARRANTIES OR CONDITIONS OF ANY KIND, either express or implied.
 # See the License for the specific language governing permissions and
 # limitations under the License.
+# Modified from espnet(https://github.com/espnet/espnet)
 from pathlib import Path
 from typing import Dict
 

paddlespeech/s2t/utils/ctc_utils.py

@@ -11,6 +11,7 @@
 # WITHOUT WARRANTIES OR CONDITIONS OF ANY KIND, either express or implied.
 # See the License for the specific language governing permissions and
 # limitations under the License.
+# Modified from wenet(https://github.com/wenet-e2e/wenet)
 from typing import List
 
 import numpy as np
@@ -138,16 +139,6 @@ def forced_align(ctc_probs: paddle.Tensor, y: paddle.Tensor,
     return output_alignment
 
 
-# ctc_align(
-# self.model,
-# self.align_loader,
-# self.config.decoding.batch_size,
-# self.align_loader.collate_fn.stride_ms,
-# self.align_loader.collate_fn.vocab_list,
-# self.args.result_file,
-# )
-
-
 def ctc_align(model, dataloader, batch_size, stride_ms, token_dict,
               result_file):
     """ctc alignment.

paddlespeech/s2t/utils/dynamic_import.py

@@ -11,6 +11,7 @@
 # WITHOUT WARRANTIES OR CONDITIONS OF ANY KIND, either express or implied.
 # See the License for the specific language governing permissions and
 # limitations under the License.
+# Modified from espnet(https://github.com/espnet/espnet)
 import importlib
 import inspect
 from typing import Any

paddlespeech/t2s/data/batch.py

@@ -52,9 +52,8 @@ def batch_text_id(minibatch, pad_id=0, dtype=np.int64):
     """
     peek_example = minibatch[0]
     assert len(peek_example.shape) == 1, "text example is an 1D tensor"
-
+    # assume (channel, n_samples) or (n_samples, )
-    lengths = [example.shape[0] for example in
+    lengths = [example.shape[0] for example in minibatch]
-               minibatch]  # assume (channel, n_samples) or (n_samples, )
     max_len = np.max(lengths)
 
     batch = []

paddlespeech/t2s/data/get_feats.py

@@ -11,6 +11,7 @@
 # WITHOUT WARRANTIES OR CONDITIONS OF ANY KIND, either express or implied.
 # See the License for the specific language governing permissions and
 # limitations under the License.
+# Modified from espnet(https://github.com/espnet/espnet)
 import librosa
 import numpy as np
 import pyworld

paddlespeech/t2s/exps/voice_cloning/tacotron2_ge2e/train.py

@@ -241,7 +241,7 @@ def main_sp(config, args):
 
 
 def main(config, args):
-    if args.gpus:
+    if args.ngpu:
         dist.spawn(main_sp, args=(config, args), nprocs=args.ngpu)
     else:
         main_sp(config, args)

paddlespeech/t2s/models/fastspeech2/fastspeech2.py

@@ -11,6 +11,7 @@
 # WITHOUT WARRANTIES OR CONDITIONS OF ANY KIND, either express or implied.
 # See the License for the specific language governing permissions and
 # limitations under the License.
+# Modified from espnet(https://github.com/espnet/espnet)
 """Fastspeech2 related modules for paddle"""
 from typing import Dict
 from typing import Sequence

paddlespeech/t2s/models/melgan/melgan.py

@@ -11,6 +11,7 @@
 # WITHOUT WARRANTIES OR CONDITIONS OF ANY KIND, either express or implied.
 # See the License for the specific language governing permissions and
 # limitations under the License.
+# Modified from espnet(https://github.com/espnet/espnet)
 """MelGAN Modules."""
 from typing import Any
 from typing import Dict

paddlespeech/t2s/models/parallel_wavegan/parallel_wavegan.py

@@ -11,6 +11,7 @@
 # WITHOUT WARRANTIES OR CONDITIONS OF ANY KIND, either express or implied.
 # See the License for the specific language governing permissions and
 # limitations under the License.
+# Modified from espnet(https://github.com/espnet/espnet)
 import math
 from typing import Any
 from typing import Dict

paddlespeech/t2s/models/transformer_tts/transformer_tts.py

@@ -11,6 +11,7 @@
 # WITHOUT WARRANTIES OR CONDITIONS OF ANY KIND, either express or implied.
 # See the License for the specific language governing permissions and
 # limitations under the License.
+# Modified from espnet(https://github.com/espnet/espnet)
 """Fastspeech2 related modules for paddle"""
 from typing import Dict
 from typing import Sequence

paddlespeech/t2s/modules/adversarial_loss.py

@@ -11,6 +11,7 @@
 # WITHOUT WARRANTIES OR CONDITIONS OF ANY KIND, either express or implied.
 # See the License for the specific language governing permissions and
 # limitations under the License.
+# Modified from espnet(https://github.com/espnet/espnet)
 """Adversarial loss modules."""
 import paddle
 import paddle.nn.functional as F

paddlespeech/t2s/modules/fastspeech2_predictor/duration_predictor.py

@@ -11,6 +11,7 @@
 # WITHOUT WARRANTIES OR CONDITIONS OF ANY KIND, either express or implied.
 # See the License for the specific language governing permissions and
 # limitations under the License.
+# Modified from espnet(https://github.com/espnet/espnet)
 """Duration predictor related modules."""
 import paddle
 from paddle import nn

paddlespeech/t2s/modules/fastspeech2_predictor/length_regulator.py

@@ -11,6 +11,7 @@
 # WITHOUT WARRANTIES OR CONDITIONS OF ANY KIND, either express or implied.
 # See the License for the specific language governing permissions and
 # limitations under the License.
+# Modified from espnet(https://github.com/espnet/espnet)
 """Length regulator related modules."""
 import paddle
 from paddle import nn

paddlespeech/t2s/modules/fastspeech2_predictor/variance_predictor.py

@@ -11,6 +11,7 @@
 # WITHOUT WARRANTIES OR CONDITIONS OF ANY KIND, either express or implied.
 # See the License for the specific language governing permissions and
 # limitations under the License.
+# Modified from espnet(https://github.com/espnet/espnet)
 """Variance predictor related modules."""
 import paddle
 from paddle import nn

paddlespeech/t2s/modules/fastspeech2_transformer/attention.py

@@ -11,6 +11,7 @@
 # WITHOUT WARRANTIES OR CONDITIONS OF ANY KIND, either express or implied.
 # See the License for the specific language governing permissions and
 # limitations under the License.
+# Modified from espnet(https://github.com/espnet/espnet)
 """Multi-Head Attention layer definition."""
 import math
 

paddlespeech/t2s/modules/fastspeech2_transformer/decoder.py

@@ -11,6 +11,7 @@
 # WITHOUT WARRANTIES OR CONDITIONS OF ANY KIND, either express or implied.
 # See the License for the specific language governing permissions and
 # limitations under the License.
+# Modified from espnet(https://github.com/espnet/espnet)
 # 暂时删除了 dyminic conv
 """Decoder definition."""
 import logging

paddlespeech/t2s/modules/fastspeech2_transformer/decoder_layer.py

@@ -11,6 +11,7 @@
 # WITHOUT WARRANTIES OR CONDITIONS OF ANY KIND, either express or implied.
 # See the License for the specific language governing permissions and
 # limitations under the License.
+# Modified from espnet(https://github.com/espnet/espnet)
 """Decoder self-attention layer definition."""
 import paddle
 from paddle import nn

paddlespeech/t2s/modules/fastspeech2_transformer/embedding.py

@@ -11,6 +11,7 @@
 # WITHOUT WARRANTIES OR CONDITIONS OF ANY KIND, either express or implied.
 # See the License for the specific language governing permissions and
 # limitations under the License.
+# Modified from espnet(https://github.com/espnet/espnet)
 """Positional Encoding Module."""
 import math
 

paddlespeech/t2s/modules/fastspeech2_transformer/encoder.py

@@ -11,6 +11,7 @@
 # WITHOUT WARRANTIES OR CONDITIONS OF ANY KIND, either express or implied.
 # See the License for the specific language governing permissions and
 # limitations under the License.
+# Modified from espnet(https://github.com/espnet/espnet)
 import logging
 
 from paddle import nn

paddlespeech/t2s/modules/fastspeech2_transformer/encoder_layer.py

@@ -11,6 +11,7 @@
 # WITHOUT WARRANTIES OR CONDITIONS OF ANY KIND, either express or implied.
 # See the License for the specific language governing permissions and
 # limitations under the License.
+# Modified from espnet(https://github.com/espnet/espnet)
 """Encoder self-attention layer definition."""
 import paddle
 from paddle import nn

paddlespeech/t2s/modules/fastspeech2_transformer/lightconv.py

@@ -11,6 +11,7 @@
 # WITHOUT WARRANTIES OR CONDITIONS OF ANY KIND, either express or implied.
 # See the License for the specific language governing permissions and
 # limitations under the License.
+# Modified from espnet(https://github.com/espnet/espnet)
 """Lightweight Convolution Module."""
 import numpy
 import paddle

paddlespeech/t2s/modules/fastspeech2_transformer/positionwise_feed_forward.py

@@ -11,6 +11,7 @@
 # WITHOUT WARRANTIES OR CONDITIONS OF ANY KIND, either express or implied.
 # See the License for the specific language governing permissions and
 # limitations under the License.
+# Modified from espnet(https://github.com/espnet/espnet)
 """Positionwise feed forward layer definition."""
 import paddle
 

paddlespeech/t2s/modules/fastspeech2_transformer/repeat.py

@@ -11,6 +11,7 @@
 # WITHOUT WARRANTIES OR CONDITIONS OF ANY KIND, either express or implied.
 # See the License for the specific language governing permissions and
 # limitations under the License.
+# Modified from espnet(https://github.com/espnet/espnet)
 """Repeat the same layer definition."""
 import paddle
 

paddlespeech/t2s/modules/nets_utils.py

@@ -11,6 +11,7 @@
 # WITHOUT WARRANTIES OR CONDITIONS OF ANY KIND, either express or implied.
 # See the License for the specific language governing permissions and
 # limitations under the License.
+# Modified from espnet(https://github.com/espnet/espnet)
 import paddle
 from paddle import nn
 from typeguard import check_argument_types

paddlespeech/t2s/modules/pqmf.py

@@ -11,6 +11,7 @@
 # WITHOUT WARRANTIES OR CONDITIONS OF ANY KIND, either express or implied.
 # See the License for the specific language governing permissions and
 # limitations under the License.
+# Modified from espnet(https://github.com/espnet/espnet)
 """Pseudo QMF modules."""
 import numpy as np
 import paddle

paddlespeech/t2s/modules/residual_stack.py

@@ -11,6 +11,7 @@
 # WITHOUT WARRANTIES OR CONDITIONS OF ANY KIND, either express or implied.
 # See the License for the specific language governing permissions and
 # limitations under the License.
+# Modified from espnet(https://github.com/espnet/espnet)
 """Residual stack module in MelGAN."""
 from typing import Any
 from typing import Dict

paddlespeech/t2s/modules/stft_loss.py

@@ -11,6 +11,7 @@
 # WITHOUT WARRANTIES OR CONDITIONS OF ANY KIND, either express or implied.
 # See the License for the specific language governing permissions and
 # limitations under the License.
+# Modified from espnet(https://github.com/espnet/espnet)
 import paddle
 from paddle import nn
 from paddle.nn import functional as F

paddlespeech/t2s/modules/style_encoder.py

@@ -11,6 +11,7 @@
 # WITHOUT WARRANTIES OR CONDITIONS OF ANY KIND, either express or implied.
 # See the License for the specific language governing permissions and
 # limitations under the License.
+# Modified from espnet(https://github.com/espnet/espnet)
 """Style encoder of GST-Tacotron."""
 from typing import Sequence
 

paddlespeech/t2s/modules/tacotron2/decoder.py

@@ -11,6 +11,7 @@
 # WITHOUT WARRANTIES OR CONDITIONS OF ANY KIND, either express or implied.
 # See the License for the specific language governing permissions and
 # limitations under the License.
+# Modified from espnet(https://github.com/espnet/espnet)
 """Tacotron2 decoder related modules."""
 import paddle.nn.functional as F
 import six

paddlespeech/t2s/modules/tacotron2/encoder.py

@@ -11,6 +11,7 @@
 # WITHOUT WARRANTIES OR CONDITIONS OF ANY KIND, either express or implied.
 # See the License for the specific language governing permissions and
 # limitations under the License.
+# Modified from espnet(https://github.com/espnet/espnet)
 """Tacotron2 encoder related modules."""
 import paddle
 import six

paddlespeech/t2s/training/extension.py

@@ -11,6 +11,7 @@
 # WITHOUT WARRANTIES OR CONDITIONS OF ANY KIND, either express or implied.
 # See the License for the specific language governing permissions and
 # limitations under the License.
+# Modified from chainer(https://github.com/chainer/chainer)
 from typing import Callable
 
 PRIORITY_WRITER = 300

paddlespeech/t2s/training/extensions/evaluator.py

@@ -11,6 +11,7 @@
 # WITHOUT WARRANTIES OR CONDITIONS OF ANY KIND, either express or implied.
 # See the License for the specific language governing permissions and
 # limitations under the License.
+# Modified from chainer(https://github.com/chainer/chainer)
 from typing import Dict
 
 import paddle

paddlespeech/t2s/training/extensions/snapshot.py

@@ -11,6 +11,7 @@
 # WITHOUT WARRANTIES OR CONDITIONS OF ANY KIND, either express or implied.
 # See the License for the specific language governing permissions and
 # limitations under the License.
+# Modified from chainer(https://github.com/chainer/chainer)
 import logging
 import os
 from datetime import datetime

paddlespeech/t2s/training/reporter.py

@@ -11,6 +11,7 @@
 # WITHOUT WARRANTIES OR CONDITIONS OF ANY KIND, either express or implied.
 # See the License for the specific language governing permissions and
 # limitations under the License.
+# Modified from chainer(https://github.com/chainer/chainer)
 import contextlib
 import math
 from collections import defaultdict

paddlespeech/t2s/training/triggers/interval_trigger.py

@@ -11,6 +11,7 @@
 # WITHOUT WARRANTIES OR CONDITIONS OF ANY KIND, either express or implied.
 # See the License for the specific language governing permissions and
 # limitations under the License.
+# Reference chainer MIT (https://opensource.org/licenses/MIT)
 
 
 class IntervalTrigger(object):

paddlespeech/t2s/training/triggers/limit_trigger.py

@@ -11,6 +11,7 @@
 # WITHOUT WARRANTIES OR CONDITIONS OF ANY KIND, either express or implied.
 # See the License for the specific language governing permissions and
 # limitations under the License.
+# Reference chainer MIT (https://opensource.org/licenses/MIT)
 
 
 class LimitTrigger(object):