Tried to allocate 20.00 MiB (GPU 0; 5.93 GiB total capacity; 4.63 GiB already allocated; 23.19 MiB free; 4.85 GiB reserved in total by PyTorch)

Asked Jul 09 '23 at 03:04

Active Jul 09 '23 at 03:14

Viewed 43 times

I am trying to run train pytorch model

this is my config file batch size, I have tried reducing batch_size to 1 but the same error

local batch_size = 3,
local num_batch_accumulated = 4,

This is the output of nvidia-smi

As we can see only 451 mib is allocated out of 6144 Mib

I have tried following different solutions mentioned in stackoverflow post here but not able to solve this.

edited Jul 09 '23 at 03:14

talonmies

asked Jul 09 '23 at 03:04

Sarde

Please, provide some additional information. What kind of model are you trying to train? Are you running training in Jupyter notebook or as a python script? What output does torch.cuda.is_available() give you? – Sherstnyov Jul 09 '23 at 11:00
I am trying to run following git project https://github.com/dmirlab-group/sadga error occurs on https://github.com/dmirlab-group/sadga#step-2-training – Sarde Jul 10 '23 at 11:39

0 Answers0