search

InternLM / InternLM-XComposer

InternLM-XComposer2 is a groundbreaking vision-language large model (VLLM) excelling in free-form text-image composition and comprehension.
1.91k stars 120 forks source link

Thanks for your feedback! Can u retry the latest version? #291

Open 2132660698 opened 2 months ago

2132660698 commented 2 months ago 
          Thanks for your feedback! Can u retry the latest version?

Originally posted by @yuhangzang in https://github.com/InternLM/InternLM-XComposer/issues/288#issuecomment-2071359528 你好，4.23日13点更新了代码，微调internlm-xcomposer2-4khd-7b报下面的错误，麻烦再看一下： image

2132660698 commented 2 months ago

image

yuhangzang commented 2 months ago

Hi, can u post the parameters of your finetune_lora.sh file?

2132660698 commented 2 months ago

image

2132660698 commented 2 months ago

这是我的微调数据，这个数据微调internlm-xcomposer2-vl-7b没有问题。 image

yuhangzang commented 2 months ago

This issue is due to the current 4khd fine-tuning code does not support multiple images in one query. Please stay tuned.

2132660698 commented 2 months ago

This issue is due to the current 4khd fine-tuning code does not support multiple images in one query. Please stay tuned.

应该是你说的这个情况，将训练数据中的多图改成一张图后，可以Lora训练完成。

2132660698 commented 2 months ago

请问Lora训练后，可以合并吗？