version 2
Browse files
README.md
CHANGED
@@ -18,8 +18,8 @@ should probably proofread and complete it, then remove this comment. -->
|
|
18 |
|
19 |
This model is a fine-tuned version of [facebook/bart-large](https://huggingface.co/facebook/bart-large) on the None dataset.
|
20 |
It achieves the following results on the evaluation set:
|
21 |
-
- Loss:
|
22 |
-
- Sacrebleu:
|
23 |
|
24 |
## Model description
|
25 |
|
@@ -39,50 +39,22 @@ More information needed
|
|
39 |
|
40 |
The following hyperparameters were used during training:
|
41 |
- learning_rate: 2e-05
|
42 |
-
- train_batch_size:
|
43 |
-
- eval_batch_size:
|
44 |
- seed: 42
|
45 |
-
- gradient_accumulation_steps:
|
46 |
-
- total_train_batch_size:
|
47 |
- optimizer: Adam with betas=(0.9,0.999) and epsilon=1e-08
|
48 |
- lr_scheduler_type: linear
|
49 |
-
- num_epochs:
|
50 |
- mixed_precision_training: Native AMP
|
51 |
|
52 |
### Training results
|
53 |
|
54 |
| Training Loss | Epoch | Step | Validation Loss | Sacrebleu |
|
55 |
|:-------------:|:-----:|:----:|:---------------:|:---------:|
|
56 |
-
| No log |
|
57 |
-
| No log |
|
58 |
-
| No log | 3.0 | 656 | 0.5662 | 66.8166 |
|
59 |
-
| No log | 4.0 | 875 | 0.6339 | 66.8911 |
|
60 |
-
| No log | 5.0 | 1093 | 0.6190 | 66.4260 |
|
61 |
-
| No log | 6.0 | 1312 | 0.6760 | 66.7698 |
|
62 |
-
| No log | 7.0 | 1531 | 0.6708 | 66.7328 |
|
63 |
-
| No log | 8.0 | 1750 | 0.7686 | 66.6153 |
|
64 |
-
| No log | 9.0 | 1968 | 0.7157 | 66.7670 |
|
65 |
-
| No log | 10.0 | 2187 | 0.7567 | 66.6510 |
|
66 |
-
| No log | 11.0 | 2406 | 0.7699 | 66.5710 |
|
67 |
-
| No log | 12.0 | 2625 | 0.8145 | 66.7658 |
|
68 |
-
| No log | 13.0 | 2843 | 0.8292 | 66.4557 |
|
69 |
-
| No log | 14.0 | 3062 | 0.8610 | 66.7477 |
|
70 |
-
| No log | 15.0 | 3281 | 0.8962 | 66.4487 |
|
71 |
-
| No log | 16.0 | 3500 | 0.9000 | 66.6798 |
|
72 |
-
| No log | 17.0 | 3718 | 0.9376 | 66.5672 |
|
73 |
-
| No log | 18.0 | 3937 | 0.8907 | 66.6538 |
|
74 |
-
| No log | 19.0 | 4156 | 0.8829 | 66.5278 |
|
75 |
-
| No log | 20.0 | 4375 | 0.9925 | 66.5495 |
|
76 |
-
| No log | 21.0 | 4593 | 0.9656 | 66.5410 |
|
77 |
-
| No log | 22.0 | 4812 | 0.9721 | 66.4741 |
|
78 |
-
| No log | 23.0 | 5031 | 0.9778 | 66.6736 |
|
79 |
-
| No log | 24.0 | 5250 | 1.0032 | 66.5801 |
|
80 |
-
| No log | 25.0 | 5468 | 1.0808 | 66.6122 |
|
81 |
-
| No log | 26.0 | 5687 | 1.0403 | 66.7292 |
|
82 |
-
| No log | 27.0 | 5906 | 1.0388 | 66.5946 |
|
83 |
-
| No log | 28.0 | 6125 | 1.0707 | 66.6240 |
|
84 |
-
| No log | 29.0 | 6343 | 1.0356 | 66.7184 |
|
85 |
-
| No log | 29.9 | 6540 | 1.0664 | 66.6476 |
|
86 |
|
87 |
|
88 |
### Framework versions
|
|
|
18 |
|
19 |
This model is a fine-tuned version of [facebook/bart-large](https://huggingface.co/facebook/bart-large) on the None dataset.
|
20 |
It achieves the following results on the evaluation set:
|
21 |
+
- Loss: 0.6032
|
22 |
+
- Sacrebleu: 65.7419
|
23 |
|
24 |
## Model description
|
25 |
|
|
|
39 |
|
40 |
The following hyperparameters were used during training:
|
41 |
- learning_rate: 2e-05
|
42 |
+
- train_batch_size: 32
|
43 |
+
- eval_batch_size: 32
|
44 |
- seed: 42
|
45 |
+
- gradient_accumulation_steps: 8
|
46 |
+
- total_train_batch_size: 256
|
47 |
- optimizer: Adam with betas=(0.9,0.999) and epsilon=1e-08
|
48 |
- lr_scheduler_type: linear
|
49 |
+
- num_epochs: 2
|
50 |
- mixed_precision_training: Native AMP
|
51 |
|
52 |
### Training results
|
53 |
|
54 |
| Training Loss | Epoch | Step | Validation Loss | Sacrebleu |
|
55 |
|:-------------:|:-----:|:----:|:---------------:|:---------:|
|
56 |
+
| No log | 0.99 | 54 | 0.5848 | 65.0804 |
|
57 |
+
| No log | 1.97 | 108 | 0.6032 | 65.7419 |
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
58 |
|
59 |
|
60 |
### Framework versions
|