matejklemen
commited on
Commit
•
d6f143f
1
Parent(s):
551223b
Update README.md
Browse files
README.md
CHANGED
@@ -3,4 +3,15 @@ language:
|
|
3 |
- sl
|
4 |
|
5 |
license: cc-by-sa-4.0
|
6 |
-
---
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
3 |
- sl
|
4 |
|
5 |
license: cc-by-sa-4.0
|
6 |
+
---
|
7 |
+
|
8 |
+
# sloberta-si-nli
|
9 |
+
|
10 |
+
SloBERTa model finetuned on the SI-NLI dataset for Slovene natural language inference.
|
11 |
+
Fine-tuned in a classic sequence pair classification setting on the official training/validation/test split for 10 epochs, using validation set accuracy for model selection.
|
12 |
+
Optimized using the AdamW optimizer (learning rate 2e-5) and cross-entropy loss.
|
13 |
+
Using batch size `82` (selected based on the available GPU memory) and maximum sequence length `102` (99th percentile of the lengths in the training set).
|
14 |
+
|
15 |
+
Achieves the following metrics:
|
16 |
+
- best validation accuracy: `0.720`
|
17 |
+
- test accuracy = `0.735`
|