Update README.md
Browse files
README.md
CHANGED
|
@@ -4,14 +4,11 @@ tags:
|
|
| 4 |
- glycebert
|
| 5 |
inference: False
|
| 6 |
---
|
| 7 |
-
|
| 8 |
-
|
| 9 |
-
## https://github.com/JunnYu/GlyceBert_pytorch
|
| 10 |
-
|
| 11 |
-
本项目主要自定义了tokenization_glycebert_fast.py文件中的GlyceBertTokenizerFast代码。从而可以从huggingface.co调用。
|
| 12 |
```python
|
| 13 |
pretrained_tokenizer_name = "junnyu/ChineseBERT-base"
|
| 14 |
-
tokenizer =
|
| 15 |
```
|
| 16 |
|
| 17 |
# Paper
|
|
@@ -20,32 +17,29 @@ tokenizer = GlyceBertTokenizerFast.from_pretrained(pretrained_tokenizer_name)
|
|
| 20 |
|
| 21 |
# Install
|
| 22 |
```bash
|
| 23 |
-
pip install
|
| 24 |
or
|
| 25 |
-
pip install git+https://github.com/JunnYu/
|
| 26 |
```
|
| 27 |
|
| 28 |
# Usage
|
| 29 |
```python
|
| 30 |
import torch
|
| 31 |
-
from transformers import BertConfig as
|
|
|
|
| 32 |
|
| 33 |
-
from glycebert import GlyceBertForMaskedLM, GlyceBertTokenizerFast
|
| 34 |
-
|
| 35 |
-
# 使用我这个里面的tokenizer config和model config
|
| 36 |
pretrained_tokenizer_name = "junnyu/ChineseBERT-base"
|
| 37 |
pretrained_model_name = "ShannonAI/ChineseBERT-base"
|
| 38 |
|
| 39 |
-
tokenizer =
|
| 40 |
-
config =
|
| 41 |
-
chinese_bert =
|
| 42 |
-
pretrained_model_name, config=config
|
| 43 |
-
)
|
| 44 |
|
| 45 |
text = "北京是[MASK]国的首都。"
|
| 46 |
inputs = tokenizer(text, return_tensors="pt")
|
| 47 |
print(inputs)
|
| 48 |
maskpos = 4
|
|
|
|
| 49 |
with torch.no_grad():
|
| 50 |
o = chinese_bert(**inputs)
|
| 51 |
value, index = o.logits.softmax(-1)[0, maskpos].topk(10)
|
|
@@ -63,4 +57,4 @@ print(outputs)
|
|
| 63 |
```
|
| 64 |
|
| 65 |
# Reference
|
| 66 |
-
https://github.com/ShannonAI/ChineseBert
|
|
|
|
| 4 |
- glycebert
|
| 5 |
inference: False
|
| 6 |
---
|
| 7 |
+
# ChineseBert_pytorch
|
| 8 |
+
本项目主要自定义了tokenization_chinesebert_fast.py文件中的ChineseBertTokenizerFast代码。从而可以从huggingface.co调用。
|
|
|
|
|
|
|
|
|
|
| 9 |
```python
|
| 10 |
pretrained_tokenizer_name = "junnyu/ChineseBERT-base"
|
| 11 |
+
tokenizer = ChineseBertTokenizerFast.from_pretrained(pretrained_tokenizer_name)
|
| 12 |
```
|
| 13 |
|
| 14 |
# Paper
|
|
|
|
| 17 |
|
| 18 |
# Install
|
| 19 |
```bash
|
| 20 |
+
pip install chinesebert
|
| 21 |
or
|
| 22 |
+
pip install git+https://github.com/JunnYu/ChineseBert_pytorch.git
|
| 23 |
```
|
| 24 |
|
| 25 |
# Usage
|
| 26 |
```python
|
| 27 |
import torch
|
| 28 |
+
from transformers import BertConfig as ChineseBertConfig
|
| 29 |
+
from chinesebert import ChineseBertForMaskedLM, ChineseBertTokenizerFast
|
| 30 |
|
|
|
|
|
|
|
|
|
|
| 31 |
pretrained_tokenizer_name = "junnyu/ChineseBERT-base"
|
| 32 |
pretrained_model_name = "ShannonAI/ChineseBERT-base"
|
| 33 |
|
| 34 |
+
tokenizer = ChineseBertTokenizerFast.from_pretrained(pretrained_tokenizer_name)
|
| 35 |
+
config = ChineseBertConfig.from_pretrained(pretrained_tokenizer_name)
|
| 36 |
+
chinese_bert = ChineseBertForMaskedLM.from_pretrained(pretrained_model_name, config=config)
|
|
|
|
|
|
|
| 37 |
|
| 38 |
text = "北京是[MASK]国的首都。"
|
| 39 |
inputs = tokenizer(text, return_tensors="pt")
|
| 40 |
print(inputs)
|
| 41 |
maskpos = 4
|
| 42 |
+
|
| 43 |
with torch.no_grad():
|
| 44 |
o = chinese_bert(**inputs)
|
| 45 |
value, index = o.logits.softmax(-1)[0, maskpos].topk(10)
|
|
|
|
| 57 |
```
|
| 58 |
|
| 59 |
# Reference
|
| 60 |
+
https://github.com/ShannonAI/ChineseBert
|