HANKUG

i have been same problem in linux docker environment use cuda 11.6 4 gpu and i solve it mount host /dev/shm volume to my docker container this is my docker...

this is my sample code def make_bin(path,save_name,shift_num,image_size): even_list, odd_list, truths = even_odd_split(path,shift_num)#replace your own test dataset code new_dataset = [] for data in zip(even_list,odd_list): new_dataset.append(data[0]) new_dataset.append(data[1]) _bins = [] for...

i review insight face validation code they split test dataset even and odd then compare even image and odd image is same so you want to make your own test...

shift_num is rotation factor of even_list or odd_list sequence if you do not rotate even_list or odd_list then matching true ratio is too increased