chore(format): run black on main

IAHispano · Feb 15, 2025 · ffa785a · ffa785a
1 parent 7a4793d
commit ffa785a
Show file tree

Hide file tree

Showing 3 changed files with 22 additions and 26 deletions.
diff --git a/rvc/lib/algorithm/generators/hifigan_mrf.py b/rvc/lib/algorithm/generators/hifigan_mrf.py
@@ -1,4 +1,3 @@
-import math
 from typing import Optional
 
 import numpy as np
@@ -274,7 +273,7 @@ def __init__(
         self.noise_convs = torch.nn.ModuleList()
 
         # f0 input gets upscaled to full segment size, then downscaled back to match each upscale step
-        
+
         # segment  upscaling mel           downscaling f0
         # 36     = input z               = 17280 / (2 x 2 x 10 x 12)
         # 432    = 36 x 12               = 17280 / (2 x 2 x 10)
@@ -284,9 +283,9 @@ def __init__(
 
         stride_f0s = [
             upsample_rates[1] * upsample_rates[2] * upsample_rates[3],
-                                upsample_rates[2] * upsample_rates[3],
-                                                    upsample_rates[3],
-                                                                    1,
+            upsample_rates[2] * upsample_rates[3],
+            upsample_rates[3],
+            1,
         ]
 
         for i, (u, k) in enumerate(zip(upsample_rates, upsample_kernel_sizes)):
@@ -307,7 +306,7 @@ def __init__(
                     torch.nn.Conv1d(
                         in_channels=1,
                         out_channels=upsample_initial_channel // (2 ** (i + 1)),
-                        kernel_size = 1
+                        kernel_size=1,
                     )
                 )
             else:
@@ -323,7 +322,7 @@ def __init__(
                         )
                     )
                 )
-            
+
                 self.noise_convs.append(
                     torch.nn.Conv1d(
                         1,
@@ -363,7 +362,9 @@ def forward(
             # in-place call
             x += self.cond(g)
 
-        for ups, upr, mrf, noise_conv in zip(self.upsamples, self.upsampler, self.mrfs, self.noise_convs):
+        for ups, upr, mrf, noise_conv in zip(
+            self.upsamples, self.upsampler, self.mrfs, self.noise_convs
+        ):
             # in-place call
             x = torch.nn.functional.leaky_relu_(x, LRELU_SLOPE)
 
@@ -379,7 +380,7 @@ def forward(
             h = noise_conv(har_source)
             if self.upp == 441:
                 h = torch.nn.functional.interpolate(h, size=x.shape[-1], mode="linear")
-            
+
             x += h
 
             def mrf_sum(x, layers):

diff --git a/rvc/lib/algorithm/generators/refinegan.py b/rvc/lib/algorithm/generators/refinegan.py
@@ -1,4 +1,3 @@
-import math
 import numpy as np
 import torch
 from torch import nn
@@ -321,7 +320,7 @@ def __init__(
         self.checkpointing = checkpointing
         self.upp = int(np.prod(upsample_rates))
         assert self.upp == sample_rate // 100
-        
+
         self.m_source = SineGenerator(sample_rate)
 
         # expanded f0 sinegen -> match mel_conv
@@ -335,9 +334,9 @@ def __init__(
                 bias=False,
             )
         )
-        
+
         # f0 input gets upscaled to full segment size, then downscaled back to match each upscale step
-        
+
         # segment  upscaling mel           downscaling f0
         # 36     = input z               = 17280 / (2 x 2 x 10 x 12)
         # 432    = 36 x 12               = 17280 / (2 x 2 x 10)
@@ -347,23 +346,23 @@ def __init__(
 
         stride_f0s = [
             upsample_rates[1] * upsample_rates[2] * upsample_rates[3],
-                                upsample_rates[2] * upsample_rates[3],
-                                                    upsample_rates[3],
-                                                                    1,
+            upsample_rates[2] * upsample_rates[3],
+            upsample_rates[3],
+            1,
         ]
 
         channels = upsample_initial_channel
 
         self.downsample_blocks = nn.ModuleList([])
         for i, u in enumerate(upsample_rates):
-            
-			# 44k f0 downsampling is done using F.interpolate in the forward call due to 2.205 multiplier
+
+            # 44k f0 downsampling is done using F.interpolate in the forward call due to 2.205 multiplier
             if self.upp == 441:
                 self.downsample_blocks.append(
                     nn.Conv1d(
                         in_channels=1,
                         out_channels=channels // 2 ** (i + 2),
-                        kernel_size = 1
+                        kernel_size=1,
                     )
                 )
             else:
@@ -466,7 +465,7 @@ def forward(self, mel: torch.Tensor, f0: torch.Tensor, g: torch.Tensor = None):
                 h = down(har_source)
                 if self.upp == 441:
                     h = F.interpolate(h, size=x.shape[-1], mode="linear")
-                x = torch.cat([x, h], dim=1)                    
+                x = torch.cat([x, h], dim=1)
                 x = checkpoint(res, x, use_reentrant=False)
             else:
                 x = ups(x)

diff --git a/tabs/inference/inference.py b/tabs/inference/inference.py
@@ -131,14 +131,10 @@ def get_presets_data(pitch, index_rate, rms_mix_rate, protect):
     }
 
 
-def export_presets_button(
-    preset_name, pitch, index_rate, rms_mix_rate, protect
-):
+def export_presets_button(preset_name, pitch, index_rate, rms_mix_rate, protect):
     if preset_name:
         file_path = os.path.join(PRESETS_DIR, f"{preset_name}.json")
-        presets_data = get_presets_data(
-            pitch, index_rate, rms_mix_rate, protect
-        )
+        presets_data = get_presets_data(pitch, index_rate, rms_mix_rate, protect)
         with open(file_path, "w", encoding="utf-8") as json_file:
             json.dump(presets_data, json_file, ensure_ascii=False, indent=4)
         return "Export successful"