Reproducible crash due to lack of memory after retraining

cramerlab / warp

GNU General Public License v3.0

54 stars 16 forks source link

Reproducible crash due to lack of memory after retraining #6

Closed bbarad closed 5 years ago

bbarad commented 5 years ago

After picking particles and retraining boxnet with them, the GPU memory does not appear to become available again and warp crashes when processing is started due to lack of memory. Quitting and reloading the software resolves this until retraining is done again.

Version is 1.0.7 beta 2.

dtegunov commented 5 years ago

Thanks for reporting! Google changed that part of TensorFlow in 1.10 and I didn't notice. Let's try beta 3: http://deployment.warpem.com/Warp_1.0.7beta3.msi

bbarad commented 5 years ago

This update resolves this issue.