Cuda error during training step.
AndresFRJ98 opened this issue · comments
AndresFRJ98 commented
Hello I was running the training step as mentioned python main.py --mode train --para_limit 2250 --batch_size 24 --init_lr 0.1 --keep_prob 1.0 --sp_lambda 1.0
I encountered this error. I am running this on GPU , is it possible that this is a memory allocation issue? Any idea how much memory is required to perform the training step?
Any pointers towards resolving this would be much appreciated.
AndresFRJ98 commented
Updated my cuda version and it worked out :)
njucsxh commented
how to do it? my environment is
torch 0.3
cuda 9.0
while my cuda version in "nvidia-smi" is 10.2,cudatookit is 10.1
i find that there is no higher cuda version for torch 0.3 while readme said they use torch=0.3
thank!