Update DPTNet_eval/DPTNet_quant_sep.py
Browse files- DPTNet_eval/DPTNet_quant_sep.py +26 -15
DPTNet_eval/DPTNet_quant_sep.py
CHANGED
@@ -66,21 +66,30 @@ import tempfile
|
|
66 |
|
67 |
def dpt_sep_process(wav_path, model=None, outfilename=None):
|
68 |
try:
|
69 |
-
|
70 |
-
|
71 |
-
|
72 |
-
|
73 |
-
|
74 |
-
x =
|
75 |
-
|
76 |
-
|
|
|
77 |
if sr != 16000:
|
78 |
-
resampler = torchaudio.transforms.Resample(sr, 16000)
|
79 |
x = resampler(x)
|
80 |
sr = 16000
|
81 |
-
|
82 |
-
|
83 |
-
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
84 |
|
85 |
# 後處理修正
|
86 |
est_sources = est_sources.squeeze(0)
|
@@ -107,9 +116,11 @@ def dpt_sep_process(wav_path, model=None, outfilename=None):
|
|
107 |
|
108 |
return final_sep1, final_sep2
|
109 |
|
110 |
-
except
|
111 |
-
|
112 |
-
|
|
|
|
|
113 |
|
114 |
if __name__ == '__main__':
|
115 |
print("This module should be used via Flask or Gradio.")
|
|
|
66 |
|
67 |
def dpt_sep_process(wav_path, model=None, outfilename=None):
|
68 |
try:
|
69 |
+
# 添加設備檢測
|
70 |
+
device = torch.device("cuda" if torch.cuda.is_available() else "cpu")
|
71 |
+
model = model.to(device)
|
72 |
+
|
73 |
+
# 強化音訊加載
|
74 |
+
x, sr = torchaudio.load(wav_path, format="wav")
|
75 |
+
x = x.mean(dim=0, keepdim=True).to(device)
|
76 |
+
|
77 |
+
# 自動重採樣
|
78 |
if sr != 16000:
|
79 |
+
resampler = torchaudio.transforms.Resample(sr, 16000).to(device)
|
80 |
x = resampler(x)
|
81 |
sr = 16000
|
82 |
+
|
83 |
+
# 分塊處理避免OOM
|
84 |
+
chunk_size = sr * 60 # 每次處理1分鐘
|
85 |
+
separated = []
|
86 |
+
for i in range(0, x.shape[1], chunk_size):
|
87 |
+
chunk = x[:, i:i+chunk_size]
|
88 |
+
with torch.no_grad():
|
89 |
+
est = model(chunk)
|
90 |
+
separated.append(est.cpu())
|
91 |
+
|
92 |
+
est_sources = torch.cat(separated, dim=2)
|
93 |
|
94 |
# 後處理修正
|
95 |
est_sources = est_sources.squeeze(0)
|
|
|
116 |
|
117 |
return final_sep1, final_sep2
|
118 |
|
119 |
+
except RuntimeError as e:
|
120 |
+
if "CUDA out of memory" in str(e):
|
121 |
+
raise RuntimeError("記憶體不足,請縮短音訊長度") from e
|
122 |
+
else:
|
123 |
+
raise
|
124 |
|
125 |
if __name__ == '__main__':
|
126 |
print("This module should be used via Flask or Gradio.")
|