dahara1 commited on
Commit
e9f5656
1 Parent(s): d48e607

Update finetune_sample/README.md

Browse files
Files changed (1) hide show
  1. finetune_sample/README.md +17 -5
finetune_sample/README.md CHANGED
@@ -1,12 +1,25 @@
1
- Lora finetune sample data and script.
2
 
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.
@@ -20,10 +33,9 @@ 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
- * 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
 
@@ -31,7 +43,7 @@ Good luck!
31
  - finetune.py gptq finetune sample file.
32
  - jawiki3.csv sample data.(Japanese)
33
  - lora_test.py after finetune, you can use lora with this script.
34
- - checkpoint-700 Created sample LoRA
35
 
36
  The model.safetensors is ../gptq_model-4bit-128g.safetensors.
37
  It's samefile. I can't find how to change script defaults model name, So I copied it.
 
1
+ autoGPTQ QLora finetune sample data and script.
2
 
3
  ```
4
+ git lfs install
5
+ git clone https://huggingface.co/dahara1/weblab-10b-instruction-sft-GPTQ
6
+ cd weblab-10b-instruction-sft-GPTQ/finetune_sample
7
+ python3 -m venv gptq_finetune
8
+ source gptq_finetune/bin/activate
9
+
10
+
11
+ pip install transformers==4.32.1
12
  pip install datasets
13
  pip install peft==0.5.0
14
  pip install trl
15
  pip install auto-gptq
16
  pip install optimum
17
+
18
+ # finetune qlora
19
+ python3 finetune.py
20
+
21
+ # use qlora sample
22
+ python3 lora_test.py
23
  ```
24
 
25
  Version is very very important.
 
33
  * auto-gptq 0.4.1+cu117
34
  * trl 0.7.1
35
  * optimum 1.12.1.dev0
 
36
  * datasets 2.14.4
37
 
38
+ The official documentation says to install from source, but sometimes that causes errors.
39
  If you can't get it to work, it might be better to wait until the stable version comes out.
40
  Good luck!
41
 
 
43
  - finetune.py gptq finetune sample file.
44
  - jawiki3.csv sample data.(Japanese)
45
  - lora_test.py after finetune, you can use lora with this script.
46
+ - checkpoint-700 Created sample LoRA for test.
47
 
48
  The model.safetensors is ../gptq_model-4bit-128g.safetensors.
49
  It's samefile. I can't find how to change script defaults model name, So I copied it.