YixuanWeng
commited on
Commit
·
c1ce8d9
1
Parent(s):
9b086ad
Update README.md
Browse files
README.md
CHANGED
|
@@ -1,3 +1,24 @@
|
|
| 1 |
---
|
| 2 |
license: mit
|
| 3 |
---
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
| 1 |
---
|
| 2 |
license: mit
|
| 3 |
---
|
| 4 |
+
|
| 5 |
+
现有的少数民族语言预训练模型仍然较为稀缺,尽管国内少数民族语言模型CINO具有较强的理解能力,但仍然缺乏面向生成与翻译领域的研究。
|
| 6 |
+
CMPT (Chinese Minority Pre-Trained Language Model) 是在BART的基础上,加入DeepNorm预训练的超深层生成模型。其最大具有128+128层。其在超过10G的汉英维藏蒙语料中进行受限预训练。其具有较强的理解与生成性能。
|
| 7 |
+
|
| 8 |
+
**Github Link:** https://github.com/WENGSYX/CMPT
|
| 9 |
+
|
| 10 |
+
## Usage
|
| 11 |
+
|
| 12 |
+
|
| 13 |
+
|
| 14 |
+
|
| 15 |
+
```python
|
| 16 |
+
>>> from modeling_cmpt import CMPTForCir
|
| 17 |
+
>>> from transformers import AutoTokenizer
|
| 18 |
+
>>> tokenizer = AutoTokenizer.from_pretrained('./CMTP')
|
| 19 |
+
>>> model = CMPTForCir.from_pretrained('./CMTP')
|
| 20 |
+
>>> inputs = tokenizer.encode("Hello world, 你好 世界", return_tensors='pt')
|
| 21 |
+
>>> pred_ids = model.generate(input_ids, num_beams=4, max_length=20)
|
| 22 |
+
>>> print(tokenizer.convert_ids_to_tokens(pred_ids[i]))
|
| 23 |
+
|
| 24 |
+
```
|