Update README.md
Browse files
README.md
CHANGED
@@ -20,9 +20,20 @@ We use [KoBEST benchmark](https://huggingface.co/datasets/skt/kobest_v1) dataset
|
|
20 |
| kullm-polyglot-5.8b | 0.5409 | 0.6072 | 0.5945 | 0.6345 | 0.6530
|
21 |
|
22 |
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
23 |
|
24 |
|
25 |
-
<img src=https://cdn-uploads.huggingface.co/production/uploads/650fecfd247f564485f8fbcf/7EKl1OAgKgPBFcSlGzBiW.png style="max-width: 800px; width: 400%"/>
|
26 |
|
27 |
|
28 |
## Training hyperparameters
|
@@ -36,7 +47,11 @@ We use [KoBEST benchmark](https://huggingface.co/datasets/skt/kobest_v1) dataset
|
|
36 |
- lr_scheduler_type: linear
|
37 |
- num_epochs: 2.0
|
38 |
|
39 |
-
|
|
|
|
|
|
|
|
|
40 |
|
41 |
## Citation
|
42 |
|
|
|
20 |
| kullm-polyglot-5.8b | 0.5409 | 0.6072 | 0.5945 | 0.6345 | 0.6530
|
21 |
|
22 |
|
23 |
+
### Evaluation results
|
24 |
+
COPA (F1)
|
25 |
+
<img src=https://cdn-uploads.huggingface.co/production/uploads/650fecfd247f564485f8fbcf/7EKl1OAgKgPBFcSlGzBiW.png style="max-width: 800px; width: 400%"/>
|
26 |
+
|
27 |
+
HellaSwag (F1)
|
28 |
+
|
29 |
+
BoolQ (F1)
|
30 |
+
|
31 |
+
SentiNeg (F1)
|
32 |
+
|
33 |
+
WiC (F1)
|
34 |
+
|
35 |
|
36 |
|
|
|
37 |
|
38 |
|
39 |
## Training hyperparameters
|
|
|
47 |
- lr_scheduler_type: linear
|
48 |
- num_epochs: 2.0
|
49 |
|
50 |
+
## Framework versions
|
51 |
+
Transformers 4.30.2
|
52 |
+
Pytorch 2.0.1+cu117
|
53 |
+
Datasets 2.11.0
|
54 |
+
deepspeed 0.9.5
|
55 |
|
56 |
## Citation
|
57 |
|