search

Coobiw / MPP-LLaVA

Personal Project: MPP-Qwen14B & MPP-Qwen-Next(Multimodal Pipeline Parallel based on Qwen-LM). Support [video/image/multi-image] {sft/conversations}. Don't let the poverty limit your imagination! Train your own 8B/14B LLaVA-training-like MLLM on RTX3090/4090 24GB.
368 stars 20 forks source link

请问如何减少训练时长 #14

Closed xiyuanhao closed 4 months ago

xiyuanhao commented 6 months ago

预训练loss 到1.4左右，step到900多 就基本不动了，然后还一直训练？是我的显卡问题么？

Coobiw commented 6 months ago

是在哪个setting跑的呢，然后loss趋于平缓也很正常，并不是说loss基本不动就会停止啊

Coobiw commented 4 months ago

I've pushed the loss curves in README. I'll close this issue.