search

TinyLLaVA / TinyLLaVA_Factory

A Framework of Small-scale Large Multimodal Models
https://arxiv.org/abs/2402.14289
Apache License 2.0
658 stars 68 forks source link

Loss #53

Open Yu-xm opened 6 months ago

Yu-xm commented 6 months ago

Can you share the loss curve during the training process?

Yu-xm commented 6 months ago

Including pretrain and finetune

cooleel commented 6 months ago

They shared the training loss curves for LLaVA-v1.5-13B at https://github.com/haotian-liu/LLaVA/issues/579#issuecomment-1763707042 Mine look the same.