update readme
Browse files
README.md
CHANGED
@@ -13,7 +13,7 @@ size_categories:
|
|
13 |
|
14 |
## 简介
|
15 |
|
16 |
-
搜集网络上的网文小说,清洗,分割后,用于训练大语言模型,共计9000本左右,大约
|
17 |
|
18 |
## 使用
|
19 |
|
@@ -32,34 +32,6 @@ size_categories:
|
|
32 |
|
33 |
```
|
34 |
|
35 |
-
### 示例代码
|
36 |
-
|
37 |
-
```python
|
38 |
-
def process_webnovel(input_dir, tokenizer):
|
39 |
-
for subdir, dirs, files in os.walk(input_dir):
|
40 |
-
all_tokens = []
|
41 |
-
for idx, file in enumerate(files):
|
42 |
-
# 只处理txt文件
|
43 |
-
if file.endswith('.jsonl'):
|
44 |
-
# 获取当前文件的绝对路径
|
45 |
-
file_path = os.path.join(subdir, file)
|
46 |
-
# 读取jsonl文件
|
47 |
-
with open(file_path, 'r', encoding='utf-8') as infile:
|
48 |
-
lines = infile.readlines()
|
49 |
-
|
50 |
-
for line in lines:
|
51 |
-
json_obj = json.loads(line) # 解析json字符串为python对象
|
52 |
-
text = json_obj['text']
|
53 |
-
tokens = tokenizer.encode(text, add_special_tokens=False)
|
54 |
-
tokens.append(tokenizer.special_tokens['<eos>'])
|
55 |
-
if len(tokens) > 5:
|
56 |
-
all_tokens += tokens
|
57 |
-
|
58 |
-
arr = np.array(all_tokens, dtype = np.uint16)
|
59 |
-
with open('./data/webnovel_{idx}.bin','wb') as f:
|
60 |
-
f.write(arr.tobytes())
|
61 |
-
```
|
62 |
-
|
63 |
|
64 |
|
65 |
|
|
|
13 |
|
14 |
## 简介
|
15 |
|
16 |
+
搜集网络上的网文小说,清洗,分割后,用于训练大语言模型,共计9000本左右,大约9B左右token。
|
17 |
|
18 |
## 使用
|
19 |
|
|
|
32 |
|
33 |
```
|
34 |
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
35 |
|
36 |
|
37 |
|