search

layer6ai-labs / xpool

https://layer6ai-labs.github.io/xpool/
110 stars 9 forks source link

Exceptions happeans when training on MSVD dataset #14

Closed ChinaWangYu666 closed 1 year ago

ChinaWangYu666 commented 1 year ago

We run the code on the MSVD dataset, but the code is exited without any errors and hints. The author can give me some help. Thanks very much!

NoelVouitsis commented 1 year ago

At one point in the code does this happen? You are likely running out of memory. What is your batch size?

ChinaWangYu666 commented 1 year ago

At one point in the code does this happen? You are likely running out of memory. What is your batch size?

It happens during the testing phase. I find it is running out of memory

NoelVouitsis commented 1 year ago

MSVD evaluation requires a lot of memory because we perform evaulation for multiple captions per video at the same time. Inference happens on the CPU. I would suggest either using a machine with more CPU memory or experimenting on less memory intensive datasets like MSR-VTT or LSMDC. Thanks!

EmreOzkose commented 1 year ago

Hi @ChinaWangYu666 , I also encountered this problem during testing. You can increase swap memory.