I want to use local data to fine-tune the mode by using ChatGLM , like this , how to w

Yes, git clone https://github.com/Wa

how it works? about medqa-chatglm HOT 3 CLOSED

wangrongsheng commented on June 3, 2024

how it works?

from medqa-chatglm.

Comments (3)

WangRongsheng commented on June 3, 2024

Yes,

git clone https://github.com/WangRongsheng/MedQA-ChatGLM.git
Install packages: pip install -r requirements.txt
Prepare data, Please reference data format: data/self_cognition.json
Add your data path: data/dataset_info.json
Fine-tuning your model:

CUDA_VISIBLE_DEVICES=0 python MedQA-ChatGLM/finetune.py \
                              --do_train \
                              --dataset self_cognition \
                              --finetuning_type lora \
                              --output_dir ./fintuning-output-lora \
                              --per_device_train_batch_size 32 \
                              --gradient_accumulation_steps 256 \
                              --lr_scheduler_type cosine \
                              --logging_steps 500 \
                              --save_steps 1000 \
                              --learning_rate 5e-5 \
                              --num_train_epochs 10.0 \
                              --fp16

from medqa-chatglm.

earthxx commented on June 3, 2024

yeah，but there are still some issues and I am very interested in your achievements, Can you leave me a contact information? For example, WeChat or email, etc？

from medqa-chatglm.

WangRongsheng commented on June 3, 2024

yes, email: [email protected]

from medqa-chatglm.

Recommend Projects