model_path = '/root/ld/ld_model_pretrained/minicpm3' # model_path or model_id
quant_path = '/root/ld/ld_model_pretrained/minicpm3_awq' # quant_save_path
quant_data_path='/root/ld/ld_project/pull_request/MiniCPM/quantize/quantize_data/alpaca'# 写入自带wikitext或者alpaca地址
quant_config = { "zero_point": True, "q_group_size": 128, "w_bit": 4, "version": "GEMM" } # "w_bit":4 or 8
quant_samples=512 # how many samples to use for calibration
custom_data=[ # first custom data 如果要使用自定义数据集,根据以下格式修改
{"role": "system", "content": "You are a helpful assistant."},
{"role": "assistant", "content": "你好,有什么我可以帮助你的吗?"},
{"role": "user", "content": "我想了解如何编写Python代码。"},
{"role": "system", "content": "You are a helpful assistant."},
{"role": "assistant", "content": "你好,有什么我可以帮助你的吗?"},
{"role": "user", "content": "我想了解如何编写Python代码。"},