2024-04-22 简化    2023-10-18 微调推理测试初步完成    2023-10-17 initial visualglm_finetuning

install

weight

data sample

p prefix  optionalq question optionala answer   must

{"id":1,"paragraph": [{"q":"<img>../assets/demo.jpeg</img>\n图中的狗是什么品种？","a":"图中是一只拉布拉多犬。"}]}

{"id":0,"conversations": [      {"from":"user","value":"<img>../assets/demo.jpeg</img>\n图中的狗是什么品种？"      },      {"from":"assistant","value":"图中是一只拉布拉多犬。"      }    ]}

infer

# infer.py 推理预训练模型# infer_finetuning.py 推理微调模型# infer_lora_finetuning.py 推理lora微调模型 python infer.py

training

    # 制作数据    cd scripts    bash train_full.sh -m dataset     or    bash train_lora.sh -m dataset     or    bash train_ptv2.sh -m dataset         注: num_process_worker 为多进程制作数据 ， 如果数据量较大 ， 适当调大至cpu数量    dataHelper.make_dataset_with_args(data_args.train_file,mixed_data=False, shuffle=True,mode='train',num_process_worker=0)        # 全参数训练         bash train_full.sh -m train            # lora adalora ia3         bash train_lora.sh -m train            # ptv2        bash train_ptv2.sh -m train

训练参数

友情链接

纯粹而干净的代码

Reference

https://github.com/THUDM/VisualGLM-6B

Star History