Hi,
I've added a few lines to allow the user to select GPU IDs in train.py and predict.py. The user specifies GPU IDs as a list or an integer in the json config files for those scripts. Importantly, I've also added a line to set_memory_growth to true for GPUs as without it I was consistently getting OOM errors if you think this is a sensible way to deal with that? The structure of this PR could probably do with some work too: I wasn't sure exactly where to place the function and how to call it within the current cryoCARE structure so I simply added a function in predict.py and also call it in train.py
Thanks,
Euan
Hi, I've added a few lines to allow the user to select GPU IDs in train.py and predict.py. The user specifies GPU IDs as a list or an integer in the json config files for those scripts. Importantly, I've also added a line to set_memory_growth to true for GPUs as without it I was consistently getting OOM errors if you think this is a sensible way to deal with that? The structure of this PR could probably do with some work too: I wasn't sure exactly where to place the function and how to call it within the current cryoCARE structure so I simply added a function in predict.py and also call it in train.py Thanks, Euan