| pip install "transformers<4.52" | |
| CUDA_VISIBLE_DEVICES=0 \ | |
| swift export \ | |
| --model Qwen/Qwen2.5-72B-Instruct \ | |
| --dataset 'AI-ModelScope/alpaca-gpt4-data-zh#500' \ | |
| 'AI-ModelScope/alpaca-gpt4-data-en#500' \ | |
| --device_map cpu \ | |
| --quant_n_samples 256 \ | |
| --quant_batch_size 1 \ | |
| --max_length 2048 \ | |
| --quant_method awq \ | |
| --quant_bits 4 \ | |
| --output_dir Qwen2.5-72B-Instruct-AWQ | |