Hugging Face's logo

Datasets:
lamooon
/
vocal

Modalities:
Audio
Tabular
Text
Size:
10K - 100K
Dataset card Data Studio Files
xet
 Community
1
lamooon commited on
Commit
6ae4c4f
·
verified ·
1 Parent(s): 01b1d08

Upload logs/rmvpe_ft/rmvpe_ft_20260329_161302.log with huggingface_hub

Browse files
Files changed (1) hide show
  1. logs/rmvpe_ft/rmvpe_ft_20260329_161302.log +138 -0
logs/rmvpe_ft/rmvpe_ft_20260329_161302.log ADDED
@@ -0,0 +1,138 @@
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
1
+ === Fine-tune RMVPE — Sun Mar 29 16:13:02 HKT 2026 ===
2
+
3
+ === Run 1/5 — Sun Mar 29 16:13:02 HKT 2026 ===
4
+ /userhome/cs/u3571824/fyp/muse/backend/ml_models/dev/vocal/finetune_rmvpe.py:373: FutureWarning: `torch.cuda.amp.GradScaler(args...)` is deprecated. Please use `torch.amp.GradScaler('cuda', args...)` instead.
5
+ scaler = GradScaler(enabled=args.fp16 and device.type == 'cuda')
6
+ Train: 4259 tracks, Val: 531 tracks
7
+ RMVPE E2E verified at /userhome/cs/u3571824/fyp/muse/backend/ml_models/prod/weights/rmvpe.pt (368MB)
8
+ Loading pretrained RMVPE from /userhome/cs/u3571824/fyp/muse/backend/ml_models/prod/weights/rmvpe.pt
9
+ RMVPE: 92.0M params
10
+
11
+ === Fine-tuning RMVPE: 20 epochs, lr=0.0001 ===
12
+ /userhome/cs/u3571824/fyp/muse/backend/ml_models/dev/vocal/finetune_rmvpe.py:405: FutureWarning: `torch.cuda.amp.autocast(args...)` is deprecated. Please use `torch.amp.autocast('cuda', args...)` instead.
13
+ with cuda_autocast(enabled=args.fp16 and device.type == 'cuda'):
14
+ Traceback (most recent call last):
15
+ File "/userhome/cs/u3571824/fyp/muse/backend/ml_models/dev/vocal/finetune_rmvpe.py", line 482, in <module>
16
+ main()
17
+ File "/userhome/cs/u3571824/fyp/muse/backend/ml_models/dev/vocal/finetune_rmvpe.py", line 420, in main
18
+ loss = focal_loss(pred, label)
19
+ File "/userhome/cs/u3571824/fyp/muse/backend/ml_models/dev/vocal/finetune_rmvpe.py", line 380, in focal_loss
20
+ bce = F.binary_cross_entropy(pred, target, reduction='none')
21
+ File "/userhome/cs/u3571824/anaconda3/envs/bass/lib/python3.10/site-packages/torch/nn/functional.py", line 3574, in binary_cross_entropy
22
+ return torch._C._nn.binary_cross_entropy(input, target, weight, reduction_enum)
23
+ RuntimeError: torch.nn.functional.binary_cross_entropy and torch.nn.BCELoss are unsafe to autocast.
24
+ Many models use a sigmoid layer right before the binary cross entropy layer.
25
+ In this case, combine the two layers using torch.nn.functional.binary_cross_entropy_with_logits
26
+ or torch.nn.BCEWithLogitsLoss. binary_cross_entropy_with_logits and BCEWithLogits are
27
+ safe to autocast.
28
+ Transient crash (exit 1). Retrying in 15s...
29
+
30
+ === Run 2/5 — Sun Mar 29 16:13:30 HKT 2026 ===
31
+ /userhome/cs/u3571824/fyp/muse/backend/ml_models/dev/vocal/finetune_rmvpe.py:373: FutureWarning: `torch.cuda.amp.GradScaler(args...)` is deprecated. Please use `torch.amp.GradScaler('cuda', args...)` instead.
32
+ scaler = GradScaler(enabled=args.fp16 and device.type == 'cuda')
33
+ Train: 4259 tracks, Val: 531 tracks
34
+ RMVPE E2E verified at /userhome/cs/u3571824/fyp/muse/backend/ml_models/prod/weights/rmvpe.pt (368MB)
35
+ Loading pretrained RMVPE from /userhome/cs/u3571824/fyp/muse/backend/ml_models/prod/weights/rmvpe.pt
36
+ RMVPE: 92.0M params
37
+
38
+ === Fine-tuning RMVPE: 20 epochs, lr=0.0001 ===
39
+ /userhome/cs/u3571824/fyp/muse/backend/ml_models/dev/vocal/finetune_rmvpe.py:405: FutureWarning: `torch.cuda.amp.autocast(args...)` is deprecated. Please use `torch.amp.autocast('cuda', args...)` instead.
40
+ with cuda_autocast(enabled=args.fp16 and device.type == 'cuda'):
41
+ Traceback (most recent call last):
42
+ File "/userhome/cs/u3571824/fyp/muse/backend/ml_models/dev/vocal/finetune_rmvpe.py", line 482, in <module>
43
+ main()
44
+ File "/userhome/cs/u3571824/fyp/muse/backend/ml_models/dev/vocal/finetune_rmvpe.py", line 420, in main
45
+ loss = focal_loss(pred, label)
46
+ File "/userhome/cs/u3571824/fyp/muse/backend/ml_models/dev/vocal/finetune_rmvpe.py", line 380, in focal_loss
47
+ bce = F.binary_cross_entropy(pred, target, reduction='none')
48
+ File "/userhome/cs/u3571824/anaconda3/envs/bass/lib/python3.10/site-packages/torch/nn/functional.py", line 3574, in binary_cross_entropy
49
+ return torch._C._nn.binary_cross_entropy(input, target, weight, reduction_enum)
50
+ RuntimeError: torch.nn.functional.binary_cross_entropy and torch.nn.BCELoss are unsafe to autocast.
51
+ Many models use a sigmoid layer right before the binary cross entropy layer.
52
+ In this case, combine the two layers using torch.nn.functional.binary_cross_entropy_with_logits
53
+ or torch.nn.BCEWithLogitsLoss. binary_cross_entropy_with_logits and BCEWithLogits are
54
+ safe to autocast.
55
+ Transient crash (exit 1). Retrying in 15s...
56
+
57
+ === Run 3/5 — Sun Mar 29 16:13:57 HKT 2026 ===
58
+ /userhome/cs/u3571824/fyp/muse/backend/ml_models/dev/vocal/finetune_rmvpe.py:373: FutureWarning: `torch.cuda.amp.GradScaler(args...)` is deprecated. Please use `torch.amp.GradScaler('cuda', args...)` instead.
59
+ scaler = GradScaler(enabled=args.fp16 and device.type == 'cuda')
60
+ Train: 4259 tracks, Val: 531 tracks
61
+ RMVPE E2E verified at /userhome/cs/u3571824/fyp/muse/backend/ml_models/prod/weights/rmvpe.pt (368MB)
62
+ Loading pretrained RMVPE from /userhome/cs/u3571824/fyp/muse/backend/ml_models/prod/weights/rmvpe.pt
63
+ RMVPE: 92.0M params
64
+
65
+ === Fine-tuning RMVPE: 20 epochs, lr=0.0001 ===
66
+ /userhome/cs/u3571824/fyp/muse/backend/ml_models/dev/vocal/finetune_rmvpe.py:405: FutureWarning: `torch.cuda.amp.autocast(args...)` is deprecated. Please use `torch.amp.autocast('cuda', args...)` instead.
67
+ with cuda_autocast(enabled=args.fp16 and device.type == 'cuda'):
68
+ Traceback (most recent call last):
69
+ File "/userhome/cs/u3571824/fyp/muse/backend/ml_models/dev/vocal/finetune_rmvpe.py", line 482, in <module>
70
+ main()
71
+ File "/userhome/cs/u3571824/fyp/muse/backend/ml_models/dev/vocal/finetune_rmvpe.py", line 420, in main
72
+ loss = focal_loss(pred, label)
73
+ File "/userhome/cs/u3571824/fyp/muse/backend/ml_models/dev/vocal/finetune_rmvpe.py", line 380, in focal_loss
74
+ bce = F.binary_cross_entropy(pred, target, reduction='none')
75
+ File "/userhome/cs/u3571824/anaconda3/envs/bass/lib/python3.10/site-packages/torch/nn/functional.py", line 3574, in binary_cross_entropy
76
+ return torch._C._nn.binary_cross_entropy(input, target, weight, reduction_enum)
77
+ RuntimeError: torch.nn.functional.binary_cross_entropy and torch.nn.BCELoss are unsafe to autocast.
78
+ Many models use a sigmoid layer right before the binary cross entropy layer.
79
+ In this case, combine the two layers using torch.nn.functional.binary_cross_entropy_with_logits
80
+ or torch.nn.BCEWithLogitsLoss. binary_cross_entropy_with_logits and BCEWithLogits are
81
+ safe to autocast.
82
+ Transient crash (exit 1). Retrying in 15s...
83
+
84
+ === Run 4/5 — Sun Mar 29 16:14:24 HKT 2026 ===
85
+ /userhome/cs/u3571824/fyp/muse/backend/ml_models/dev/vocal/finetune_rmvpe.py:373: FutureWarning: `torch.cuda.amp.GradScaler(args...)` is deprecated. Please use `torch.amp.GradScaler('cuda', args...)` instead.
86
+ scaler = GradScaler(enabled=args.fp16 and device.type == 'cuda')
87
+ Train: 4259 tracks, Val: 531 tracks
88
+ RMVPE E2E verified at /userhome/cs/u3571824/fyp/muse/backend/ml_models/prod/weights/rmvpe.pt (368MB)
89
+ Loading pretrained RMVPE from /userhome/cs/u3571824/fyp/muse/backend/ml_models/prod/weights/rmvpe.pt
90
+ RMVPE: 92.0M params
91
+
92
+ === Fine-tuning RMVPE: 20 epochs, lr=0.0001 ===
93
+ /userhome/cs/u3571824/fyp/muse/backend/ml_models/dev/vocal/finetune_rmvpe.py:405: FutureWarning: `torch.cuda.amp.autocast(args...)` is deprecated. Please use `torch.amp.autocast('cuda', args...)` instead.
94
+ with cuda_autocast(enabled=args.fp16 and device.type == 'cuda'):
95
+ Traceback (most recent call last):
96
+ File "/userhome/cs/u3571824/fyp/muse/backend/ml_models/dev/vocal/finetune_rmvpe.py", line 482, in <module>
97
+ main()
98
+ File "/userhome/cs/u3571824/fyp/muse/backend/ml_models/dev/vocal/finetune_rmvpe.py", line 420, in main
99
+ loss = focal_loss(pred, label)
100
+ File "/userhome/cs/u3571824/fyp/muse/backend/ml_models/dev/vocal/finetune_rmvpe.py", line 380, in focal_loss
101
+ bce = F.binary_cross_entropy(pred, target, reduction='none')
102
+ File "/userhome/cs/u3571824/anaconda3/envs/bass/lib/python3.10/site-packages/torch/nn/functional.py", line 3574, in binary_cross_entropy
103
+ return torch._C._nn.binary_cross_entropy(input, target, weight, reduction_enum)
104
+ RuntimeError: torch.nn.functional.binary_cross_entropy and torch.nn.BCELoss are unsafe to autocast.
105
+ Many models use a sigmoid layer right before the binary cross entropy layer.
106
+ In this case, combine the two layers using torch.nn.functional.binary_cross_entropy_with_logits
107
+ or torch.nn.BCEWithLogitsLoss. binary_cross_entropy_with_logits and BCEWithLogits are
108
+ safe to autocast.
109
+ Transient crash (exit 1). Retrying in 15s...
110
+
111
+ === Run 5/5 — Sun Mar 29 16:14:49 HKT 2026 ===
112
+ /userhome/cs/u3571824/fyp/muse/backend/ml_models/dev/vocal/finetune_rmvpe.py:373: FutureWarning: `torch.cuda.amp.GradScaler(args...)` is deprecated. Please use `torch.amp.GradScaler('cuda', args...)` instead.
113
+ scaler = GradScaler(enabled=args.fp16 and device.type == 'cuda')
114
+ Train: 4259 tracks, Val: 531 tracks
115
+ RMVPE E2E verified at /userhome/cs/u3571824/fyp/muse/backend/ml_models/prod/weights/rmvpe.pt (368MB)
116
+ Loading pretrained RMVPE from /userhome/cs/u3571824/fyp/muse/backend/ml_models/prod/weights/rmvpe.pt
117
+ RMVPE: 92.0M params
118
+
119
+ === Fine-tuning RMVPE: 20 epochs, lr=0.0001 ===
120
+ /userhome/cs/u3571824/fyp/muse/backend/ml_models/dev/vocal/finetune_rmvpe.py:405: FutureWarning: `torch.cuda.amp.autocast(args...)` is deprecated. Please use `torch.amp.autocast('cuda', args...)` instead.
121
+ with cuda_autocast(enabled=args.fp16 and device.type == 'cuda'):
122
+ Traceback (most recent call last):
123
+ File "/userhome/cs/u3571824/fyp/muse/backend/ml_models/dev/vocal/finetune_rmvpe.py", line 482, in <module>
124
+ main()
125
+ File "/userhome/cs/u3571824/fyp/muse/backend/ml_models/dev/vocal/finetune_rmvpe.py", line 420, in main
126
+ loss = focal_loss(pred, label)
127
+ File "/userhome/cs/u3571824/fyp/muse/backend/ml_models/dev/vocal/finetune_rmvpe.py", line 380, in focal_loss
128
+ bce = F.binary_cross_entropy(pred, target, reduction='none')
129
+ File "/userhome/cs/u3571824/anaconda3/envs/bass/lib/python3.10/site-packages/torch/nn/functional.py", line 3574, in binary_cross_entropy
130
+ return torch._C._nn.binary_cross_entropy(input, target, weight, reduction_enum)
131
+ RuntimeError: torch.nn.functional.binary_cross_entropy and torch.nn.BCELoss are unsafe to autocast.
132
+ Many models use a sigmoid layer right before the binary cross entropy layer.
133
+ In this case, combine the two layers using torch.nn.functional.binary_cross_entropy_with_logits
134
+ or torch.nn.BCEWithLogitsLoss. binary_cross_entropy_with_logits and BCEWithLogits are
135
+ safe to autocast.
136
+ Transient crash (exit 1). Retrying in 15s...
137
+
138
+ === Upload to HuggingFace ===