LLM: add alpaca qlora finetuning example #9276
Conversation
|
This is the original code https://github.com/tloen/alpaca-lora/blob/main/finetune.py, maybe we should also add their license. |
python/llm/example/GPU/QLoRA-FineTuning/alpaca-qlora/alpaca_qlora_finetuning.py
Outdated
Show resolved
Hide resolved
python/llm/example/GPU/QLoRA-FineTuning/alpaca-qlora/alpaca_qlora_finetuning.py
Outdated
Show resolved
Hide resolved
plusbang
force-pushed
the
add-alpaca-qlora-example
branch
from
bc68139 to
61b3752
Compare
yeah, have added. |
|
Please @jason-dai take a look. Example usage on single Arc is provided. |
|
@jason-dai Now, the minimal necessary components to run Alpaca QloRA are updated, but I'm not sure whether need to add license comments in the JSON files of templates folder. |
Add comment "This file is copied from ..." |
OK, have added comment in JSON files. |
| # you can install specific ipex/torch version for your need | ||
| pip install --pre --upgrade bigdl-llm[xpu] -f https://developer.intel.com/ipex-whl-stable-xpu | ||
| pip install transformers==4.34.0 | ||
| pip install fire datasets peft==0.5.0 |
There was a problem hiding this comment.
add pip install accelerate==0.23.0 ?
There was a problem hiding this comment.
add
pip install accelerate==0.23.0?
Following the provided commands, accelerate==0.21.0 is installed and experiments work fine. Is it necessary?
| @@ -0,0 +1,7 @@ | |||
| { | |||
There was a problem hiding this comment.
add comment for where this file is copied from
There was a problem hiding this comment.
add comment for where this file is copied from
Have updated.
Description
(According to comparison in https://github.com/analytics-zoo/nano/issues/706#issuecomment-1793989686, use ["q_proj","v_proj","k_proj","o_proj","up_proj","down_proj","gate_proj"] as default target modules)
4. How to test?