fix packages

.gitignore

@@ -0,0 +1,2 @@
+venv/
+__pycache__

indextts/BigVGAN/ECAPA_TDNN.py

@@ -12,6 +12,7 @@ from indextts.BigVGAN.nnet.CNN import Conv1d as _Conv1d
 from indextts.BigVGAN.nnet.linear import Linear
 from indextts.BigVGAN.nnet.normalization import BatchNorm1d as _BatchNorm1d
 
+
 def length_to_mask(length, max_len=None, dtype=None, device=None):
     """Creates a binary mask for each sequence.
 

indextts/BigVGAN/activations.py

@@ -2,7 +2,7 @@
 #   LICENSE is in incl_licenses directory.
 
 import torch
-from torch import nn, sin, pow
+from torch import nn, pow, sin
 from torch.nn import Parameter
 
 
@@ -22,6 +22,7 @@ class Snake(nn.Module):
         >>> x = torch.randn(256)
         >>> x = a1(x)
     '''
+
     def __init__(self, in_features, alpha=1.0, alpha_trainable=True, alpha_logscale=False):
         '''
         Initialization.
@@ -76,6 +77,7 @@ class SnakeBeta(nn.Module):
         >>> x = torch.randn(256)
         >>> x = a1(x)
     '''
+
     def __init__(self, in_features, alpha=1.0, alpha_trainable=True, alpha_logscale=False):
         '''
         Initialization.

indextts/BigVGAN/alias_free_activation/cuda/activation1d.py

@@ -3,10 +3,9 @@
 
 import torch
 import torch.nn as nn
-from alias_free_activation.torch.resample import UpSample1d, DownSample1d
-
 # load fused CUDA kernel: this enables importing anti_alias_activation_cuda
 from alias_free_activation.cuda import load
+from alias_free_activation.torch.resample import DownSample1d, UpSample1d
 
 anti_alias_activation_cuda = load.load()
 

indextts/BigVGAN/alias_free_activation/torch/__init__.py

@@ -1,6 +1,6 @@
 # Adapted from https://github.com/junjun3518/alias-free-torch under the Apache License 2.0
 #   LICENSE is in incl_licenses directory.
 
+from .act import *
 from .filter import *
 from .resample import *
-from .act import *

indextts/BigVGAN/alias_free_activation/torch/act.py

@@ -2,7 +2,8 @@
 #   LICENSE is in incl_licenses directory.
 
 import torch.nn as nn
-from .resample import UpSample1d, DownSample1d
+
+from .resample import DownSample1d, UpSample1d
 
 
 class Activation1d(nn.Module):

indextts/BigVGAN/alias_free_activation/torch/filter.py

@@ -1,10 +1,11 @@
 # Adapted from https://github.com/junjun3518/alias-free-torch under the Apache License 2.0
 #   LICENSE is in incl_licenses directory.
 
+import math
+
 import torch
 import torch.nn as nn
 import torch.nn.functional as F
-import math
 
 if "sinc" in dir(torch):
     sinc = torch.sinc

indextts/BigVGAN/alias_free_activation/torch/resample.py

@@ -3,8 +3,8 @@
 
 import torch.nn as nn
 from torch.nn import functional as F
-from .filter import LowPassFilter1d
-from .filter import kaiser_sinc_filter1d
+
+from .filter import LowPassFilter1d, kaiser_sinc_filter1d
 
 
 class UpSample1d(nn.Module):

indextts/BigVGAN/alias_free_torch/__init__.py

@@ -1,6 +1,6 @@
 # Adapted from https://github.com/junjun3518/alias-free-torch under the Apache License 2.0
 #   LICENSE is in incl_licenses directory.
 
+from .act import *
 from .filter import *
 from .resample import *
-from .act import *

indextts/BigVGAN/alias_free_torch/act.py

@@ -2,7 +2,8 @@
 #   LICENSE is in incl_licenses directory.
 
 import torch.nn as nn
-from .resample import UpSample1d, DownSample1d
+
+from .resample import DownSample1d, UpSample1d
 
 
 class Activation1d(nn.Module):

indextts/BigVGAN/alias_free_torch/filter.py

@@ -1,10 +1,11 @@
 # Adapted from https://github.com/junjun3518/alias-free-torch under the Apache License 2.0
 #   LICENSE is in incl_licenses directory.
 
+import math
+
 import torch
 import torch.nn as nn
 import torch.nn.functional as F
-import math
 
 if 'sinc' in dir(torch):
     sinc = torch.sinc

indextts/BigVGAN/alias_free_torch/resample.py

@@ -3,8 +3,8 @@
 
 import torch.nn as nn
 from torch.nn import functional as F
-from .filter import LowPassFilter1d
-from .filter import kaiser_sinc_filter1d
+
+from .filter import LowPassFilter1d, kaiser_sinc_filter1d
 
 
 class UpSample1d(nn.Module):

indextts/BigVGAN/bigvgan.py

@@ -4,23 +4,23 @@
 # Adapted from https://github.com/jik876/hifi-gan under the MIT license.
 #   LICENSE is in incl_licenses directory.
 
-import os
 import json
+import os
 from pathlib import Path
-from typing import Optional, Union, Dict
+from typing import Dict, Optional, Union
 
 import torch
 import torch.nn as nn
+from huggingface_hub import PyTorchModelHubMixin, hf_hub_download
 from torch.nn import Conv1d, ConvTranspose1d
-from torch.nn.utils import weight_norm, remove_weight_norm
+from torch.nn.utils import remove_weight_norm, weight_norm
 
 import indextts.BigVGAN.activations as activations
-from indextts.BigVGAN.utils import init_weights, get_padding
-from indextts.BigVGAN.alias_free_activation.torch.act import Activation1d as TorchActivation1d
-from indextts.BigVGAN.env import AttrDict
-
-from huggingface_hub import PyTorchModelHubMixin, hf_hub_download
+from indextts.BigVGAN.alias_free_activation.torch.act import \
+    Activation1d as TorchActivation1d
 from indextts.BigVGAN.ECAPA_TDNN import ECAPA_TDNN
+from indextts.BigVGAN.env import AttrDict
+from indextts.BigVGAN.utils import get_padding, init_weights
 
 
 def load_hparams_from_json(path) -> AttrDict:
@@ -94,9 +94,8 @@ class AMPBlock1(torch.nn.Module):
 
         # Select which Activation1d, lazy-load cuda version to ensure backward compatibility
         if self.h.get("use_cuda_kernel", False):
-            from alias_free_activation.cuda.activation1d import (
-                Activation1d as CudaActivation1d,
-            )
+            from alias_free_activation.cuda.activation1d import \
+                Activation1d as CudaActivation1d
 
             Activation1d = CudaActivation1d
         else:
@@ -194,9 +193,8 @@ class AMPBlock2(torch.nn.Module):
 
         # Select which Activation1d, lazy-load cuda version to ensure backward compatibility
         if self.h.get("use_cuda_kernel", False):
-            from alias_free_activation.cuda.activation1d import (
-                Activation1d as CudaActivation1d,
-            )
+            from alias_free_activation.cuda.activation1d import \
+                Activation1d as CudaActivation1d
 
             Activation1d = CudaActivation1d
         else:
@@ -241,6 +239,7 @@ class AMPBlock2(torch.nn.Module):
         for l in self.convs:
             remove_weight_norm(l)
 
+
 '''
     PyTorchModelHubMixin,
     library_name="bigvgan",
@@ -251,6 +250,7 @@ class AMPBlock2(torch.nn.Module):
     tags=["neural-vocoder", "audio-generation", "arxiv:2206.04658"],
 '''
 
+
 class BigVGAN(
     torch.nn.Module,
 ):
@@ -274,9 +274,8 @@ class BigVGAN(
 
         # Select which Activation1d, lazy-load cuda version to ensure backward compatibility
         if self.h.get("use_cuda_kernel", False):
-            from alias_free_activation.cuda.activation1d import (
-                Activation1d as CudaActivation1d,
-            )
+            from alias_free_activation.cuda.activation1d import \
+                Activation1d as CudaActivation1d
 
             Activation1d = CudaActivation1d
         else:

indextts/BigVGAN/models.py

@@ -4,14 +4,13 @@
 # Adapted from https://github.com/jik876/hifi-gan under the MIT license.
 #   LICENSE is in incl_licenses directory.
 
-from torch.nn import Conv1d, ConvTranspose1d, Conv2d
-from torch.nn.utils import weight_norm, remove_weight_norm, spectral_norm
+from torch.nn import Conv1d, Conv2d, ConvTranspose1d
+from torch.nn.utils import remove_weight_norm, spectral_norm, weight_norm
 
 import indextts.BigVGAN.activations as activations
-from indextts.BigVGAN.utils import init_weights, get_padding
 from indextts.BigVGAN.alias_free_torch import *
-
 from indextts.BigVGAN.ECAPA_TDNN import ECAPA_TDNN
+from indextts.BigVGAN.utils import get_padding, init_weights
 
 LRELU_SLOPE = 0.1
 
@@ -180,7 +179,6 @@ class BigVGAN(torch.nn.Module):
 
         # self.logit_scale = nn.Parameter(torch.ones([]) * np.log(1 / 0.07))
 
-
     def forward(self, x, mel_ref, lens=None):
         speaker_embedding = self.speaker_encoder(mel_ref, lens)
         n_batch = x.size(0)
@@ -432,4 +430,3 @@ def generator_loss(disc_outputs):
         loss += l
 
     return loss, gen_losses
-

indextts/BigVGAN/nnet/CNN.py

@@ -19,6 +19,7 @@ import torch.nn as nn
 import torch.nn.functional as F
 import torchaudio
 
+
 class SincConv(nn.Module):
     """This function implements SincConv (SincNet).
 

indextts/BigVGAN/utils.py

@@ -3,13 +3,14 @@
 
 import glob
 import os
+
 import matplotlib
+import matplotlib.pylab as plt
 import torch
+from scipy.io.wavfile import write
 from torch.nn.utils import weight_norm
 
 matplotlib.use("Agg")
-import matplotlib.pylab as plt
-from scipy.io.wavfile import write
 
 MAX_WAV_VALUE = 32768.0
 

indextts/gpt/conformer/embedding.py

@@ -21,6 +21,7 @@ from typing import Tuple, Union
 import torch
 import torch.nn.functional as F
 
+
 class PositionalEncoding(torch.nn.Module):
     """Positional encoding.
 

indextts/gpt/conformer_encoder.py

@@ -3,11 +3,18 @@ from typing import Optional, Tuple
 
 import torch
 import torch.nn as nn
-from gpt.conformer.subsampling import Conv2dSubsampling4, Conv2dSubsampling6, \
-    Conv2dSubsampling8, LinearNoSubsampling, Conv2dSubsampling2
-from gpt.conformer.embedding import PositionalEncoding, RelPositionalEncoding, NoPositionalEncoding
-from gpt.conformer.attention import MultiHeadedAttention, RelPositionMultiHeadedAttention
-from utils.utils import make_pad_mask
+
+from indextts.gpt.conformer.attention import (MultiHeadedAttention,
+                                              RelPositionMultiHeadedAttention)
+from indextts.gpt.conformer.embedding import (NoPositionalEncoding,
+                                              PositionalEncoding,
+                                              RelPositionalEncoding)
+from indextts.gpt.conformer.subsampling import (Conv2dSubsampling2,
+                                                Conv2dSubsampling4,
+                                                Conv2dSubsampling6,
+                                                Conv2dSubsampling8,
+                                                LinearNoSubsampling)
+from indextts.utils.utils import make_pad_mask
 
 
 class PositionwiseFeedForward(torch.nn.Module):
@@ -22,6 +29,7 @@ class PositionwiseFeedForward(torch.nn.Module):
         dropout_rate (float): Dropout rate.
         activation (torch.nn.Module): Activation function
     """
+
     def __init__(self,
                  idim: int,
                  hidden_units: int,
@@ -47,6 +55,7 @@ class PositionwiseFeedForward(torch.nn.Module):
 
 class ConvolutionModule(nn.Module):
     """ConvolutionModule in Conformer model."""
+
     def __init__(self,
                  channels: int,
                  kernel_size: int = 15,
@@ -181,6 +190,7 @@ class ConformerEncoderLayer(nn.Module):
             True: x -> x + linear(concat(x, att(x)))
             False: x -> x + att(x)
     """
+
     def __init__(
         self,
         size: int,
@@ -428,6 +438,7 @@ class BaseEncoder(torch.nn.Module):
 
 class ConformerEncoder(BaseEncoder):
     """Conformer encoder module."""
+
     def __init__(
         self,
         input_size: int,
@@ -507,4 +518,3 @@ class ConformerEncoder(BaseEncoder):
                 concat_after,
             ) for _ in range(num_blocks)
         ])
-

indextts/gpt/model.py

@@ -5,11 +5,13 @@ import torch.nn as nn
 import torch.nn.functional as F
 from transformers import GPT2Config, GPT2PreTrainedModel, LogitsProcessorList
 from transformers.modeling_outputs import CausalLMOutputWithCrossAttentions
-from transformers.utils.model_parallel_utils import get_device_map, assert_device_map
-from gpt.perceiver import PerceiverResampler
-from gpt.conformer_encoder import ConformerEncoder
+from transformers.utils.model_parallel_utils import (assert_device_map,
+                                                     get_device_map)
+
+from indextts.gpt.conformer_encoder import ConformerEncoder
+from indextts.gpt.perceiver import PerceiverResampler
 from indextts.utils.arch_util import AttentionBlock
-from utils.typical_sampling import TypicalLogitsWarper
+from indextts.utils.typical_sampling import TypicalLogitsWarper
 
 
 def null_position_embeddings(range, dim):
@@ -20,6 +22,7 @@ class ResBlock(nn.Module):
     """
     Basic residual convolutional block that uses GroupNorm.
     """
+
     def __init__(self, chan):
         super().__init__()
         self.net = nn.Sequential(
@@ -619,7 +622,3 @@ class UnifiedVoice(nn.Module):
                                             max_length=max_length, logits_processor=logits_processor,
                                             num_return_sequences=num_return_sequences, **hf_generate_kwargs)
         return gen[:, trunc_index:]
-
-
-
-

indextts/infer.py

@@ -1,16 +1,18 @@
 import os
 import re
 import sys
+
+import sentencepiece as spm
 import torch
 import torchaudio
 from omegaconf import OmegaConf
-import sentencepiece as spm
-from utils.utils import tokenize_by_CJK_char
-from utils.feature_extractors import MelSpectrogramFeatures
-from indextts.vqvae.xtts_dvae import DiscreteVAE
-from indextts.utils.checkpoint import load_checkpoint
-from indextts.gpt.model import UnifiedVoice
+
 from indextts.BigVGAN.models import BigVGAN as Generator
+from indextts.gpt.model import UnifiedVoice
+from indextts.utils.checkpoint import load_checkpoint
+from indextts.utils.feature_extractors import MelSpectrogramFeatures
+from indextts.utils.utils import tokenize_by_CJK_char
+from indextts.vqvae.xtts_dvae import DiscreteVAE
 
 
 class IndexTTS:

indextts/utils/arch_util.py

@@ -1,6 +1,8 @@
+import math
+
 import torch
 import torch.nn as nn
-import math
+
 from indextts.utils.xtransformers import RelativePositionBias
 
 

indextts/utils/checkpoint.py

@@ -12,15 +12,14 @@
 # See the License for the specific language governing permissions and
 # limitations under the License.
 
+import datetime
 import logging
 import os
 import re
-
-import yaml
-import torch
 from collections import OrderedDict
 
-import datetime
+import torch
+import yaml
 
 
 def load_checkpoint(model: torch.nn.Module, model_pth: str) -> dict:

indextts/utils/feature_extractors.py

@@ -1,7 +1,8 @@
 import torch
 import torchaudio
 from torch import nn
-from utils import safe_log
+
+from indextts.utils.utils import safe_log
 
 
 class FeatureExtractor(nn.Module):

indextts/utils/utils.py

@@ -1,6 +1,7 @@
 import os
-import re
 import random
+import re
+
 import torch
 import torchaudio
 

indextts/utils/xtransformers.py

@@ -6,7 +6,7 @@ from inspect import isfunction
 import torch
 import torch.nn.functional as F
 from einops import rearrange, repeat
-from torch import nn, einsum
+from torch import einsum, nn
 
 DEFAULT_DIM_HEAD = 64
 

requirements.txt

@@ -14,8 +14,8 @@ matplotlib==3.8.2
 opencv-python==4.9.0.80
 vocos==0.1.0
 accelerate==0.25.0
-omegaconf==2.0.6
 tensorboard==2.9.1
+omegaconf
 sentencepiece
 pypinyin
 librosa

webui.py

@@ -1,16 +1,18 @@
 import os
 import shutil
+import sys
 import threading
 import time
-import sys
+
 current_dir = os.path.dirname(os.path.abspath(__file__))
 sys.path.append(current_dir)
 sys.path.append(os.path.join(current_dir, "indextts"))
 
 import gradio as gr
+from utils.webui_utils import next_page, prev_page
+
 from indextts.infer import IndexTTS
 from tools.i18n.i18n import I18nAuto
-from utils.webui_utils import next_page, prev_page
 
 i18n = I18nAuto(language="zh_CN")
 MODE = 'local'