Giters

NVlabs / FB-BEV

Official PyTorch implementation of FB-BEV & FB-OCC - Forward-backward view transformation for vision-centric autonomous driving perception

Geek Repo:Geek Repo

Github PK Tool:Github PK Tool

memory overflow of evaluation

rkdckddnjs9 opened this issue · comments

commented

Thanks for your excellent works!

When I test with dist_test.sh, the memory continues to increase, causing the evaluation to stop.
The same thing happens when fbocc.py's pred_occcupancy_category = pred_occcupancy_category.cpu().numpy() is performed without changing to cpu.
I wonder if there were any problems like this, do you know how to solve them?

commented

Even if the evaluation was conducted on another server with 128G of RAM, it was confirmed that the memory was exceeded.

commented

Even if the evaluation was conducted on another server with 128G of RAM, it was confirmed that the memory was exceeded.

Have you solved this? The memory overflow on the 160g RAM too.

commented

Even if the evaluation was conducted on another server with 128G of RAM, it was confirmed that the memory was exceeded.

Have you solved this? The memory overflow on the 160g RAM too.

I tried several methods, but couldn't solve it.

commented

This work preserves the history information in the RAM, so more memory is needed.

commented

This work preserves the history information in the RAM, so more memory is needed.

Could you give me your computer specifications and environmental information?

commented

This work preserves the history information in the RAM, so more memory is needed.

Could you give me your computer specifications and environmental information?

I run the code on the cluster, similar errors occur when I specify the default quantity memory, but this problem is solved when I allocate 500g memory.

commented

This work preserves the history information in the RAM, so more memory is needed.

Could you give me your computer specifications and environmental information?

I run the code on the cluster, similar errors occur when I specify the default quantity memory, but this problem is solved when I allocate 500g memory.

Thank you for your reply :)