Websight

Prepare for finetune

cd LLaMA-Factory
uv pip install -e .
cd ..
uv run -m model.prepare_data

Put the following in data/dataset_info.json

{
  "waveui_clicks": {
    "file_name": "waveui_train.jsonl",
    "eval_file_name": "waveui_val.jsonl",
    "template": "qwen2_vl",
    "formatting": "sharegpt",
    "columns": { "messages": "conversations", "images": null }
  }
}

Train with:

llamafactory-cli train train_waveui_lora.yaml