Upload logs/vocal_onset_exp/vocal_onset_exp_20260404_021015.log with huggingface_hub

logs/vocal_onset_exp/vocal_onset_exp_20260404_021015.log

@@ -14,3 +14,95 @@ Train: 11283  Val: 39
 Feature extraction: 11322 new, 0 cached, 0 failed
 extract_only done
 Phase 2 done — cache at /userhome/cs/u3571824/fyp/muse/backend/ml_models/dev/vocal/cache/onset_v3
+Uploaded: logs/vocal_onset_exp/vocal_onset_exp_20260404_021015.log
+
+===============================================
+  Phase 3: Train onset_v3 (OnsetF0GRU, f0 derivatives) -- Sat Apr  4 02:20:40 HKT 2026
+===============================================
+/userhome/cs/u3571824/fyp/muse/backend/ml_models/dev/vocal/train_onset_v3.py:434: FutureWarning: `torch.cuda.amp.GradScaler(args...)` is deprecated. Please use `torch.amp.GradScaler('cuda', args...)` instead.
+  scaler = GradScaler(enabled=(args.fp16 and device == 'cuda'))
+Device: cuda
+Train: 11283  Val: 39
+
+--- Feature extraction (11322 tracks) ---
+Feature extraction: 0 new, 11322 cached, 0 failed
+OnsetF0GRU: 0.103M params
+
+=== Training: 40 epochs ===
+/userhome/cs/u3571824/fyp/muse/backend/ml_models/dev/vocal/train_onset_v3.py:304: FutureWarning: `torch.cuda.amp.autocast(args...)` is deprecated. Please use `torch.amp.autocast('cuda', args...)` instead.
+  with cuda_autocast(enabled=scaler.is_enabled()):
+/userhome/cs/u3571824/anaconda3/envs/bass/lib/python3.10/site-packages/mir_eval/onset.py:49: UserWarning: Reference onsets are empty.
+  warnings.warn("Reference onsets are empty.")
+/userhome/cs/u3571824/anaconda3/envs/bass/lib/python3.10/site-packages/mir_eval/onset.py:51: UserWarning: Estimated onsets are empty.
+  warnings.warn("Estimated onsets are empty.")
+Ep 1/40 (14s) lr=1.00e-03 | train_loss=0.0289 | val_f1=0.1360 P=0.1375 R=0.1793
+  ★ New best val_f1: 0.1360
+Ep 2/40 (13s) lr=1.00e-03 | train_loss=0.0246 | val_f1=0.1933 P=0.1676 R=0.2725
+  ★ New best val_f1: 0.1933
+Ep 3/40 (13s) lr=9.98e-04 | train_loss=0.0236 | val_f1=0.1884 P=0.1833 R=0.2412
+Ep 4/40 (13s) lr=9.93e-04 | train_loss=0.0232 | val_f1=0.1890 P=0.1813 R=0.2361
+Ep 5/40 (13s) lr=9.85e-04 | train_loss=0.0230 | val_f1=0.2000 P=0.1915 R=0.2513
+  ★ New best val_f1: 0.2000
+Ep 6/40 (13s) lr=9.73e-04 | train_loss=0.0228 | val_f1=0.1784 P=0.1903 R=0.2050
+Ep 7/40 (13s) lr=9.58e-04 | train_loss=0.0227 | val_f1=0.2163 P=0.2192 R=0.2596
+  ★ New best val_f1: 0.2163
+Ep 8/40 (12s) lr=9.40e-04 | train_loss=0.0225 | val_f1=0.2002 P=0.2217 R=0.2206
+Ep 9/40 (12s) lr=9.19e-04 | train_loss=0.0224 | val_f1=0.2088 P=0.2090 R=0.2477
+Ep 10/40 (13s) lr=8.95e-04 | train_loss=0.0223 | val_f1=0.3016 P=0.2992 R=0.3595
+  ★ New best val_f1: 0.3016
+Ep 11/40 (13s) lr=8.68e-04 | train_loss=0.0222 | val_f1=0.2979 P=0.2958 R=0.3520
+Ep 12/40 (13s) lr=8.39e-04 | train_loss=0.0222 | val_f1=0.3117 P=0.2836 R=0.4019
+  ★ New best val_f1: 0.3117
+Ep 13/40 (12s) lr=8.07e-04 | train_loss=0.0221 | val_f1=0.3307 P=0.3247 R=0.4003
+  ★ New best val_f1: 0.3307
+Ep 14/40 (12s) lr=7.73e-04 | train_loss=0.0219 | val_f1=0.2752 P=0.2980 R=0.3110
+Ep 15/40 (13s) lr=7.38e-04 | train_loss=0.0219 | val_f1=0.3097 P=0.3004 R=0.3900
+Ep 16/40 (13s) lr=7.01e-04 | train_loss=0.0218 | val_f1=0.3623 P=0.3972 R=0.4043
+  ★ New best val_f1: 0.3623
+Ep 17/40 (12s) lr=6.62e-04 | train_loss=0.0218 | val_f1=0.3210 P=0.2994 R=0.4091
+Ep 18/40 (13s) lr=6.23e-04 | train_loss=0.0217 | val_f1=0.3766 P=0.3614 R=0.4659
+  ★ New best val_f1: 0.3766
+Ep 19/40 (13s) lr=5.82e-04 | train_loss=0.0216 | val_f1=0.3701 P=0.3575 R=0.4494
+Ep 20/40 (12s) lr=5.41e-04 | train_loss=0.0216 | val_f1=0.3724 P=0.3636 R=0.4469
+Ep 21/40 (12s) lr=5.00e-04 | train_loss=0.0215 | val_f1=0.3802 P=0.3634 R=0.4645
+  ★ New best val_f1: 0.3802
+Ep 22/40 (12s) lr=4.59e-04 | train_loss=0.0215 | val_f1=0.3700 P=0.3506 R=0.4571
+Ep 23/40 (13s) lr=4.18e-04 | train_loss=0.0214 | val_f1=0.4029 P=0.3758 R=0.5015
+  ★ New best val_f1: 0.4029
+Ep 24/40 (13s) lr=3.77e-04 | train_loss=0.0214 | val_f1=0.3903 P=0.3795 R=0.4691
+Ep 25/40 (12s) lr=3.38e-04 | train_loss=0.0213 | val_f1=0.4093 P=0.3795 R=0.5124
+  ★ New best val_f1: 0.4093
+Ep 26/40 (12s) lr=2.99e-04 | train_loss=0.0213 | val_f1=0.3981 P=0.3836 R=0.4737
+Ep 27/40 (14s) lr=2.62e-04 | train_loss=0.0213 | val_f1=0.3967 P=0.3624 R=0.5063
+Ep 28/40 (12s) lr=2.27e-04 | train_loss=0.0212 | val_f1=0.4077 P=0.4086 R=0.4736
+Ep 29/40 (12s) lr=1.93e-04 | train_loss=0.0212 | val_f1=0.4258 P=0.4034 R=0.5231
+  ★ New best val_f1: 0.4258
+Ep 30/40 (13s) lr=1.61e-04 | train_loss=0.0212 | val_f1=0.4319 P=0.4344 R=0.4933
+  ★ New best val_f1: 0.4319
+Ep 31/40 (12s) lr=1.32e-04 | train_loss=0.0211 | val_f1=0.4179 P=0.4075 R=0.4981
+Ep 32/40 (12s) lr=1.05e-04 | train_loss=0.0211 | val_f1=0.3984 P=0.3847 R=0.4753
+Ep 33/40 (12s) lr=1.00e-04 | train_loss=0.0212 | val_f1=0.4205 P=0.4033 R=0.5111
+Ep 34/40 (12s) lr=1.00e-04 | train_loss=0.0211 | val_f1=0.4198 P=0.4136 R=0.4890
+Ep 35/40 (12s) lr=1.00e-04 | train_loss=0.0211 | val_f1=0.4134 P=0.3840 R=0.5122
+Ep 36/40 (12s) lr=1.00e-04 | train_loss=0.0211 | val_f1=0.4206 P=0.4000 R=0.5083
+Ep 37/40 (13s) lr=1.00e-04 | train_loss=0.0211 | val_f1=0.4231 P=0.4184 R=0.4926
+  Early stopping at epoch 37
+
+Done. Best val_f1=0.4319  ->  /userhome/cs/u3571824/fyp/muse/backend/ml_models/dev/vocal/outputs/onset_v3/best.pt
+
+
+
+
  .../outputs/onset_v3/best.pt: 100%|██████████|  416kB /  416kB            
+
+
  .../outputs/onset_v3/best.pt: 100%|██████████|  416kB /  416kB            
+
+
+
  .../outputs/onset_v3/best.pt: 100%|██████████|  416kB /  416kB            
+
+
  .../outputs/onset_v3/best.pt: 100%|██████████|  416kB /  416kB            
+
+
  .../outputs/onset_v3/best.pt: 100%|██████████|  416kB /  416kB            
+
+
  .../outputs/onset_v3/best.pt: 100%|██████████|  416kB /  416kB            
+No files have been modified since last commit. Skipping to prevent empty commit.
+Phase 3 complete — /userhome/cs/u3571824/fyp/muse/backend/ml_models/dev/vocal/outputs/onset_v3/best.pt