| Step | Action |
|---|---|
| 1 | Ensure GPU (CUDA) runtime |
| 2 | Install dependencies:pip install imagebind llama-index vllm |
| 3 | Download LLaMA weight files (from HF or Meta office) |
| 4 | Download ImageBind “huge” model checkpoint |
| 5 | Point your script’s paths to both checkpoints |
| 6 | Run again in GPU environment |