You signed in with another tab or window. Reload to refresh your session.You signed out in another tab or window. Reload to refresh your session.You switched accounts on another tab or window. Reload to refresh your session.Dismiss alert
watch_torch_model() does not work entirely. For all images, I somehow only see the very first training instance the model encountered, though the distributions and also some stuff in the weights still seem to get updated.
Setup:
Windows 10
Deepkit Release 2020.1.5
Deepkit SDK 1.0.1
Running experiment on anything, from local script run, to local UI start to deployed docker run, eg. on pytorch/pytorch:1.4-cuda10.1-cudnn7-runtime