Extending LaikaLLM
August 02, 2024
October 17, 2024
LaikaLLM is a software, for researchers, that helps in setting up a repeatable, reproducible, replicable protocol for training and evaluating multitask LLM for recommendation!
- Two different model family implemented at the moment of writing (T5 and GPT2)
- Fully vectorized Ranking (NDCG, MAP, HitRate, …) and Error (RMSE, MAE) metrics
- Fully integrated with WandB monitoring service
- Full use of transformers and datasets libraries
- Easy to use (via .yaml configuration or Python api)
- Fast (Intended to be used for consumer gpus)
- Fully modular and easily extensible!
Goal
Extending the framework with novel LLMs Models and Recommendation Approaches. The aim is to publish a co-authored conference or journal research paper.