[英]Why is looping through pytorch tensors so slow (compared to Numpy)?
I've been working with image transformations recently and came to a situation where I have a large array (shape of 100,000 x 3) where each row represents a point in 3D space like:我最近一直在处理图像转换,遇到了一个大数组(形状为 100,000 x 3)的情况,其中每一行代表 3D 空间中的一个点,例如:
pnt = [x y z]
All I'm trying to do is iterating through each point and matrix multiplying each point with a matrix called T (shape = 3 X 3).我想要做的就是迭代每个点,并将每个点与一个名为 T 的矩阵相乘(形状 = 3 X 3)。
def transform(pnt_cloud, T):
i = 0
for pnt in pnt_cloud:
xyz_pnt = np.dot(T, pnt)
if xyz_pnt[0] > 0:
arr[i] = xyz_pnt[0]
i += 1
return arr
Calling the following code and calculating runtime (using %time) gives the output:调用以下代码并计算运行时间(使用 %time)给出输出:
Out[190]: CPU times: user 670 ms, sys: 7.91 ms, total: 678 ms
Wall time: 674 ms
import torch
tensor_cld = torch.tensor(pnt_cloud)
tensor_T = torch.tensor(T)
def transform(pnt_cloud, T):
depth_array = torch.tensor(np.zeros(pnt_cloud.shape[0]))
i = 0
for pnt in pnt_cloud:
xyz_pnt = torch.matmul(T, pnt)
if xyz_pnt[0] > 0:
depth_array[i] = xyz_pnt[0]
i += 1
return depth_array
Calling the following code and calculating runtime (using %time) gives the output:调用以下代码并计算运行时间(使用 %time)给出输出:
Out[199]: CPU times: user 6.15 s, sys: 28.1 ms, total: 6.18 s
Wall time: 6.09 s
NOTE: Doing the same with torch.jit only reduces 2s注意:对 torch.jit 做同样的事情只会减少 2s
I would have thought that PyTorch tensor computations would be much faster due to the way PyTorch breaks its code down in the compiling stage.由于 PyTorch 在编译阶段分解其代码的方式,我原以为 PyTorch 张量计算会快得多。 What am I missing here?
我在这里缺少什么?
Would there be any faster way to do this other than using Numba?除了使用 Numba 之外,还有什么更快的方法可以做到这一点吗?
Why are you using a for loop??为什么要使用for循环??
Why do you compute a 3x3 dot product and only uses the first element of the result??为什么要计算 3x3 点积并且只使用结果的第一个元素?
You can do all the math in a single matmul
:您可以在单个
matmul
完成所有数学运算:
with torch.no_grad():
depth_array = torch.matmul(pnt_cloud, T[:1, :].T) # nx3 dot 3x1 -> nx1
# since you only want non negative results
depth_array = torch.maximum(depth_array, 0)
Since you want to compare runtime to numpy, you should disable gradient accumulation .由于您想将运行时与 numpy 进行比较,您应该禁用梯度累积。
For the speed, I got this reply from the PyTorch forums:对于速度,我从 PyTorch 论坛得到了这个回复:
operations of 1-3 elements are generally rather expensive in PyTorch as the overhead of Tensor creation becomes significant (this includes setting single elements), I think this is the main thing here.在 PyTorch 中,1-3 个元素的操作通常相当昂贵,因为 Tensor 创建的开销变得很大(这包括设置单个元素),我认为这是这里的主要内容。 This is also the reason why the JIT doesn't help a whole lot (it only takes away the Python overhead) and Numby shines (where eg the assignment to depth_array[i] is just memory write).
这也是为什么 JIT 没有多大帮助(它只带走 Python 开销)和 Numby 闪耀的原因(例如,对 depth_array[i] 的分配只是内存写入)。
the matmul itself might differ in speed if you have different BLAS backends for it in PyTorch vs. NumPy.如果在 PyTorch 和 NumPy 中使用不同的 BLAS 后端,matmul 本身的速度可能会有所不同。
声明:本站的技术帖子网页,遵循CC BY-SA 4.0协议,如果您需要转载,请注明本站网址或者原文地址。任何问题请咨询:yoyou2525@163.com.