Haokun Lin

icon indicating a website link https://felixmessi.github.io

icon company Institute of Automation, CAS & CityU icon location Hong Kong SAR icon location Joint PhD student at CASIA & CityU.

Results 1 issues of Haokun Lin

Training loss NaN

1 comment

Hi, thanks for the amazing work! I've encountered an issue with NaN losses when using AMP in MambaVision, particularly when I reduce the training epochs to 30. The problem seems...