heimoshuiyu/ultimatevocalremovergui
1
0
Fork 0
Code Issues Pull Requests Packages Projects Releases Wiki Activity

Add files via upload

Browse Source
This commit is contained in:
Anjok07
2022-12-29 21:07:23 -06:00
committed by GitHub
parent 00ae0157ff
commit f968803f89
2 changed files with 59 additions and 54 deletions
Download Patch File Download Diff File Expand all files Collapse all files

9
separate.py
View File

@@ -224,13 +224,14 @@ class SeperateAttributes:
class SeperateMDX(SeperateAttributes):
def seperate(self):
samplerate = 44100
if self.primary_model_name == self.model_basename and self.primary_sources:
self.primary_source, self.secondary_source = self.load_cached_sources()
else:
self.start_inference()
if self.is_gpu_conversion >= 0:
self.device = torch.device('cuda:0' if torch.cuda.is_available() else 'cpu')
run_type = ['CUDAExecutionProvider'] if torch.cuda.is_available() else ['CPUExecutionProvider']
@@ -668,10 +669,10 @@ class SeperateVR(SeperateAttributes):
for d in range(bands_n, 0, -1):
bp = self.mp.param['band'][d]
if OPERATING_SYSTEM == 'Windows':
wav_resolution = bp['res_type']
if OPERATING_SYSTEM == 'Darwin':
wav_resolution = 'polyphase' if SYSTEM_PROC == ARM or ARM in SYSTEM_ARCH else bp['res_type']
else:
wav_resolution = 'polyphase'
wav_resolution = bp['res_type']
if d == bands_n: # high-end band
X_wave[d], _ = librosa.load(