RuntimeError: CUDNN_STATUS_EXECUTION_FAILED

rowanz / neural-motifs

Code for Neural Motifs: Scene Graph Parsing with Global Context (CVPR 2018)

https://rowanzellers.com/neuralmotifs

MIT License

525 stars 115 forks source link

RuntimeError: CUDNN_STATUS_EXECUTION_FAILED #95

Open progvasu opened 4 years ago

progvasu commented 4 years ago

Hi,

I am running into the below error while running the evaluation script "eval_models_sgdet.sh"

I am using torch 0.3 with cuda 9.0 (ran the same conda cmd you provided in readme)

Any idea why it might be happening?

Thanks and Regards, Vasu

nullkatar commented 4 years ago

@progvasu Hey there, did you solve this issue? I have the same thing on my desktop. I read around it in the internet, probably it could be caused by RTX graphics card...

wtt0213 commented 3 years ago

maybe the gpu runs out of the memery, try to set nworker to 0, or set batch size smaller

congvmit commented 3 years ago

Hello @progvasu, Have you solve this problem? I am stuck at this step for a day. Could you please help me?

progvasu commented 3 years ago

Hey @congvm-it, no I didn't solve it. I had to switch to a different machine and didn't get any error there.