Commit
•
01d8f26
1
Parent(s):
e0c19fd
Update README.md
Browse files
README.md
CHANGED
@@ -15,10 +15,30 @@ tags:
|
|
15 |
|
16 |
## 模型:THUDM/chatglm2-6b-int4: [chat_glm2](https://huggingface.co/THUDM/chatglm2-6b-int4)
|
17 |
|
18 |
-
|
|
|
|
|
19 |
|
20 |
-
|
|
|
|
|
|
|
|
|
|
|
|
|
21 |
|
22 |
-
Tuning
|
23 |
|
24 |
-
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
15 |
|
16 |
## 模型:THUDM/chatglm2-6b-int4: [chat_glm2](https://huggingface.co/THUDM/chatglm2-6b-int4)
|
17 |
|
18 |
+
* base model: chatglm2-6b-int4
|
19 |
+
* pretrained_ckpt: facebook/THUDM/chatglm2-6b-int4
|
20 |
+
* This model was trained for [rpy dl translate](https://github.com/O5-7/rpy_dl_translate)
|
21 |
|
22 |
+
## Model description
|
23 |
+
* source group: English
|
24 |
+
* target group: Chinese
|
25 |
+
* model: transformer
|
26 |
+
* source language(s): eng
|
27 |
+
* target language(s): cjy_Hans cjy_Hant cmn cmn_Hans cmn_Hant gan lzh lzh_Hans nan wuu yue yue_Hans yue_Hant
|
28 |
+
* fine_tune: On the basis of mbart-large-50-one-to-many-mmt checkpoints, train English original text with renpy text features (including but not limited to {i} [text] {/i}) to Chinese with the same reserved flag, as well as training for English name retention for LIL
|
29 |
|
30 |
+
## Fine Tuning
|
31 |
|
32 |
+
* 这是chatglm2-6b-int4通过P-Tuning v2微调后的PrefixEncoder部分的参数,不含有chatglm2-6b-int4本体
|
33 |
+
* This is the parameter of the PrefixEncoder section of chatglm2-6b-int4 after fine-tuning through P-Tuning v2, without the chatglm2-6b-int4 body
|
34 |
+
* Tuning脚本:[ChatGLM2-6B/ptuning](https://github.com/THUDM/ChatGLM2-6B/tree/main/ptuning)
|
35 |
+
* 此微调未经过测试,请自行测试
|
36 |
+
|
37 |
+
## How to use
|
38 |
+
|
39 |
+
* https://github.com/THUDM/ChatGLM2-6B/blob/main/ptuning/README.md
|
40 |
+
|
41 |
+
## Contact
|
42 |
+
|
43 |
+
* 517205163@qq.com
|
44 |
+
* a4564563@gmail.com
|