Closed irene622 closed 1 year ago
Thank you for the PR! Please see my comments in the code.
Thanks, I change them! Please merge them. :)
@irene622 The black code formatter failed. Could you please fix it (black prover/evaluate.py
)? Thx
Oh, I done black prover/evaluate.py
.
Hi, I experiment your code, I think the some of code would be good if it is changed.
After evaluation is done, in my experiment, all theorem are discarded, so I meet
ZeroDivisionError
. Hence, by usingif
, adding the case of all theorem are discarded can avoid theZeroDivisionError
.When evaluating with gpus and cpus more than 2 respectively,
ray.init
takes the argumentnum_gpus
asnum_cpus
. It cause the error in case of num gpus and num cpus are not equal. I think it is typo, because it is easy to confuse num_cpus and num_gpus. So, addtorch.cuda.device_count()
to find num_gpus instead of getting num_gpus as arguments.Thank you for spending your time, I hope to merge my modification soon. :)