search

westlake-repl / IDvs.MoRec

End-to-end Training for Multimodal Recommendation Systems
Apache License 2.0
117 stars 15 forks source link

Fine-tuning Script #1

Open BoPeng112 opened 1 year ago

BoPeng112 commented 1 year ago

Thanks for sharing this interesting work. I was wondering if you are going to share the scripts for fine-tuning LLMs.

zyuanat commented 1 year ago

Thank you for your attention and sorry for the late reply.

We have released the code and script for training and testing. Take the End2end MoRec for text recommendation (Bert-base + SASRec) as an example, you can run 'cd bce_text/main-end2end', then 'python train_bert_base.py'. The files looked like 'train_xxx.py' are the scripts, you can change the parameters in these files, the details maybe need to be adjusted based on your GPU resources and Python environment.

It is worth noting that "LLMs" usually refer to models with comparable parameters to GPT-3, much larger than the Bert and Roberta used in our paper. Due to the GPU limitation and the challenges of fine-tuning, LLMs are not considered in our paper. How to apply End2end LLMs to Recsys is a valuable research direction.