Views
No views yet
pip install -U 'xtuner[deepspeed]'1xtuner chat lmsys/vicuna-13b-v1.5 \
2 --visual-encoder openai/clip-vit-large-patch14-336 \
3 --llava xtuner/llava-v1.5-13b-xtuner \
4 --prompt-template vicuna \
5 --image $IMAGE_PATH./work_dirs/)NPROC_PER_NODE=8 xtuner train llava_vicuna_13b_v15_clip_vit_large_p14_336_e1_gpu8_pretrain --deepspeed deepspeed_zero2./work_dirs/)NPROC_PER_NODE=8 xtuner train llava_vicuna_13b_v15_qlora_clip_vit_large_p14_336_lora_e1_gpu8_finetune --deepspeed deepspeed_zero21xtuner mmbench lmsys/vicuna-13b-v1.5 \
2 --visual-encoder openai/clip-vit-large-patch14-336 \
3 --llava xtuner/llava-v1.5-13b-xtuner \
4 --prompt-template vicuna \
5 --data-path $MMBENCH_DATA_PATH \
6 --work-dir $RESULT_PATHmmbench_result.xlsx to the official MMBench for final evaluation to obtain precision results!1@misc{2023xtuner,
2 title={XTuner: A Toolkit for Efficiently Fine-tuning LLM},
3 author={XTuner Contributors},
4 howpublished = {\url{https://github.com/InternLM/xtuner}},
5 year={2023}
6}