Meta Llama 2 and Llama 3 fine-tuned using QLoRA on a subset of the Amazon Reviews '23 dataset to generate titles using Amazon review data.
Models available on Hugging Face: Llama 2 and Llama 3. Note: model card currently unfinished.
This repo includes the main experimental notebook and utils.py helper code. It also contains the final processed data subsets used for training and evaluation.