File size: 506 Bytes
66e7f4a
 
 
 
 
 
 
 
 
 
 
 
 
 
039ea66
 
66e7f4a
 
 
 
 
 
 
 
ef45ae6
 
1
2
3
4
5
6
7
8
9
10
11
12
13
14
15
16
17
18
19
20
21
22
23
24
25
26
27
---
language: ko
tags:
- bart
license: mit
---

## KoBART-base-v2

With the addition of chatting data, the model is trained to handle the semantics of sequences longer than KoBART.

```python
from transformers import PreTrainedTokenizerFast, BartModel

tokenizer = PreTrainedTokenizerFast.from_pretrained('hyunwoongko/kobart')
model = BartModel.from_pretrained('hyunwoongko/kobart')
```

### Performance 

NSMC
- acc. : 0.901

### hyunwoongko/kobart
- Added bos/eos post processor
- Removed token_type_ids