Update README.md
Browse files
README.md
CHANGED
|
@@ -3,7 +3,16 @@ language: ru
|
|
| 3 |
license: apache-2.0
|
| 4 |
tags:
|
| 5 |
- transformers
|
|
|
|
| 6 |
widget:
|
| 7 |
- text: "Он решил ту или иную сложную задачу."
|
| 8 |
---
|
| 9 |
-
This is a finetuned version of [RuRoBERTa-large](https://huggingface.co/sberbank-ai/ruRoberta-large) for the task of linguistic acceptability classification on the [RuCoLA](https://rucola-benchmark.com/) benchmark.
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
| 3 |
license: apache-2.0
|
| 4 |
tags:
|
| 5 |
- transformers
|
| 6 |
+
thumbnail: "https://github.com/RussianNLP/RuCoLA/blob/main/logo.png"
|
| 7 |
widget:
|
| 8 |
- text: "Он решил ту или иную сложную задачу."
|
| 9 |
---
|
| 10 |
+
This is a finetuned version of [RuRoBERTa-large](https://huggingface.co/sberbank-ai/ruRoberta-large) for the task of linguistic acceptability classification on the [RuCoLA](https://rucola-benchmark.com/) benchmark.
|
| 11 |
+
|
| 12 |
+
The hyperparameters used for finetuning are as follows:
|
| 13 |
+
* 5 training epochs (with early stopping based on validation MCC)
|
| 14 |
+
* Peak learning rate: 1e-5, linear warmup for 10% of total training time
|
| 15 |
+
* Weight decay: 1e-4
|
| 16 |
+
* Batch size: 32
|
| 17 |
+
* Random seed: 5
|
| 18 |
+
* Optimizer: [torch.optim.AdamW](https://pytorch.org/docs/stable/generated/torch.optim.AdamW.html)
|