Update finetune_sample/README.md
Browse files- finetune_sample/README.md +18 -4
finetune_sample/README.md
CHANGED
@@ -3,15 +3,29 @@ finetune sample data and script.
|
|
3 |
```
|
4 |
pip install transformers
|
5 |
pip install datasets
|
6 |
-
pip install peft
|
7 |
pip install trl
|
8 |
pip install auto-gptq
|
9 |
pip install optimum
|
10 |
```
|
11 |
|
12 |
-
|
13 |
-
|
14 |
-
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
15 |
|
16 |
- finetune.py gptq finetune sample file.
|
17 |
- jawiki3.csv sample data.(Japanese)
|
|
|
3 |
```
|
4 |
pip install transformers
|
5 |
pip install datasets
|
6 |
+
pip install peft==0.5.0
|
7 |
pip install trl
|
8 |
pip install auto-gptq
|
9 |
pip install optimum
|
10 |
```
|
11 |
|
12 |
+
Version is very very important.
|
13 |
+
For example if you get something like
|
14 |
+
```
|
15 |
+
ValueError: Target module QuantLinear() is not supported. Currently, only `torch.nn.Linear` and `Conv1D` are supported.
|
16 |
+
```
|
17 |
+
It's because peft old version.
|
18 |
+
|
19 |
+
I don't know if it's required, but the version of my running environment.
|
20 |
+
* auto-gptq 0.4.1+cu117
|
21 |
+
* trl 0.7.1
|
22 |
+
* optimum 1.12.1.dev0
|
23 |
+
* pip install transformers 4.32.1
|
24 |
+
* datasets 2.14.4
|
25 |
+
|
26 |
+
The documentation says to install from source, but sometimes that causes errors.
|
27 |
+
If you can't get it to work, it might be better to wait until the stable version comes out.
|
28 |
+
Good luck!
|
29 |
|
30 |
- finetune.py gptq finetune sample file.
|
31 |
- jawiki3.csv sample data.(Japanese)
|