# Retrieval-Augmented Finetuning (RAFT)

**Ablauf**:

## Vorbereiten des Retrieval-Corpus

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

## Erstellen des RAFT-Datensatzes

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

## Training der QLoRA-Adapter

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

## Inferenz

### Per Baseline Mistral 7B + PEFT-Adapter

```bash
python rag_chat.py --lora_dir out/mistral_balitwin_lora
```

### Pre-Merged Modell + Adapter

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