Required Hardware Configuration
Running WeClone (especially during the model fine-tuning phase) has high requirements for video memory. It is recommended to use a dedicated GPU device or a cloud GPU rental service. Running on integrated graphics or with only a CPU is not recommended.
The project uses the Qwen2.5-7B-Instruct model by default and employs the LoRA method for fine-tuning, which requires approximately 16GB of video memory.
Additionally, please reserve more than 20GB of hard disk space to store model files, intermediate results, and cached data.
The following table lists the estimated video memory requirements for different model sizes and fine-tuning methods. Multimodal models will occupy more video memory depending on the size and number of images:
| Fine-tuning Method | Precision (bits) | 7B Model | 14B Model | 30B Model | 70B Model | xB Model |
|---|---|---|---|---|---|---|
Full (bf16 / fp16) | 32 | 120GB | 240GB | 600GB | 1200GB | 18x GB |
Full (pure_bf16) | 16 | 60GB | 120GB | 300GB | 600GB | 8x GB |
| Freeze / LoRA / GaLore / APOLLO | 16 | 16GB | 32GB | 64GB | 160GB | 2x GB |
| QLoRA | 8 | 10GB | 20GB | 40GB | 80GB | x GB |
| QLoRA | 4 | 6GB | 12GB | 24GB | 48GB | x/2 GB |
| QLoRA | 2 | 4GB | 8GB | 16GB | 24GB | x/4 GB |
If you wish to enable the QLoRA fine-tuning method, please refer to the subsequent chapter "Enable QLoRA (Optional Configuration)" to learn how to switch the fine-tuning strategy.
WARNING
The weights fine-tuned using the quantized model cannot be merged back into the original model.
TIP
Video Memory ≥16GB: The default LoRA fine-tuning scheme is recommended. Video Memory <16GB: Consider switching to QLoRA or choosing a model with a smaller number of parameters.