cupy 加速

这个本质是cuda计算，训练的时候gpu是紧缺资源，这个也需要gpu，所以gpu冲突了。

 

import cupy as cp

def to_cupy(tensor):
    """Convert PyTorch tensor to CuPy array.
    """
    return cp.fromDlpack(to_dlpack(tensor))

def to_tensor(cp_array):
    """Convert CuPy array to PyTorch tensor.
    """
    return from_dlpack(cp_array.toDlpack())

合并拼接：

cp_data=cp.hstack((cp_data, cp_data))

numpy 与cupy互转:

import cupy as cp
import numpy as np
x_cpu = np.random.randn(100, 100).astype(np.float32)
x_gpu = cp.asarray(x_cpu)


n_data = cp.asnumpy(x_gpu)
print(n_data)


x_cpu*x_cpu
2.41 µs ± 19.7 ns per loop (mean ± std. dev. of 7 runs, 100000 loops each)
%%timeit
x_gpu*x_gpu
14.3 µs ± 53.4 ns per loop (mean ± std. dev. of 7 runs, 100000 loops each)
import torch
x_tensor = torch.from_numpy(x