fix(dml): train extract_f0_print error

ModuleNotFoundError: No module named 'torch.privateuseone' due to new prosess

Author: fumiama

infer/modules/train/extract_f0_print.py

@@ -2,6 +2,7 @@
 import sys
 import traceback
 from pathlib import Path
+import importlib.util
 
 from dotenv import load_dotenv
 
@@ -38,6 +39,8 @@ def printt(strr):
 device = sys.argv[4]
 is_half = sys.argv[5] == "True"
 
+if importlib.util.find_spec("torch_directml") is not None:
+    import torch_directml # use side effect
 
 class FeatureInput(object):
     def __init__(self, is_half: bool, device="cpu", samplerate=16000, hop_size=160):

rvc/f0/f0.py

@@ -11,14 +11,14 @@ def __init__(
         f0_min=50,
         f0_max=1100,
         sampling_rate=44100,
-        device: Optional[str] = None,
+        device: Optional[Union[str, torch.device]] = None,
     ):
         self.hop_length = hop_length
         self.f0_min = f0_min
         self.f0_max = f0_max
         self.sampling_rate = sampling_rate
-        if device is None:
-            device = "cuda:0" if torch.cuda.is_available() else "cpu"
+        if not device:
+            device = torch.device("cuda:0" if torch.cuda.is_available() else "cpu")
         self.device = device
 
     def compute_f0(

rvc/f0/mel.py

@@ -1,4 +1,4 @@
-from typing import Optional
+from typing import Optional, Union
 
 import torch
 import numpy as np
@@ -17,9 +17,9 @@ def __init__(
         hop_length: int,
         n_fft: Optional[int] = None,
         mel_fmin: int = 0,
-        mel_fmax: int = None,
+        mel_fmax: Optional[int] = None,
         clamp: float = 1e-5,
-        device=torch.device("cpu"),
+        device: Union[str, torch.device] = torch.device("cpu"),
     ):
         super().__init__()
         if n_fft is None:

rvc/f0/rmvpe.py

@@ -1,6 +1,6 @@
 from io import BytesIO
 import os
-from typing import Any, Optional, Union
+from typing import Optional, Union
 
 import numpy as np
 import torch

rvc/layers/attentions.py

@@ -12,8 +12,8 @@ def __init__(
         channels: int,
         out_channels: int,
         n_heads: int,
+        window_size: int,
         p_dropout: float = 0.0,
-        window_size: Optional[int] = None,
         heads_share: bool = True,
         block_length: Optional[int] = None,
         proximal_bias: bool = False,

rvc/layers/encoders.py

@@ -42,8 +42,8 @@ def __init__(
                     hidden_channels,
                     hidden_channels,
                     n_heads,
+                    window_size,
                     p_dropout=p_dropout,
-                    window_size=window_size,
                 )
             )
             self.norm_layers_1.append(LayerNorm(hidden_channels))
@@ -121,7 +121,7 @@ def __init__(
     def __call__(
         self,
         phone: torch.Tensor,
-        pitch: torch.Tensor,
+        pitch: Optional[torch.Tensor],
         lengths: torch.Tensor,
         skip_head: Optional[int] = None,
     ) -> Tuple[torch.Tensor, torch.Tensor, torch.Tensor]:
@@ -135,7 +135,7 @@ def __call__(
     def forward(
         self,
         phone: torch.Tensor,
-        pitch: torch.Tensor,
+        pitch: Optional[torch.Tensor],
         lengths: torch.Tensor,
         skip_head: Optional[int] = None,
     ) -> Tuple[torch.Tensor, torch.Tensor, torch.Tensor]:

rvc/layers/generators.py

@@ -46,6 +46,7 @@ def __init__(
 
         self.resblocks = nn.ModuleList()
         resblock_module = ResBlock1 if resblock == "1" else ResBlock2
+        ch = 0
         for i in range(len(self.ups)):
             ch = upsample_initial_channel // (2 ** (i + 1))
             for k, d in zip(resblock_kernel_sizes, resblock_dilation_sizes):

rvc/layers/norms.py

@@ -30,7 +30,7 @@ def __init__(
         dilation_rate: int,
         n_layers: int,
         gin_channels: int = 0,
-        p_dropout: int = 0,
+        p_dropout: float = 0,
     ):
         super(WN, self).__init__()
         assert kernel_size % 2 == 1

rvc/layers/nsf.py

@@ -1,4 +1,4 @@
-from typing import Optional, List
+from typing import Optional, List, Union
 import math
 
 import torch
@@ -83,7 +83,7 @@ def __init__(
         self.conv_pre = Conv1d(
             initial_channel, upsample_initial_channel, 7, 1, padding=3
         )
-        resblock = ResBlock1 if resblock == "1" else ResBlock2
+        resblockcls = ResBlock1 if resblock == "1" else ResBlock2
 
         self.ups = nn.ModuleList()
         for i, (u, k) in enumerate(zip(upsample_rates, upsample_kernel_sizes)):
@@ -114,12 +114,13 @@ def __init__(
                 self.noise_convs.append(Conv1d(1, c_cur, kernel_size=1))
 
         self.resblocks = nn.ModuleList()
+        ch = 0
         for i in range(len(self.ups)):
-            ch: int = upsample_initial_channel // (2 ** (i + 1))
+            ch = upsample_initial_channel // (2 ** (i + 1))
             for j, (k, d) in enumerate(
                 zip(resblock_kernel_sizes, resblock_dilation_sizes)
             ):
-                self.resblocks.append(resblock(ch, k, d))
+                self.resblocks.append(resblockcls(ch, k, d))
 
         self.conv_post = Conv1d(ch, 1, 7, 1, padding=3, bias=False)
         self.ups.apply(call_weight_data_normal_if_Conv)

rvc/layers/residuals.py

@@ -20,7 +20,7 @@ def __init__(
         self,
         channels: int,
         kernel_size: int = 3,
-        dilation: List[int] = (1, 3, 5),
+        dilation: List[int] = [1, 3, 5],
     ):
         super(ResBlock1, self).__init__()
 
@@ -117,7 +117,7 @@ def __init__(
         self,
         channels: int,
         kernel_size=3,
-        dilation: List[int] = (1, 3),
+        dilation: List[int] = [1, 3],
     ):
         super(ResBlock2, self).__init__()
         self.convs = nn.ModuleList()
@@ -182,7 +182,7 @@ def __init__(
         kernel_size: int,
         dilation_rate: int,
         n_layers: int,
-        p_dropout: int = 0,
+        p_dropout: float = 0,
         gin_channels: int = 0,
         mean_only: bool = False,
     ):