optimize(f0): move fcpe into rvc.f0

infer/lib/rtrvc.py

@@ -13,7 +13,6 @@ import scipy.signal as signal
 import torch
 import torch.nn as nn
 import torch.nn.functional as F
-import torchcrepe
 from torchaudio.transforms import Resample
 
 from rvc.synthesizer import load_synthesizer
@@ -323,20 +322,17 @@ class RVC:
 
     def get_f0_fcpe(self, x, f0_up_key):
         if hasattr(self, "model_fcpe") == False:
-            from torchfcpe import spawn_bundled_infer_model
+            from rvc.f0 import FCPE
 
             printt("Loading fcpe model")
-            if "privateuseone" in str(self.device):
-                self.device_fcpe = "cpu"
-            else:
-                self.device_fcpe = self.device
-            self.model_fcpe = spawn_bundled_infer_model(self.device_fcpe)
-        f0 = self.model_fcpe.infer(
-            x.to(self.device_fcpe).unsqueeze(0).float(),
-            sr=16000,
-            decoder_mode="local_argmax",
-            threshold=0.006,
-        )
+            self.model_fcpe = FCPE(
+                160,
+                self.f0_min,
+                self.f0_max,
+                16000,
+                self.device,
+            )
+        f0 = self.model_fcpe.compute_f0(x)
         f0 *= pow(2, f0_up_key / 12)
         return self.get_f0_post(f0)
 

infer/modules/vc/pipeline.py

@@ -14,7 +14,7 @@ import torch
 import torch.nn.functional as F
 from scipy import signal
 
-from rvc.f0 import PM, Harvest, RMVPE, CRePE, Dio
+from rvc.f0 import PM, Harvest, RMVPE, CRePE, Dio, FCPE
 
 now_dir = os.getcwd()
 sys.path.append(now_dir)
@@ -118,21 +118,15 @@ class Pipeline(object):
 
         elif f0_method == "fcpe":
             if not hasattr(self, "model_fcpe"):
-                from torchfcpe import spawn_bundled_infer_model
-
                 logger.info("Loading fcpe model")
-                self.model_fcpe = spawn_bundled_infer_model(self.device)
-            f0 = (
-                self.model_fcpe.infer(
-                    torch.from_numpy(x).to(self.device).unsqueeze(0).float(),
-                    sr=16000,
-                    decoder_mode="local_argmax",
-                    threshold=0.006,
+                self.model_fcpe = FCPE(
+                    self.window,
+                    f0_min,
+                    f0_max,
+                    self.sr,
+                    self.device,
                 )
-                .squeeze()
-                .cpu()
-                .numpy()
-            )
+            f0 = self.model_fcpe.compute_f0(x, p_len=p_len)
 
         f0 *= pow(2, f0_up_key / 12)
         # with open("test.txt","w")as f:f.write("\n".join([str(i)for i in f0.tolist()]))