in the origin server
first install SimpleTornadoServer
pip install SimpleTornadoServer
then generate the manifest of the dir to be download
#In process 1 | |
import torch | |
shape = (100000,) | |
a=torch.rand(shape, device="cuda:0").share_memory_() | |
deviceId, handle_ptr, size, offset,*_ =a.storage()._share_cuda_() | |
# in process 2 | |
import numpy as np | |
import numba.cuda |
in the origin server
first install SimpleTornadoServer
pip install SimpleTornadoServer
then generate the manifest of the dir to be download