masterthesis-playground/raft

Retrieval-Augmented Finetuning (RAFT)

Ablauf:

Vorbereiten des Retrieval-Corpus

python prepare_corpus.py --input_tab ../data/intermediate/selected_topics_documents.csv --out_dir out

Erstellen des RAFT-Datensatzes

python make_raft_data.py --out_dir out --n_examples 100

Training der QLoRA-Adapter

python train_mistral_raft.py --train_jsonl out/raft_train.jsonl --out_dir out/mistral_balitwin_lora

Inferenz

Per Baseline Mistral 7B + PEFT-Adapter

python rag_chat.py --lora_dir out/mistral_balitwin_lora

Pre-Merged Modell + Adapter

python rag_chat_merged.py --model_dir /path/to/model_folder --out_dir out