I am using a NVidia GeForce 830M graphics card of 2GB. Therefore I encountered the out of memory error as I ran learn.fit(lr, 3, cycle_len=1, cycle_mult=2)
In a different thread I found a solution for minimising the batch size in the function
data = ImageClassifierData.from_paths(PATH,bs=32,tfms=tfms_from_model(arch, sz)) where such problems arrised.
But for the function learn.fit() , can I perform some other operation that lets me use less memory for it or will I have to use that much memory?
Hi,
Try to play with ‘bs’ and ‘sz’ I had this problem on 12GB after I made it working on small bs =4
I was able to gradually increase it and problem disapee.
to check free memory run nvidia-smi
Cheers
M