LiLT-RE-EN
Browse files- README.md +17 -27
- model.safetensors +1 -1
- training_args.bin +1 -1
README.md
CHANGED
@@ -21,10 +21,10 @@ should probably proofread and complete it, then remove this comment. -->
|
|
21 |
|
22 |
This model is a fine-tuned version of [nielsr/lilt-xlm-roberta-base](https://huggingface.co/nielsr/lilt-xlm-roberta-base) on the funsd_re dataset.
|
23 |
It achieves the following results on the evaluation set:
|
24 |
-
- Precision: 0.
|
25 |
-
- Recall: 0.
|
26 |
-
- F1: 0.
|
27 |
-
- Loss: 0.
|
28 |
|
29 |
## Model description
|
30 |
|
@@ -50,32 +50,22 @@ The following hyperparameters were used during training:
|
|
50 |
- optimizer: Adam with betas=(0.9,0.999) and epsilon=1e-08
|
51 |
- lr_scheduler_type: linear
|
52 |
- lr_scheduler_warmup_ratio: 0.1
|
53 |
-
- training_steps:
|
54 |
|
55 |
### Training results
|
56 |
|
57 |
-
| Training Loss | Epoch | Step
|
58 |
-
|
59 |
-
| 0.1604 | 26.32 | 500
|
60 |
-
| 0.1012 | 52.63 | 1000
|
61 |
-
| 0.0994 | 78.95 | 1500
|
62 |
-
| 0.0694 | 105.26 | 2000
|
63 |
-
| 0.0771 | 131.58 | 2500
|
64 |
-
| 0.0565 | 157.89 | 3000
|
65 |
-
| 0.0411 | 184.21 | 3500
|
66 |
-
| 0.0463 | 210.53 | 4000
|
67 |
-
| 0.0414 | 236.84 | 4500
|
68 |
-
| 0.036 | 263.16 | 5000
|
69 |
-
| 0.062 | 289.47 | 5500 | 0.3196 | 0.4802 | 0.3838 | 0.7425 |
|
70 |
-
| 0.0208 | 315.79 | 6000 | 0.3252 | 0.4778 | 0.387 | 0.4978 |
|
71 |
-
| 0.0245 | 342.11 | 6500 | 0.3229 | 0.4716 | 0.3833 | 0.7313 |
|
72 |
-
| 0.0274 | 368.42 | 7000 | 0.3275 | 0.4815 | 0.3898 | 0.5429 |
|
73 |
-
| 0.0232 | 394.74 | 7500 | 0.3262 | 0.4901 | 0.3917 | 0.4830 |
|
74 |
-
| 0.0207 | 421.05 | 8000 | 0.3162 | 0.4790 | 0.3810 | 0.5178 |
|
75 |
-
| 0.0346 | 447.37 | 8500 | 0.3274 | 0.4741 | 0.3873 | 0.4925 |
|
76 |
-
| 0.0229 | 473.68 | 9000 | 0.3219 | 0.4654 | 0.3806 | 0.5514 |
|
77 |
-
| 0.0253 | 500.0 | 9500 | 0.3261 | 0.4802 | 0.3884 | 0.5878 |
|
78 |
-
| 0.0186 | 526.32 | 10000 | 0.3202 | 0.4728 | 0.3819 | 0.5508 |
|
79 |
|
80 |
|
81 |
### Framework versions
|
|
|
21 |
|
22 |
This model is a fine-tuned version of [nielsr/lilt-xlm-roberta-base](https://huggingface.co/nielsr/lilt-xlm-roberta-base) on the funsd_re dataset.
|
23 |
It achieves the following results on the evaluation set:
|
24 |
+
- Precision: 0.3264
|
25 |
+
- Recall: 0.4864
|
26 |
+
- F1: 0.3907
|
27 |
+
- Loss: 0.4377
|
28 |
|
29 |
## Model description
|
30 |
|
|
|
50 |
- optimizer: Adam with betas=(0.9,0.999) and epsilon=1e-08
|
51 |
- lr_scheduler_type: linear
|
52 |
- lr_scheduler_warmup_ratio: 0.1
|
53 |
+
- training_steps: 5000
|
54 |
|
55 |
### Training results
|
56 |
|
57 |
+
| Training Loss | Epoch | Step | F1 | Validation Loss | Precision | Recall |
|
58 |
+
|:-------------:|:------:|:----:|:------:|:---------------:|:---------:|:------:|
|
59 |
+
| 0.1604 | 26.32 | 500 | 0 | 0.1513 | 0 | 0 |
|
60 |
+
| 0.1012 | 52.63 | 1000 | 0.0098 | 0.0786 | 0.5 | 0.0049 |
|
61 |
+
| 0.0994 | 78.95 | 1500 | 0.2518 | 0.1847 | 0.3729 | 0.1901 |
|
62 |
+
| 0.0694 | 105.26 | 2000 | 0.3499 | 0.1926 | 0.3667 | 0.3346 |
|
63 |
+
| 0.0771 | 131.58 | 2500 | 0.3856 | 0.3295 | 0.3450 | 0.4370 |
|
64 |
+
| 0.0565 | 157.89 | 3000 | 0.3865 | 0.4137 | 0.3293 | 0.4679 |
|
65 |
+
| 0.0411 | 184.21 | 3500 | 0.3808 | 0.3624 | 0.3252 | 0.4593 |
|
66 |
+
| 0.0463 | 210.53 | 4000 | 0.3832 | 0.5089 | 0.3221 | 0.4728 |
|
67 |
+
| 0.0414 | 236.84 | 4500 | 0.3911 | 0.6137 | 0.3305 | 0.4790 |
|
68 |
+
| 0.036 | 263.16 | 5000 | 0.3910 | 0.4428 | 0.3275 | 0.4852 |
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
69 |
|
70 |
|
71 |
### Framework versions
|
model.safetensors
CHANGED
@@ -1,3 +1,3 @@
|
|
1 |
version https://git-lfs.github.com/spec/v1
|
2 |
-
oid sha256:
|
3 |
size 1147300280
|
|
|
1 |
version https://git-lfs.github.com/spec/v1
|
2 |
+
oid sha256:8d34e1117ad326258b80a79ad96690f7ee2a91d871d63635aafbcab3c9c1a062
|
3 |
size 1147300280
|
training_args.bin
CHANGED
@@ -1,3 +1,3 @@
|
|
1 |
version https://git-lfs.github.com/spec/v1
|
2 |
-
oid sha256:
|
3 |
size 4664
|
|
|
1 |
version https://git-lfs.github.com/spec/v1
|
2 |
+
oid sha256:dee326b1a350900c4f55a05fc5857c3d142d66e2a6b1d0d4de6eaa4e9b7afd8c
|
3 |
size 4664
|