lijn14 commited on
Commit
2b2147f
·
1 Parent(s): 0cda4b5

训练log

Browse files
Files changed (2) hide show
  1. result/link.txt +1 -0
  2. result/train_output.txt +102 -0
result/link.txt ADDED
@@ -0,0 +1 @@
 
 
1
+ https://colab.research.google.com/drive/1wCVTnNuULJkBIS5tzPu5uQY4zd-akEd4?authuser=1#scrollTo=SssPN0dpTHJX&fullscreenOutput=true
result/train_output.txt ADDED
@@ -0,0 +1,102 @@
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
1
+ ============================================================
2
+ Starting Training...
3
+ ============================================================
4
+ [05/13/26 07:55:00] INFO Starting training for 30 epochs trainer.py:193
5
+
6
+ INFO Device: cuda, FP16: False, BF16: True trainer.py:194
7
+
8
+ INFO Gradient accumulation steps: 4 trainer.py:195
9
+
10
+ INFO ================================================== trainer.py:199
11
+
12
+ INFO Epoch 1/30 trainer.py:200
13
+
14
+ [05/13/26 07:57:11] INFO Epoch 1 - Train Loss: 8.2498 trainer.py:279
15
+
16
+ Evaluating: 100%|██████████| 122/122 [17:49<00:00, 8.76s/it]
17
+ [05/13/26 08:15:01] WARNING Evaluation failed during validation: references cannot be empty trainer.py:311
18
+
19
+ INFO Epoch 1 - Val Loss: 7.9030 trainer.py:313
20
+
21
+ [05/13/26 08:15:02] INFO Checkpoint saved: checkpoints/checkpoint_epoch_1.pt trainer.py:341
22
+
23
+ [05/13/26 08:15:03] INFO New best model saved: checkpoints/best_model.pt (metric: 7.9030) trainer.py:352
24
+
25
+ INFO No improvement for 1 epochs (best: 7.9030, current: 7.9030) trainer.py:398
26
+
27
+ INFO ================================================== trainer.py:199
28
+
29
+ INFO Epoch 2/30 trainer.py:200
30
+
31
+ [05/13/26 08:17:12] INFO Epoch 2 - Train Loss: 6.6109 trainer.py:279
32
+
33
+ Evaluating: 100%|██████████| 122/122 [06:10<00:00, 3.04s/it]
34
+ [05/13/26 08:23:23] WARNING Evaluation failed during validation: references cannot be empty trainer.py:311
35
+
36
+ INFO Epoch 2 - Val Loss: 6.9446 trainer.py:313
37
+
38
+ [05/13/26 08:23:24] INFO Checkpoint saved: checkpoints/checkpoint_epoch_2.pt trainer.py:341
39
+
40
+ INFO No improvement for 2 epochs (best: 7.9030, current: 6.9446) trainer.py:398
41
+
42
+ INFO ================================================== trainer.py:199
43
+
44
+ INFO Epoch 3/30 trainer.py:200
45
+
46
+ [05/13/26 08:25:33] INFO Epoch 3 - Train Loss: 5.2802 trainer.py:279
47
+
48
+ Evaluating: 100%|██████████| 122/122 [06:55<00:00, 3.40s/it]
49
+ [05/13/26 08:32:29] WARNING Evaluation failed during validation: references cannot be empty trainer.py:311
50
+
51
+ INFO Epoch 3 - Val Loss: 6.0977 trainer.py:313
52
+
53
+ INFO Checkpoint saved: checkpoints/checkpoint_epoch_3.pt trainer.py:341
54
+
55
+ INFO No improvement for 3 epochs (best: 7.9030, current: 6.0977) trainer.py:398
56
+
57
+ INFO ================================================== trainer.py:199
58
+
59
+ INFO Epoch 4/30 trainer.py:200
60
+
61
+ Train Epoch 4: 28%|██▊ | 1687/6104 [00:35<01:30, 48.68it/s, loss=4.5668, lr=3.00e-04, step=5000][05/13/26 08:33:06] INFO Checkpoint saved: checkpoints/checkpoint_step_5000.pt trainer.py:341
62
+
63
+ INFO New best model saved: checkpoints/best_model.pt (metric: inf) trainer.py:352
64
+
65
+ [05/13/26 08:34:39] INFO Epoch 4 - Train Loss: 4.4749 trainer.py:279
66
+
67
+ Evaluating: 100%|██████████| 122/122 [03:08<00:00, 1.55s/it]
68
+ [05/13/26 08:37:49] WARNING Evaluation failed during validation: references cannot be empty trainer.py:311
69
+
70
+ INFO Epoch 4 - Val Loss: 5.6509 trainer.py:313
71
+
72
+ [05/13/26 08:37:50] INFO Checkpoint saved: checkpoints/checkpoint_epoch_4.pt trainer.py:341
73
+
74
+ INFO No improvement for 4 epochs (best: inf, current: 5.6509) trainer.py:398
75
+
76
+ INFO ================================================== trainer.py:199
77
+
78
+ INFO Epoch 5/30 trainer.py:200
79
+
80
+ [05/13/26 08:40:00] INFO Epoch 5 - Train Loss: 4.0513 trainer.py:279
81
+
82
+ Evaluating: 100%|██████████| 122/122 [03:28<00:00, 1.71s/it]
83
+ [05/13/26 08:43:29] WARNING Evaluation failed during validation: references cannot be empty trainer.py:311
84
+
85
+ INFO Epoch 5 - Val Loss: 5.4433 trainer.py:313
86
+
87
+ INFO Checkpoint saved: checkpoints/checkpoint_epoch_5.pt trainer.py:341
88
+
89
+ INFO No improvement for 5 epochs (best: inf, current: 5.4433) trainer.py:398
90
+
91
+ INFO Early stopping triggered at epoch 5 trainer.py:212
92
+
93
+ INFO ================================================== trainer.py:431
94
+
95
+ INFO Training completed! trainer.py:432
96
+
97
+ INFO Best epoch: 4 trainer.py:433
98
+
99
+ INFO Best bleu: inf trainer.py:434
100
+
101
+ INFO Training summary saved to checkpoints/training_summary.json trainer.py:447
102
+