refactor: apply unsafe automatic ruff lint fixes

Manually checked and adjusted after generating with:
uv run ruff check tests/ TTS/ notebooks/ recipes/ --fix --unsafe-fixes

TTS/bin/train_encoder.py

@@ -296,7 +296,7 @@ def main(args):  # pylint: disable=redefined-outer-name
         criterion, args.restore_step = model.load_checkpoint(
             c, args.restore_path, eval=False, use_cuda=use_cuda, criterion=criterion
         )
-        print(" > Model restored from step %d" % args.restore_step, flush=True)
+        print(f" > Model restored from step {args.restore_step}", flush=True)
     else:
         args.restore_step = 0
 

TTS/config/__init__.py

@@ -1,7 +1,7 @@
 import json
 import os
 import re
-from typing import Any, Dict, Union
+from typing import Any, Union
 
 import fsspec
 import yaml
@@ -58,7 +58,7 @@ def _process_model_name(config_dict: dict) -> str:
     """Format the model name as expected. It is a band-aid for the old `vocoder` model names.
 
     Args:
-        config_dict (Dict): A dictionary including the config fields.
+        config_dict (dict): A dictionary including the config fields.
 
     Returns:
         str: Formatted modelname.

TTS/encoder/models/base_encoder.py

@@ -34,7 +34,7 @@ class BaseEncoder(nn.Module):
 
     # pylint: disable=W0102
     def __init__(self):
-        super(BaseEncoder, self).__init__()
+        super().__init__()
 
     def get_torch_mel_spectrogram_class(self, audio_config):
         return torch.nn.Sequential(
@@ -107,7 +107,7 @@ def get_criterion(self, c: Coqpit, num_classes=None):
         elif c.loss == "softmaxproto":
             criterion = SoftmaxAngleProtoLoss(c.model_params["proj_dim"], num_classes)
         else:
-            raise Exception("The %s  not is a loss supported" % c.loss)
+            raise Exception(f"The {c.loss}  not is a loss supported")
         return criterion
 
     def load_checkpoint(

TTS/encoder/models/resnet.py

@@ -7,7 +7,7 @@
 
 class SELayer(nn.Module):
     def __init__(self, channel, reduction=8):
-        super(SELayer, self).__init__()
+        super().__init__()
         self.avg_pool = nn.AdaptiveAvgPool2d(1)
         self.fc = nn.Sequential(
             nn.Linear(channel, channel // reduction),
@@ -27,7 +27,7 @@ class SEBasicBlock(nn.Module):
     expansion = 1
 
     def __init__(self, inplanes, planes, stride=1, downsample=None, reduction=8):
-        super(SEBasicBlock, self).__init__()
+        super().__init__()
         self.conv1 = nn.Conv2d(inplanes, planes, kernel_size=3, stride=stride, padding=1, bias=False)
         self.bn1 = nn.BatchNorm2d(planes)
         self.conv2 = nn.Conv2d(planes, planes, kernel_size=3, padding=1, bias=False)
@@ -73,7 +73,7 @@ def __init__(
         use_torch_spec=False,
         audio_config=None,
     ):
-        super(ResNetSpeakerEncoder, self).__init__()
+        super().__init__()
 
         self.encoder_type = encoder_type
         self.input_dim = input_dim

TTS/encoder/utils/prepare_voxceleb.py

@@ -82,30 +82,30 @@ def download_and_extract(directory, subset, urls):
                 continue
             logger.info("Downloading %s to %s", url, zip_filepath)
             subprocess.call(
-                "wget %s --user %s --password %s -O %s" % (url, USER["user"], USER["password"], zip_filepath),
+                "wget {} --user {} --password {} -O {}".format(url, USER["user"], USER["password"], zip_filepath),
                 shell=True,
             )
 
             statinfo = os.stat(zip_filepath)
-            logger.info("Successfully downloaded %s, size(bytes): %d" % (url, statinfo.st_size))
+            logger.info("Successfully downloaded %s, size(bytes): %d", url, statinfo.st_size)
 
         # concatenate all parts into zip files
         if ".zip" not in zip_filepath:
             zip_filepath = "_".join(zip_filepath.split("_")[:-1])
-            subprocess.call("cat %s* > %s.zip" % (zip_filepath, zip_filepath), shell=True)
+            subprocess.call(f"cat {zip_filepath}* > {zip_filepath}.zip", shell=True)
             zip_filepath += ".zip"
         extract_path = zip_filepath.strip(".zip")
 
         # check zip file md5sum
         with open(zip_filepath, "rb") as f_zip:
             md5 = hashlib.md5(f_zip.read()).hexdigest()
         if md5 != MD5SUM[subset]:
-            raise ValueError("md5sum of %s mismatch" % zip_filepath)
+            raise ValueError(f"md5sum of {zip_filepath} mismatch")
 
         with zipfile.ZipFile(zip_filepath, "r") as zfile:
             zfile.extractall(directory)
             extract_path_ori = os.path.join(directory, zfile.infolist()[0].filename)
-            subprocess.call("mv %s %s" % (extract_path_ori, extract_path), shell=True)
+            subprocess.call(f"mv {extract_path_ori} {extract_path}", shell=True)
     finally:
         # os.remove(zip_filepath)
         pass
@@ -193,7 +193,7 @@ def convert_audio_and_make_label(input_dir, subset, output_dir, output_file):
         writer.writerow(["wav_filename", "wav_length_ms", "speaker_id", "speaker_name"])
         for wav_file in files:
             writer.writerow(wav_file)
-    logger.info(f"Successfully generated csv file {csv_file_path}")
+    logger.info("Successfully generated csv file %s", csv_file_path)
 
 
 def processor(directory, subset, force_process):

TTS/tts/datasets/__init__.py

@@ -4,7 +4,6 @@
 from collections import Counter
 from collections.abc import Callable
 from pathlib import Path
-from typing import Dict, List, Tuple, Union
 
 import numpy as np
 
@@ -18,7 +17,7 @@ def split_dataset(items, eval_split_max_size=None, eval_split_size=0.01):
     """Split a dataset into train and eval. Consider speaker distribution in multi-speaker training.
 
     Args:
-        items (List[List]):
+        items (list[list]):
             A list of samples. Each sample is a list of `[audio_path, text, speaker_id]`.
 
         eval_split_max_size (int):
@@ -76,12 +75,12 @@ def load_tts_samples(
     eval_split_max_size=None,
     eval_split_size=0.01,
 ) -> tuple[list[list], list[list]]:
-    """Parse the dataset from the datasets config, load the samples as a List and load the attention alignments if provided.
+    """Parse the dataset from the datasets config, load the samples as a list and load the attention alignments if provided.
     If `formatter` is not None, apply the formatter to the samples else pick the formatter from the available ones based
     on the dataset name.
 
     Args:
-        datasets (List[Dict], Dict): A list of datasets or a single dataset dictionary. If multiple datasets are
+        datasets (list[dict], dict): A list of datasets or a single dataset dictionary. If multiple datasets are
             in the list, they are all merged.
 
         eval_split (bool, optional): If true, create a evaluation split. If an eval split provided explicitly, generate
@@ -100,7 +99,7 @@ def load_tts_samples(
             If > 1, represents the absolute number of evaluation samples. Defaults to 0.01 (1%).
 
     Returns:
-        Tuple[List[List], List[List]: training and evaluation splits of the dataset.
+        tuple[list[list], list[list]: training and evaluation splits of the dataset.
     """
     meta_data_train_all = []
     meta_data_eval_all = [] if eval_split else None

TTS/tts/layers/delightful_tts/conv_layers.py

@@ -50,7 +50,7 @@ def __init__(
         w_init_gain="linear",
         use_weight_norm=False,
     ):
-        super(ConvNorm, self).__init__()  # pylint: disable=super-with-arguments
+        super().__init__()
         if padding is None:
             assert kernel_size % 2 == 1
             padding = int(dilation * (kernel_size - 1) / 2)
@@ -92,7 +92,7 @@ def __init__(
         lstm_type="bilstm",
         use_linear=True,
     ):
-        super(ConvLSTMLinear, self).__init__()  # pylint: disable=super-with-arguments
+        super().__init__()
         self.out_dim = out_dim
         self.lstm_type = lstm_type
         self.use_linear = use_linear

TTS/tts/layers/delightful_tts/networks.py

@@ -51,7 +51,7 @@ def __init__(
         kernel_size=3,
         use_partial_padding=False,  # pylint: disable=unused-argument
     ):
-        super(BottleneckLayer, self).__init__()  # pylint: disable=super-with-arguments
+        super().__init__()
 
         self.reduction_factor = reduction_factor
         reduced_dim = int(in_dim / reduction_factor)
@@ -194,7 +194,7 @@ class STL(nn.Module):
     """
 
     def __init__(self, n_hidden: int, token_num: int):
-        super(STL, self).__init__()  # pylint: disable=super-with-arguments
+        super().__init__()
 
         num_heads = 1
         E = n_hidden

TTS/tts/layers/tortoise/diffusion.py

@@ -653,7 +653,7 @@ def p_sample_loop_progressive(
         """
         if device is None:
             device = next(model.parameters()).device
-        assert isinstance(shape, (tuple, list))
+        assert isinstance(shape, tuple | list)
         if noise is not None:
             img = noise
         else:
@@ -805,7 +805,7 @@ def ddim_sample_loop_progressive(
         """
         if device is None:
             device = next(model.parameters()).device
-        assert isinstance(shape, (tuple, list))
+        assert isinstance(shape, tuple | list)
         if noise is not None:
             img = noise
         else:

TTS/tts/layers/tortoise/vocoder.py

@@ -293,7 +293,7 @@ def __init__(
         hop_length=256,
         n_mel_channels=100,
     ):
-        super(UnivNetGenerator, self).__init__()
+        super().__init__()
         self.mel_channel = n_mel_channels
         self.noise_dim = noise_dim
         self.hop_length = hop_length
@@ -344,7 +344,7 @@ def forward(self, c, z):
         return z
 
     def eval(self, inference=False):
-        super(UnivNetGenerator, self).eval()
+        super().eval()
         # don't remove weight norm while validation in training loop
         if inference:
             self.remove_weight_norm()