Using LLM-powered Neural Information Retrieval for Target Determination

To run indexing (8 GPUs): torchrun --standalone --nnodes=1 --nproc-per-node=8 indexer.py --experiment-name codellama-try3

To run retrieval (1 GPU): torchrun --standalone --nnodes=1 --nproc-per-node=1 retriever.py --experiment-name codellama-try3 --pr-parse-format GITDIFF