optimize(rvc): move . into layers

infer/lib/infer_pack/models_onnx.py

@@ -1,9 +1,9 @@
 import torch
 from torch import nn
 
-from rvc.nsf import NSFGenerator
-from rvc.encoders import TextEncoder, PosteriorEncoder
-from rvc.residuals import ResidualCouplingBlock
+from rvc.layers.nsf import NSFGenerator
+from rvc.layers.encoders import TextEncoder, PosteriorEncoder
+from rvc.layers.residuals import ResidualCouplingBlock
 
 
 class SynthesizerTrnMsNSFsidM(nn.Module):

infer/lib/jit/synthesizer.py

@@ -1,6 +1,6 @@
 import torch
 
-from rvc.synthesizers import SynthesizerTrnMsNSFsid
+from rvc.layers.synthesizers import SynthesizerTrnMsNSFsid
 
 
 def get_synthesizer_ckpt(cpt, device=torch.device("cpu")):

infer/modules/train/train.py

@@ -56,15 +56,15 @@ from infer.lib.train.data_utils import (
     TextAudioLoaderMultiNSFsid,
 )
 
-from rvc.discriminators import MultiPeriodDiscriminator
+from rvc.layers.discriminators import MultiPeriodDiscriminator
 
 if hps.version == "v1":
-    from rvc.synthesizers import SynthesizerTrnMs256NSFsid as RVC_Model_f0
-    from rvc.synthesizers import (
+    from rvc.layers.synthesizers import SynthesizerTrnMs256NSFsid as RVC_Model_f0
+    from rvc.layers.synthesizers import (
         SynthesizerTrnMs256NSFsid_nono as RVC_Model_nof0,
     )
 else:
-    from rvc.synthesizers import (
+    from rvc.layers.synthesizers import (
         SynthesizerTrnMs768NSFsid as RVC_Model_f0,
         SynthesizerTrnMs768NSFsid_nono as RVC_Model_nof0,
     )

rvc/layers/synthesizers.py

@@ -3,14 +3,15 @@ from typing import Optional, List, Union
 import torch
 from torch import nn
 
-from rvc.residuals import ResidualCouplingBlock
-from rvc.utils import (
+
+from .encoders import TextEncoder, PosteriorEncoder
+from .generators import Generator
+from .nsf import NSFGenerator
+from .residuals import ResidualCouplingBlock
+from .utils import (
     slice_on_last_dim,
     rand_slice_segments_on_last_dim,
 )
-from rvc.encoders import TextEncoder, PosteriorEncoder
-from rvc.generators import Generator
-from rvc.nsf import NSFGenerator
 
 
 class SynthesizerTrnMsNSFsid(nn.Module):

rvc/onnx/infer.py

@@ -75,7 +75,7 @@ class RVC(Model):
         self.vec_model = ContentVec(vec_path, device)
         self.hop_len = hop_len
 
-    def inference(
+    def infer(
         self,
         wav: np.ndarray[typing.Any, np.dtype],
         wav_sr: int,
@@ -119,7 +119,7 @@ class RVC(Model):
         rnd = np.random.randn(1, 192, hubert_length).astype(np.float32)
         hubert_length = np.array([hubert_length]).astype(np.int64)
 
-        out_wav = self.__forward(
+        out_wav = self.forward(
             hubert, hubert_length, pitch, pitchf, ds, rnd
         ).squeeze()
 
@@ -127,7 +127,7 @@ class RVC(Model):
 
         return out_wav[0:org_length]
 
-    def __forward(
+    def forward(
         self,
         hubert: np.ndarray[typing.Any, np.dtype[np.float32]],
         hubert_length: int,

tools/cmd/infer-pm-index256.py

@@ -24,7 +24,7 @@ from fairseq import checkpoint_utils
 
 # from models import SynthesizerTrn256#hifigan_nonsf
 # from lib.infer_pack.models import SynthesizerTrn256NSF as SynthesizerTrn256#hifigan_nsf
-from rvc.synthesizers import (
+from rvc.layers.synthesizers import (
     SynthesizerTrnMs256NSFsid as SynthesizerTrn256,
 )  # hifigan_nsf
 from scipy.io import wavfile

tools/onnx/infer.py

@@ -19,6 +19,6 @@ model = RVC(
 
 wav, sr = librosa.load(wav_path, sr=sampling_rate)
 
-audio = model.inference(wav, sr, sid, f0_method=f0_method, f0_up_key=f0_up_key)
+audio = model.infer(wav, sr, sid, f0_method=f0_method, f0_up_key=f0_up_key)
 
 soundfile.write(out_path, audio, sampling_rate)