reconstructed data sequence
Closed this issue · 1 comments
tonyyunyang commented
Hi,
I am now using my own data and your architecture for the pretrain process.
Have you every visualized the reconstructed data?
You could find the reconstructed data of mine.
It seems that the reconstructed data is like a set of "identical" values that minimizes the difference among all masked values.
Is this what you have seen as well when applying on the IMU data?
tonyyunyang commented
No problem! Figured that this is due to not setting enough epoch.