NumPy、Torch和Tensorflow 代码对比

2023-05-16

深度学习

在深入学习的基本单位上实施初级到高级操作。

> Excerpts

我习惯于为不同的问题创建新的深度学习架构，但选择哪个框架（Keras、Pytorch、TensorFlow）通常比较困难。

由于其中存在不确定性，因此了解这些框架的基本单元（NumPy、Torch、Tensor）的基本操作是件好事。

在这篇文章中，我跨 3 个框架执行了几个相同的操作，也尝试了大多数框架的可视化操作。

这是一个初学者友好的帖子，所以让我们开始。

1. 安装

pip install numpy
pip install tensorflow
pip install torch

2. 版本检查

import numpy as np
import tensorflow as tf
import torch

print(np.__version__)
print(tf.__version__)
print(torch.__version__)

### OUTPUT ###

2.3.0 
1.18.5 
1.6.0+cu101
view raw

3. 阵列初始化 = 1-D、2-D、3-D

标量和一维阵列

> Scalar, 1-D, 2-D arrays

Numpy

# Numpy

a = np.array(10)
print(a)
print(a.shape, a.dtype) #shape of the array and type of the elements

a = np.array([10])
print(a)
print(a.shape, a.dtype) #shape of the array and type of the elements

a = np.array([10], dtype=np.float32)
print(a)
print(a.shape, a.dtype) #shape of the array and type of the elements

############################# OUTPUT ##############################

10 () int64
[10] (1,) int64
[10.] (1,) float32

TensorFlow

# TensorFlow

b = tf.constant(10) # As Scalar
print(b)
b = tf.constant(10, shape=(1,1)) # As 1-D Vector
print(b)
b = tf.constant(10, shape=(1,1), dtype=tf.float32) # with Data-type
print(b)

############################# OUTPUT ##############################

tf.Tensor(10, shape=(), dtype=int32)
tf.Tensor([[10]], shape=(1, 1), dtype=int32)
tf.Tensor([[10.]], shape=(1, 1), dtype=float32)

Torch

# Torch

c = torch.tensor(10, ) # As Scalar
print(c)
c = torch.tensor([10]) # As 1-D Vector
print(c, c.shape, c.dtype)
c = torch.tensor([10], dtype=torch.float32) # With Data-type
print(c)

############################# OUTPUT ##############################

tensor(10)  
tensor([10])torch.Size([1]) torch.int64
tensor([10.])

二维矢量阵列

> 2-D Array

Numpy

# Numpy

a = np.array([[1,2,3], [4,5,6]])
print(a)
print(a.shape, a.dtype)

############################# OUTPUT ##############################

[[1 2 3]  
 [4 5 6]] 
(2, 3) int64

TensorFlow

# TensorFlow

b = tf.constant([[1,2,3], [4,5,6]])
print(b)
print(b.shape)

############################# OUTPUT ##############################

tf.Tensor([[1 2 3]
           [4 5 6]], shape=(2, 3), dtype=int32)
(2, 3)

Torch

# Torch

c = torch.tensor([[1,2,3], [4,5,6]])
print(c)
print(c.shape)

############################# OUTPUT ##############################

tensor([[1, 2, 3],
        [4, 5, 6]])
torch.Size([2, 3])

4. 生成数据

> Zeros and Ones

> Diagonal & Same element fill

Numpy

# Numpy

a = np.zeros((3,3))
print(a, a.shape, a.dtype)

a = np.ones((3,3))
print(a, a.shape, a.dtype)

a = np.eye(3)
print(a, a.shape, a.dtype)

a = np.full((3,3),10.0)
print(a, a.shape, a.dtype)

############################# OUTPUT ##############################

[[0. 0. 0.]
 [0. 0. 0.]
 [0. 0. 0.]] (3, 3) float64

[[1. 1. 1.] 
 [1. 1. 1.] 
 [1. 1. 1.]] (3, 3) float64

[[1. 0. 0.] 
 [0. 1. 0.] 
 [0. 0. 1.]] (3, 3) float64

[[10. 10. 10.] 
 [10. 10. 10.] 
 [10. 10. 10.]] (3, 3) float64

Tensorflow

# TensorFlow

b = tf.zeros((3,3))
print(b)

b = tf.ones((3,3))
print(b)

b = tf.eye(3)
print(b)

b = tf.fill([3,3], 10)
print(b)

############################# OUTPUT ##############################

tf.Tensor( [[0. 0. 0.]
            [0. 0. 0.]
            [0. 0. 0.]], shape=(3, 3), dtype=float32)

tf.Tensor( [[1. 1. 1.]
            [1. 1. 1.]
            [1. 1. 1.]], shape=(3, 3), dtype=float32)

tf.Tensor( [[1. 0. 0.]
            [0. 1. 0.]
            [0. 0. 1.]], shape=(3, 3), dtype=float32)

tf.Tensor( [[10 10 10]
            [10 10 10]
            [10 10 10]], shape=(3, 3), dtype=int32)

Torch

# Torch

c = torch.zeros((3,3))
print(c)

c = torch.ones((3,3))
print(c)

c = torch.eye(3)
print(c)

c = c.new_full([3,3], 10)
print(c)

############################# OUTPUT ##############################

tensor([[0., 0., 0.],
        [0., 0., 0.],
        [0., 0., 0.]])

tensor([[1., 1., 1.],
        [1., 1., 1.],
        [1., 1., 1.]])

tensor([[1., 0., 0.],
        [0., 1., 0.],
        [0., 0., 1.]])

tensor([[10., 10., 10.],
        [10., 10., 10.],
        [10., 10., 10.]])

从正态分布绘制随机样本

> Normal Dist'n Bell Curve

> Samples were drawn from Normal Dist'n

Numpy

# Numpy

a = np.random.randn(3,3)
print(a, a.shape, a.dtype)
print(a.mean(), a.std())

############################# OUTPUT ##############################

[[ 0.41406362 -1.51382214  0.55400531]
 [-0.95226975 -0.50038461  1.29014057]
 [ 0.90320426 -1.65923581 -1.03100388]] (3, 3) float64

-0.27725582657643905 1.0291132313341855

Tensorflow

# TensorFlow

b = tf.random.normal((3,3),mean=0, stddev=1)
print(b)
print(tf.reduce_mean(b), tf.math.reduce_std(b))

############################# OUTPUT ##############################

tf.Tensor([[0.04017716 -2.2840774 -0.3615016 ]
           [-1.9259684  1.2054121 0.02211744]
           [0.96204025 0.07906733 -2.2352242 ]], shape=(3, 3), dtype=float32)

tf.Tensor(-0.49977303, shape=(), dtype=float32) 
tf.Tensor(1.2557517, shape=(), dtype=float32)

Torch

# Torch

c = torch.normal(mean=0, std=1, size=(3, 3))
print(c)
print(torch.mean(c), torch.std(c))

############################# OUTPUT ##############################

tensor([[-0.1682,  0.9610,  1.1005],
        [-1.0462,  0.4431,  0.6005], 
        [-1.2714, -1.1894,  0.7221]])
tensor(0.0169) tensor(0.9595)

从均匀分布中抽取样本

> Uniform Dist'n Curve

> Samples were drawn from Uniform Dist'n

Numpy

# Numpy

a = np.random.uniform(low=0, high=1, size=(3,3))
print(a, a.shape, a.dtype)
print(a.mean(), a.std())

############################# OUTPUT ##############################

[[0.00738313 0.56734768 0.92893306]
 [0.36580515 0.91986967 0.70643149]
 [0.72854485 0.73961837 0.88049091]] (3, 3) float64

0.6493804787345917 0.28317558657389263

Tensorflow

# TensorFlow

b = tf.random.uniform((3,3),minval=0,maxval=1)#Values are always > 1
print(b)
print(tf.reduce_mean(b), tf.math.reduce_std(b))

############################# OUTPUT ##############################

tf.Tensor( [[0.11186028 0.04624796 0.59104955]
           [0.5344571  0.1144793  0.8468257 ]
           [0.5247066  0.61488223 0.7592212 ]], shape=(3, 3), dtype=float32)

tf.Tensor(0.4604144, shape=(), dtype=float32)
tf.Tensor(0.2792521, shape=(), dtype=float32)

Torch

# Torch

num_samples = 3
Dim = 3
c = torch.distributions.Uniform(0, +1).sample((num_samples, Dim))
print(c)
print(torch.mean(c), torch.std(c))

############################# OUTPUT ##############################

tensor([[0.5842, 0.5787, 0.3526],
        [0.2647, 0.6233, 0.4482],
        [0.3495, 0.0562, 0.0495]])

tensor(0.3674) tensor(0.2158)

向量排列

Numpy

# Numpy

a = np.arange(0,9)
print(a)
a = np.arange(start=1, stop=20, step=2, dtype=np.float32)
print(a, a.dtype)

############################# OUTPUT ##############################

[0 1 2 3 4 5 6 7 8]
[ 1.  3.  5.  7.  9. 11. 13. 15. 17. 19.] float32

Tensorflow

# TensorFlow

b = tf.range(9)
print(b)
b = tf.range(start=1, limit=20, delta=2, dtype=tf.float64)
print(b)

############################# OUTPUT ##############################

tf.Tensor([0 1 2 3 4 5 6 7 8], shape=(9,), dtype=int32)
tf.Tensor([ 1.  3.  5.  7.  9. 11. 13. 15. 17. 19.], shape=(10,), dtype=float64)

Torch


# Torch

c = torch.arange(start=0, end=9)
print(c)
c = torch.arange(start=1, end=20, step=2, dtype=torch.float64)
print(c)

############################# OUTPUT ##############################

tensor([0, 1, 2, 3, 4, 5, 6, 7, 8])
tensor([ 1.,  3.,  5.,  7.,  9., 11., 13., 15., 17., 19.], dtype=torch.float64)

7. 数据类型 + 转换

uint8/16/32/64 ← →浮子8/16/32/64

Numpy

# Numpy

a = a.astype(np.uint8)
print(a, a.dtype)

############################# OUTPUT ##############################

[ 1  3  5  7  9 11 13 15 17 19] uint8

Tensorflow

# TensorFlow

b = tf.cast(b, dtype=tf.uint8)
print(b)

############################# OUTPUT ##############################

tf.Tensor([ 1  3  5  7  9 11 13 15 17 19], shape=(10,), dtype=uint8)

Torch

# Torch

c = torch.tensor(c)
c = c.type(torch.int64)
print(c)

############################# OUTPUT ##############################

tensor([ 1,  3,  5,  7,  9, 11, 13, 15, 17, 19])

8. 数学运算

> Sum and Subtract operations

> multiply and divide operations

Numpy

# Numpy

a = np.array([1,2,3,4,5])
b = np.array([6,7,8,9,10])

c = np.add(a, b)
print(c, c.dtype)

c = np.subtract(b,a)
print(c, c.dtype)

c = np.divide(b,a)
print(c, c.dtype)

c = np.multiply(b,a)
print(c, c.dtype)

c = (a**2)
print(c)

############################# OUTPUT ##############################

[ 7  9 11 13 15] int64  [5 5 5 5 5] int64
[6.         3.5        2.66666667 2.25       2.        ] float64
[ 6 14 24 36 50] int64
[ 1  4  9 16 25]

Tensorflow

# TensorFlow

x = tf.constant([1,2,3,4,5])
y = tf.constant([6,7,8,9,10])

z = tf.add(x,y)
print(z)

z = tf.subtract(y,x)
print(z)

z = tf.divide(y,x)
print(z)

z = tf.multiply(y,x)
print(z)

z = (x **2)
print(z)
############################# OUTPUT ##############################

tf.Tensor([ 7  9 11 13 15], shape=(5,), dtype=int32)
tf.Tensor([5 5 5 5 5], shape=(5,), dtype=int32)
tf.Tensor([6.         3.5        2.66666667 2.25       2.        ], shape=(5,), dtype=float64)
tf.Tensor([ 6 14 24 36 50], shape=(5,), dtype=int32)
tf.Tensor([ 1  4  9 16 25], shape=(5,), dtype=int32)

Torch

# Torch

t = torch.tensor([1,2,3,4,5])
u = torch.tensor([6,7,8,9,10])

v = torch.add(t, u) 
print(v)

v = torch.sub(u,t)
print(v)

v = torch.true_divide (u, t)
print(v)

v = torch.mul(u,t)
print(v)

v = (t **2)
print(v)

############################# OUTPUT ##############################

tensor([ 7,  9, 11, 13, 15])
tensor([5, 5, 5, 5, 5])
tensor([6.0000, 3.5000, 2.6667, 2.2500, 2.0000])
tensor([ 6, 14, 24, 36, 50])
tensor([ 1,  4,  9, 16, 25])

9. 点积

> Dot Product

Numpy

# Numpy

a = np.array([1,2,3,4,5])
b = np.array([6,7,8,9,10])
c = np.dot(a, b)
print(c, c.dtype)

############################# OUTPUT ##############################

130 int64

Tensorflow

# TensorFlow

x = tf.constant([1,2,3,4,5])
y = tf.constant([6,7,8,9,10])
z = tf.tensordot(x,y, axes=1)
print(z)

############################# OUTPUT ##############################

tf.Tensor(130, shape=(), dtype=int32)

Torch

# Torch

t = torch.tensor([1,2,3,4,5])
u = torch.tensor([6,7,8,9,10])
v = torch.dot(t,u)
print(v)

############################# OUTPUT ##############################

tensor(130)

10. 矩阵乘法

> Matrix Multiplication

Numpy

# Numpy

a = np.array([[1,2,3], [4,5,6]])
b = np.array([[1,2,3], [4,5,6], [7,8,9]])
c = np.matmul(a,b) # (2,3) @ (3,3) --> (2,3) output shape
print(c)

############################# OUTPUT ##############################

[[30 36 42]
 [66 81 96]]

Tensorflow

# TensorFlow

x = tf.constant([[1,2,3], [4,5,6]])
y = tf.constant([[1,2,3], [4,5,6], [7,8,9]])
z = tf.matmul(x,y) # (2,3) @ (3,3) --> (2,3) output shape
print(z)

############################# OUTPUT ##############################

tf.Tensor([[30 36 42]
            [66 81 96]], shape=(2, 3), dtype=int32)

Torch

# Torch

t = torch.tensor([[1,2,3], [4,5,6]])
u = torch.tensor([[1,2,3], [4,5,6], [7,8,9]])
v = torch.matmul(t,u) # (2,3) @ (3,3) --> (2,3) output shape
print(v)

############################# OUTPUT ##############################

tensor([[30, 36, 42],
        [66, 81, 96]])

11. 索引和切片（2D）

> Indexing and Slicing

Numpy

# Numpy

a = np.array([1,2,3,4,5,6,7,8])
print(a[:])
print(a[2:-3])
print(a[3:-1])
print(a[::2])

indices = np.array([0,3,5])
x_indices = a[indices]
print(x_indices)
############################# OUTPUT ##############################

[1 2 3 4 5 6 7 8]
[3 4 5]
[4 5 6 7]
[1 3 5 7]
[1 4 6]

Tensorflow

# TensorFlow

b = tf.constant([1,2,3,4,5,6,7,8])
print(b[:])
print(b[2:-3])
print(b[3:-1])
print(b[::2])

indices = tf.constant([0,3,5])
x_indices = tf.gather(b, indices)
print(x_indices)

############################# OUTPUT ##############################

tf.Tensor([1 2 3 4 5 6 7 8], shape=(8,), dtype=int32)
tf.Tensor([3 4 5], shape=(3,), dtype=int32) 
tf.Tensor([4 5 6 7], shape=(4,), dtype=int32) 
tf.Tensor([1 3 5 7], shape=(4,), dtype=int32)
tf.Tensor([1 4 6], shape=(3,), dtype=int32)

Torch

# Torch

c = torch.tensor([1,2,3,4,5,6,7,8])
print(c[:])
print(c[2:-3])
print(c[3:-1])
print(c[::2])

indices = torch.tensor([0,3,5])
x_indices = c[indices]
print(x_indices)

############################# OUTPUT ##############################

tensor([1, 2, 3, 4, 5, 6, 7, 8]) 
tensor([3, 4, 5]) 
tensor([4, 5, 6, 7]) 
tensor([1, 3, 5, 7])
tensor([1, 4, 6])

12. 索引和切片（2D + 矩阵）

> Matrix Slicing

Numpy

# Numpy

a = np.array([[1,2,3],[4,5,6],[7,8,9]])

# Matrix Indexing
# Print all individual Rows and Columns
print("Row-1",a[0, :])
print("Row-2",a[1, :])
print("Row-3",a[2, :])
print("Col-1",a[:, 0])
print("Col-2",a[:, 1])
print("Col-3",a[:, 2])

# Print the sub-diagonal matrix
print("Upper-Left",a[0:2,0:2])
print("Upper-Right",a[0:2,1:3])
print("Bottom-Left",a[1:3,0:2])
print("Bottom-Right",a[1:3,1:3])

############################# OUTPUT ##############################

Row-1 [1 2 3]
Row-2 [4 5 6] 
Row-3 [7 8 9]
Col-1 [1 4 7] 
Col-2 [2 5 8] 
Col-3 [3 6 9]

Upper-Left [[1 2]
            [4 5]]
Upper-Right [[2 3]
             [5 6]]
Bottom-Left [[4 5]
             [7 8]]
Bottom-Right [[5 6]
              [8 9]]

Tensorflow

# TensorFlow

b = tf.constant([[1,2,3],[4,5,6],[7,8,9]])

# Matrix Indexing
# Print all individual Rows and Columns
print("Row-1",b[0, :])
print("Row-2",b[1, :])
print("Row-3",b[2, :])
print("Col-1",b[:, 0])
print("Col-2",b[:, 1])
print("Col-3",b[:, 2])

# Print the sub-diagonal matrix
print("Upper-Left",b[0:2,0:2])
print("Upper-Right",b[0:2,1:3])
print("Bottom-Left",b[1:3,0:2])
print("Bottom-Right",b[1:3,1:3])

############################# OUTPUT ##############################

Row-1 tf.Tensor([1 2 3], shape=(3,), dtype=int32)
Row-2 tf.Tensor([4 5 6], shape=(3,), dtype=int32)
Row-3 tf.Tensor([7 8 9], shape=(3,), dtype=int32)
Col-1 tf.Tensor([1 4 7], shape=(3,), dtype=int32) 
Col-2 tf.Tensor([2 5 8], shape=(3,), dtype=int32) 
Col-3 tf.Tensor([3 6 9], shape=(3,), dtype=int32)

Upper-Left tf.Tensor( [[1 2]  [4 5]], shape=(2, 2), dtype=int32) 
Upper-Right tf.Tensor( [[2 3]  [5 6]], shape=(2, 2), dtype=int32)  
Bottom-Left tf.Tensor( [[4 5]  [7 8]], shape=(2, 2), dtype=int32) 
Bottom-Right tf.Tensor( [[5 6]  [8 9]], shape=(2, 2), dtype=int32)

Torch

# Torch

c = torch.tensor([[1,2,3],[4,5,6],[7,8,9]])

# Matrix Indexing
# Print all individual Rows and Columns
print("Row-1",c[0, :])
print("Row-2",c[1, :])
print("Row-3",c[2, :])
print("Col-1",c[:, 0])
print("Col-2",c[:, 1])
print("Col-3",c[:, 2])

# Print the sub-diagonal matrix
print("Upper-Left",c[0:2,0:2])
print("Upper-Right",c[0:2,1:3])
print("Bottom-Left",c[1:3,0:2])
print("Bottom-Right",c[1:3,1:3])

############################# OUTPUT ##############################

Row-1 tensor([1, 2, 3]) 
Row-2 tensor([4, 5, 6]) 
Row-3 tensor([7, 8, 9])
Col-1 tensor([1, 4, 7]) 
Col-2 tensor([2, 5, 8]) 
Col-3 tensor([3, 6, 9])

Upper-Left tensor([[1, 2],[4, 5]]) 
Upper-Right tensor([[2, 3],[5, 6]])  
Bottom-Left tensor([[4, 5],[7, 8]]) 
Bottom-Right tensor([[5, 6],[8, 9]])

13. 轴的重塑和转置

> Reshape & Transpose

Numpy

# Numpy

a = np.arange(9)
print(a)
a = np.reshape(a, (3,3))
print(a)
a = np.transpose(a, (1,0)) # Swap axes (1,0)
print(a)

############################# OUTPUT ##############################

[0 1 2 3 4 5 6 7 8]
[[0 1 2]  
 [3 4 5]  
 [6 7 8]]
[[0 3 6]
 [1 4 7]
 [2 5 8]]

Tensorflow

# TensorFlow

b = tf.range(9)
print(b)
b = tf.reshape(b, (3,3))
print(b)
b = tf.transpose(b, perm=[1,0]) # Swap axes in perm (1,0)
print(b)

############################# OUTPUT ##############################

tf.Tensor([0 1 2 3 4 5 6 7 8], shape=(9,), dtype=int32)
tf.Tensor( [[0 1 2]  [3 4 5]  [6 7 8]], shape=(3, 3), dtype=int32) 
tf.Tensor( [[0 3 6]  [1 4 7]  [2 5 8]], shape=(3, 3), dtype=int32)

Torch

# Torch

c = torch.arange(9)
print(c)
c = torch.reshape(c, (3,3))
print(c)
c = c.permute(1,0) # Swap axes in perm (1,0)

############################# OUTPUT ##############################

tensor([0, 1, 2, 3, 4, 5, 6, 7, 8]) 
tensor([[0, 1, 2],         [3, 4, 5],         [6, 7, 8]]) 
tensor([[0, 3, 6],         [1, 4, 7],         [2, 5, 8]])

14. 串联

> Matrix Concatenation

Numpy

# Numpy

a = np.array([[1, 2], [3, 4]])
print("a",a)
b = np.array([[5, 6]])
print("b",b)

d = np.concatenate((a, b), axis=0)
print("Concat (axis=0 - Row)")
print(d)

e = np.concatenate((a, b.T), axis=1)
print("Concat (axis=1 - Column)")
print(e)

############################# OUTPUT ##############################

a [[1 2]  [3 4]] 
b [[5 6]]

Concat (axis=0 - Row) 
[[1 2]  
 [3 4]  
 [5 6]]

Concat (axis=1 - Column) 
[[1 2 5]  
 [3 4 6]]

Tensorflow

# TensorFlow

x = tf.constant([[1, 2], [3, 4]])
print("x",x)
y = tf.constant([[5, 6]])
print("y",y)

z = tf.concat((x, y), axis=0)
print("Concat (axis=0 - Row)")
print(z)

z = tf.concat((x, tf.transpose(y)), axis=1)
print("Concat (axis=1 - Column)")
print(z)

############################# OUTPUT ##############################

x tf.Tensor( [[1 2]  [3 4]], shape=(2, 2), dtype=int32) 
y tf.Tensor([[5 6]], shape=(1, 2), dtype=int32)

Concat (axis=0 - Row) 
tf.Tensor( [[1 2]  [3 4]  [5 6]], shape=(3, 2), dtype=int32)

Concat (axis=1 - Column) 
tf.Tensor( [[1 2 5]  [3 4 6]], shape=(2, 3), dtype=int32)

Torch

# Torch

t = torch.tensor([[1, 2], [3, 4]])
print("x",t)
u = torch.tensor([[5, 6]])
print("y",u)

v = torch.cat((t , u), axis=0)
print("Concat (axis=0 - Row)")
print(v)

v = torch.cat((t , u.T), axis=1)
print("Concat (axis=1 - Column)")
print(v)

############################# OUTPUT ##############################

x tensor([[1, 2], [3, 4]]) 
y tensor([[5, 6]])

Concat (axis=0 - Row) 
tensor([[1, 2],
        [3, 4],         
        [5, 6]])

Concat (axis=1 - Column) 
tensor([[1, 2, 5],
        [3, 4, 6]])

15. 跨轴求和

> Axes — Sum

Numpy

# Numpy

a = np.array([[1,2,3,4,5], [10,10,10,10,10]])
print(a)

print("Overall flattened Sum", a.sum())
print("Sum across Columns",a.sum(axis=0))
print("Sum across Rows",a.sum(axis=1))

############################# OUTPUT ##############################

[[ 1  2  3  4  5]  [10 10 10 10 10]]
Overall flattened Sum 65 
Sum across Columns [11 12 13 14 15]
Sum across Rows [15 50]

Tensorflow

# TensorFlow

b = tf.constant([[1,2,3,4,5], [10,10,10,10,10]])
print(b)

print("Overall flattened Sum",tf.math.reduce_sum(b))
print("Sum across Columns",tf.math.reduce_sum(b, axis=0))
print("Sum across Rows",tf.math.reduce_sum(b, axis=1))

############################# OUTPUT ##############################

tf.Tensor([[ 1  2  3  4  5]  [10 10 10 10 10]], shape=(2, 5), dtype=int32)

Overall flattened Sum tf.Tensor(65, shape=(), dtype=int32) 
Sum across Columns tf.Tensor([11 12 13 14 15], shape=(5,), dtype=int32) 
Sum across Rows tf.Tensor([15 50], shape=(2,), dtype=int32)

Torch

# Torch

c = torch.tensor([[1,2,3,4,5], [10,10,10,10,10]])
#print(c)

print("Overall flattened Sum",torch.sum(c))
print("Sum across Columns",torch.sum(c, axis=0))
print("Sum across Rows",torch.sum(c, axis=1))

############################# OUTPUT ##############################

Overall flattened Sum tensor(65) 
Sum across Columns tensor([11, 12, 13, 14, 15]) 
Sum across Rows tensor([15, 50])

16. 轴的均值

Numpy

# Numpy

a = np.array([[1,2,3,4,5], 
               [10,10,10,10,10]])
print(a)

print("Overall flattened mean", a.mean())
print("Sum across Columns",a.mean(axis=0))
print("Sum across Rows",a.mean(axis=1))

############################# OUTPUT ##############################

[[ 1  2  3  4  5]  [10 10 10 10 10]]
Overall flattened mean 6.5 
Sum across Columns [5.5 6.  6.5 7.  7.5] 
Sum across Rows [ 3. 10.]

Tensorflow

# TensorFlow

b = tf.constant([[1,2,3,4,5], 
                 [10,10,10,10,10]])
print(b)

print("Overall flattened mean",tf.math.reduce_mean(b))
print("Sum across Columns",tf.math.reduce_mean(b, axis=0))
print("Sum across Rows",tf.math.reduce_mean(b, axis=1))

############################# OUTPUT ##############################

tf.Tensor( [[ 1  2  3  4  5]  [10 10 10 10 10]], shape=(2, 5), dtype=int32)
Overall flattened mean tf.Tensor(6, shape=(), dtype=int32) 
Sum across Columns tf.Tensor([5 6 6 7 7], shape=(5,), dtype=int32) Sum across Rows tf.Tensor([ 3 10], shape=(2,), dtype=int32)

Torch

# Torch

c = torch.tensor([[1,2,3,4,5], [10,10,10,10,10]], dtype=torch.float32)
print(c)

print("Overall flattened mean",torch.mean(c))
print("Sum across Columns",torch.mean(c, axis=0))
print("Sum across Rows",torch.mean(c, axis=1))

############################# OUTPUT ##############################

tensor([[ 1.,  2.,  3.,  4.,  5.],
        [10., 10., 10., 10., 10.]])  
Overall flattened mean tensor(6.5000) 
Sum across Columns tensor([5.5000, 6.0000, 6.5000, 7.0000, 7.5000])
Sum across Rows tensor([ 3., 10.])

17. 尺寸扩展和移动轴

> Concat and move axes

Numpy

# Numpy

a = np.full((3,3),10.0)
print(a)
print(a.shape)
a = np.expand_dims(a, axis=0)
print(a)
print(a.shape)

b = np.full((3,3),20.0)
print(b)
b = np.expand_dims(b, axis=0)
print(b.shape)

c = np.concatenate((a,b), axis=0)
c = np.moveaxis(c,2,0) # Move 2nd dimension to 0th position
print(c)
print(c.shape)

############################# OUTPUT ##############################

[[10. 10. 10.]  
 [10. 10. 10.]  
 [10. 10. 10.]] 
(3, 3)

[[[10. 10. 10.]   
  [10. 10. 10.]   
  [10. 10. 10.]]] 
(1, 3, 3)

[[20. 20. 20.]  
 [20. 20. 20.]  
 [20. 20. 20.]] 
(1, 3, 3)

[[[10. 10. 10.]   [20. 20. 20.]]
 [[10. 10. 10.]   [20. 20. 20.]]
 [[10. 10. 10.]   [20. 20. 20.]]]
(3, 2, 3)

Tensorflow

# TensorFlow

x = tf.fill((3,3),10.0)
print(x)
print(x.shape)
x = tf.expand_dims(x, axis=0)
print(x.shape)

y = tf.fill((3,3),20.0)
print(y)
print(y.shape)
y = tf.expand_dims(y, axis=0)
print(y.shape)

z = tf.concat((x,y), axis=0)
z = tf.transpose(z, [1, 0, 2])
print(z.shape)

############################# OUTPUT ##############################

tf.Tensor( [[10. 10. 10.]  
            [10. 10. 10.]  
            [10. 10. 10.]], shape=(3, 3), dtype=float32) 
(3, 3) 
(1, 3, 3) 

tf.Tensor( [[20. 20. 20.]  
            [20. 20. 20.]  
            [20. 20. 20.]], shape=(3, 3), dtype=float32) 

(3, 3)
(1, 3, 3)
(3, 2, 3)

Torch

# Torch

m1 = torch.ones((2,), dtype=torch.int32)
m1 = m1.new_full((3, 3), 10)
m1 = torch.unsqueeze(m1, axis=0)
print(m1)
print(m1.shape)

m2 = torch.ones((2,), dtype=torch.int32)
m2 = m2.new_full((3, 3), 20)
print(m2)
m2 = torch.unsqueeze(m2, axis=0)
print(m2.shape)

m = torch.cat((m1,m2), axis=0)
m = m.permute([1,0,2])
print(m)
print(m.shape)

############################# OUTPUT ##############################

tensor([[[10, 10, 10],
         [10, 10, 10],
         [10, 10, 10]]], dtype=torch.int32)
torch.Size([1, 3, 3])

tensor([[20, 20, 20],
        [20, 20, 20],
        [20, 20, 20]], dtype=torch.int32)
torch.Size([1, 3, 3])

tensor([[[10, 10, 10],
         [20, 20, 20]],
        
        [[10, 10, 10],
         [20, 20, 20]],

        [[10, 10, 10],
         [20, 20, 20]]], dtype=torch.int32)
torch.Size([3, 2, 3])

最大最小值

> Max for axis=0

> Max for axis=1

Numpy

# Numpy

a = np.array([[5,10,15],
              [20,25,30]])
b = np.array([[6,69,35],
              [70,10,82]])
c = np.array([[25,45,48],
              [4,100,89]])
print(a)

final = np.zeros((3,2,3))
print(final.shape)
final[0, :, :] = a
final[1, :, :] = b
final[2, :, :] = c
print(final)

print("Overall flattened max", final.max())
print("max across Columns",final.max(axis=0))
print("max across Rows",final.max(axis=1))
print("Index of max value across the flattened max", final.argmax())
print("Index of max value across Columns",final.argmax(axis=0))
print("Index of max value across Rows",final.argmax(axis=1))

############################# OUTPUT ##############################

[[ 5 10 15]  
 [20 25 30]]
(3, 2, 3)

[[[  5.  10.  15.]
  [ 20.  25.  30.]]
[[  6.  69.  35.]
 [ 70.  10.  82.]]
[[ 25.  45.  48.]
 [  4. 100.  89.]]]

Overall flattened max 100.0

max across Columns 
[[ 25.  69.  48.]  
 [ 70. 100.  89.]]

max across Rows 
[[ 20.  25.  30.]  
 [ 70.  69.  82.]  
 [ 25. 100.  89.]]

Index of max value across the flattened max 16 
Index of max value across Columns 
[[2 1 2]  
[1 2 2]]

Index of max value across Rows 
[[1 1 1]  
 [1 0 1]
 [0 1 1]]

Tensorflow

# Tensorflow

final = tf.constant([[[5,10,15],
                     [20,25,30]],
                     [[6,69,35],
                      [70,10,82]],
                     [[25,45,48],
                      [4,100,89]]])
print(final)

print("Overall flattened max", tf.math.reduce_max(final))
print("max across Columns",tf.math.reduce_max(final,axis=0))
print("max across Rows",tf.math.reduce_max(final, axis=1))
print("Index of max value across the flattened max", tf.math.argmax(final))
print("Index of max value across Columns",tf.math.argmax(final, axis=0))
print("Index of max value across Rows",tf.math.argmax(final, axis=1))

############################# OUTPUT ##############################

tf.Tensor( [[[  5  10  15]
             [ 20  25  30]]
            [[  6  69  35]
             [ 70  10  82]] 
            [[ 25  45  48]
             [  4 100  89]]], shape=(3, 2, 3), dtype=int32)

Overall flattened max tf.Tensor(100, shape=(), dtype=int32) 
max across Columns 
tf.Tensor( [[ 25  69  48]  
            [ 70 100  89]], shape=(2, 3), dtype=int32)

max across Rows 
tf.Tensor( [[ 20  25  30]  
            [ 70  69  82]  
            [ 25 100  89]], shape=(3, 3), dtype=int32)

Index of max value across the flattened max 
tf.Tensor( [[2 1 2]  
           [1 2 2]], shape=(2, 3), dtype=int64)

Index of max value across Columns 
tf.Tensor( [[2 1 2]  
           [1 2 2]], shape=(2, 3), dtype=int64)

Index of max value across Rows 
tf.Tensor( [[1 1 1]  
            [1 0 1]
            [0 1 1]], shape=(3, 3), dtype=int64)

Torch

# Torch

final = torch.tensor([[[5,10,15],
                      [20,25,30]],
                     [[6,69,35],
                      [70,10,82]],
                     [[25,45,48],
                      [4,100,89]]])

print(final)
print("Overall flattened max", torch.max(final))
print("max across Columns",torch.max(final,axis=0))
print("max across Rows",torch.max(final, axis=1))
print("Index of max value across the flattened max", torch.argmax(final))
print("Index of max value across Columns",torch.argmax(final, axis=0))
print("Index of max value across Rows",torch.argmax(final, axis=1))

############################# OUTPUT ##############################

tensor([[[  5,  10,  15],
         [ 20,  25,  30]],          
        [[  6,  69,  35],     
         [ 70,  10,  82]],    
        [[ 25,  45,  48],       
         [  4, 100,  89]]])

Overall flattened max tensor(100) 

max across Columns torch.return_types.max( values=
tensor([[ 25,  69,  48],
        [ 70, 100,  89]]), 
indices=tensor([[2, 1, 2],
               [1, 2, 2]]))

max across Rows torch.return_types.max( values=
tensor([[ 20,  25,  30],         
        [ 70,  69,  82],         
        [ 25, 100,  89]]), 
indices=tensor([[1, 1, 1],         
                [1, 0, 1],         
                [0, 1, 1]]))

Index of max value across the flattened max tensor(16) 
Index of max value across Columns 
tensor([[2, 1, 2], 
        [1, 2, 2]])

Index of max value across Rows 
tensor([[1, 1, 1],
        [1, 0, 1],        
        [0, 1, 1]])

19. 切片和索引（3D矩阵）

> 3x3 Matrix and its indices

> Upper-Left & Lower-Right

> Middle Elements & Inverse Middle Element

Numpy

# Numpy

a = np.array([[10,10,10],[10,10,10],
              [10,10,10],[10,10,10]])
b = np.array([[20,20,20],[20,20,20],
              [20,20,20],[20,20,20]])
c = np.array([[30,30,30],[30,30,30],
              [30,30,30],[30,30,30]])
final = np.zeros((3,4,3))
final[0, :, :] = a
final[1, :, :] = b
final[2, :, :] = c

print("Upper-Left",final[:, 0:2, 0:2])
print("Lower-Right",final[:, 2:, 1:])
print("Middle Elements",final[:,1:3, 1])

# Ignore Middle Elements
a = np.array([[1,2,3],[4,5,6],[7,8,9],[10,11,12]])
b = np.array([[13,14,15],[16,17,18],[19,20,21],[22,23,24]])
c = np.array([[25,26,27],[28,29,30],[31,32,33],[34,35,36]])

final = np.zeros((3,4,3))
print(final.shape)
final[0, :, :] = a
final[1, :, :] = b
final[2, :, :] = c

# Though may work,but not efficient
print("Ignore Middle",final[:,[0,0,0,1,1,2,2,3,3,3], [0,1,2,0,2,0,2,0,1,2]])
      
############################# OUTPUT ##############################

Upper-Left [[[10. 10.]
  [10. 10.]]

 [[20. 20.]
  [20. 20.]]

 [[30. 30.]
  [30. 30.]]]

Lower-Right [[[10. 10.]
  [10. 10.]]

 [[20. 20.]
  [20. 20.]]

 [[30. 30.]
  [30. 30.]]]

Middle Elements [[10. 10.]
 [20. 20.]
 [30. 30.]]
(3, 4, 3)

Ignore Middle [[ 1.  2.  3.  4.  6.  7.  9. 10. 11. 12.]
 [13. 14. 15. 16. 18. 19. 21. 22. 23. 24.]
 [25. 26. 27. 28. 30. 31. 33. 34. 35. 36.]]

Tensorflow

# Tensorflow

final = tf.constant([[[10,10,10],[10,10,10],
                  [10,10,10],[10,10,10]],

                  [[20,20,20],[20,20,20],
                  [20,20,20],[20,20,20]],

                  [[30,30,30],[30,30,30],
                  [30,30,30],[30,30,30]]])

print("Upper-Left",final[:, 0:2, 0:2])
print("Lower-Right",final[:, 2:, 1:])
print("Middle Elements",final[:,1:3, 1])

############################# OUTPUT ##############################

Upper-Left tf.Tensor(
[[[10 10]
  [10 10]]

 [[20 20]
  [20 20]]

 [[30 30]
  [30 30]]], shape=(3, 2, 2), dtype=int32)

Lower-Right tf.Tensor(
[[[10 10]
  [10 10]]

 [[20 20]
  [20 20]]

 [[30 30]
  [30 30]]], shape=(3, 2, 2), dtype=int32)

Middle Elements tf.Tensor(
[[10 10]
 [20 20]
 [30 30]], shape=(3, 2), dtype=int32)

Torch

# Torch

a = torch.Tensor([[10,10,10],[10,10,10],
              [10,10,10],[10,10,10]])

b = torch.Tensor([[20,20,20],[20,20,20],
              [20,20,20],[20,20,20]])

c = torch.Tensor([[30,30,30],[30,30,30],
              [30,30,30],[30,30,30]])
final = np.zeros((3,4,3))
final[0, :, :] = a
final[1, :, :] = b
final[2, :, :] = c

print("Upper-Left",final[:, 0:2, 0:2])
print("Lower-Right",final[:, 2:, 1:])
print("Middle Elements",final[:,1:3, 1])

# Ignore Middle Elements
a = torch.Tensor([[1,2,3],[4,5,6],[7,8,9],[10,11,12]])
b = torch.Tensor([[13,14,15],[16,17,18],[19,20,21],[22,23,24]])
c = torch.Tensor([[25,26,27],[28,29,30],[31,32,33],[34,35,36]])

final = np.zeros((3,4,3))
print(final.shape)
final[0, :, :] = a
final[1, :, :] = b
final[2, :, :] = c
# Though may work,but not efficient
print("Ignore Middle",final[:,[0,0,0,1,1,2,2,3,3,3], [0,1,2,0,2,0,2,0,1,2]])

############################# OUTPUT ##############################

Upper-Left [[[10. 10.]
  [10. 10.]]

 [[20. 20.]
  [20. 20.]]

 [[30. 30.]
  [30. 30.]]]

Lower-Right [[[10. 10.]
  [10. 10.]]

 [[20. 20.]
  [20. 20.]]

 [[30. 30.]
  [30. 30.]]]

Middle Elements [[10. 10.]
 [20. 20.]
 [30. 30.]]
(3, 4, 3)

Ignore Middle [[ 1.  2.  3.  4.  6.  7.  9. 10. 11. 12.]
 [13. 14. 15. 16. 18. 19. 21. 22. 23. 24.]
 [25. 26. 27. 28. 30. 31. 33. 34. 35. 36.]]

本文内容由网友自发贡献，版权归原作者所有，本站不承担相应法律责任。如您发现有涉嫌抄袭侵权的内容，请联系:hwhale#tublm.com(使用前将#替换为@)

NumPy

torch

tensorflow

代码对比

NumPy、Torch和Tensorflow 代码对比的相关文章

Keras Maxpooling2d 层给出 ValueError

我正在尝试在 keras 中复制 VGG16 模型以下是我的代码 model Sequential model add ZeroPadding2D 1 1 input shape 3 224 224 model add Convoluti
使用基于 ConvLSTM2D 的 Keras 模型从较低分辨率图像估计高分辨率图像

我正在尝试使用以下内容ConvLSTM2D从低分辨率图像序列估计高分辨率图像序列的架构 import numpy as np scipy ndimage matplotlib pyplot as plt from keras models
Keras 中的 Tensorflow 自定义损失函数 - 张量循环

我正在尝试在 Keras 中编写自定义损失函数如下所示 Keras 中的自定义损失函数 https stackoverflow com questions 43818584 custom loss function in keras 我的
NumPy 使用索引列表选择每行的特定列索引

我正在努力选择 NumPy 矩阵每行的特定列假设我有以下矩阵我称之为X 1 2 3 4 5 6 7 8 9 我也有一个list我称之为每行的列索引Y 1 0 2 我需要获取值 2 4 9 代替list带索引Y 我还可以生成一个形状相同的
scipy.stats....rvs 和 numpy.random 随机抽取之间的区别

看起来是否是相同的分布从中抽取随机样本numpy random比这样做更快scipy stats rvs 我想知道是什么原因导致两者之间的速度差异 scipy stats uniform实际上使用了numpy 这里是stats中对应的函数
Tensorflow批处理：将结果保留为字符串

这个简单的程序 import tensorflow as tf input string batch tf train batch tf constant input batch size 1 with tf Session as sess
奇异矩阵 - python

下面的代码显示了矩阵的奇异性问题因为在 Pycharm 中工作我得到 raise LinAlgError Singular matrix numpy linalg linalg LinAlgError Singular matrix 我猜
在 Chaquopy 中转换数组和张量

我该怎么做呢我看到你的帖子说你可以将 java 对象传递给 Python 方法但这不适用于 numpy 数组和 TensorFlow 张量以下以及其各种变体是我尝试过的但没有成功 double anchors new double
当我想在电脑中加载该模型时，我可以在 colab bu 中加载我的深度模型，但我不能

我在colab中通过keras 2 3 1和tensorflow 2 1 0训练了一个深度模型我用JSON和Keras保存了我的模型 saveWeightPath content drive My Drive model info mod
如何在Pycharm中运行Tensorflow GPU？

我想在 Linux Mint 上的 Pycharm 中运行 Tensorflow GPU 我尝试了一些像这样的指南 https medium com p venkata kishore install anaconda tenserflow
如何使用 pandas 选择所有非 NaN 列和非 NaN 最后一列？

如果标题有点令人困惑请原谅我假设我有test h5 下面是使用读取该文件的结果df read hdf test h5 testdata 0 1 2 3 4 5 6 0 123 444 111 321 NaN NaN NaN 1 12 2
向量化 numpy bincount

我有一个 2d numpy 数组 A我要申请np bincount 到矩阵的每一列A生成另一个二维数组B由原始矩阵每列的 bincounts 组成A 我的问题是 np bincount 是一个采用一维数组的函数它不是像这样的数组方法B A
如何“缩放”numpy 数组？

我想将形状 h w 的数组缩放 n 倍从而得到形状 h n w n 的数组其中假设我有一个 2x2 数组 array 1 1 0 1 我想将数组缩放为 4x4 array 1 1 1 1 1 1 1 1 0 0 1 1 0 0 1 1
打印总和时出现“内置方法总和”奇怪的消息

使用 Python NumPy 计算矩阵的列之和 import numpy from StringIO import StringIO fileName test2 csv myFile open fileName r print Read
如何将one-hot向量转换为多标签？

我有一项多分类任务并且我得到了像这样的单热类型预测 0 1 1 0 1 0 1 0 1 我希望将这个单热向量转换为标签例如 1 2 1 0 2 我已经尝试过 tf argmax 但它不起作用那么我该如何处理呢使用列表理解 oheLi
DataFrame 中的字符串，但 dtype 是对象

为什么 Pandas 告诉我我有对象尽管所选列中的每个项目都是一个字符串即使在显式转换之后也是如此这是我的数据框
以编程方式将列名称添加到 numpy ndarray

我正在尝试将列名称添加到 numpy ndarray 然后按名称选择列但这不起作用我无法判断问题是在添加名称时出现还是在稍后尝试调用它们时出现这是我的代码 data np genfromtxt csv file delimiter
这可能是因为 cuDNN 初始化失败，因此请尝试查看上面是否打印了警告日志消息。 [操作：Conv2D]

我在 anaconda 中安装了 TensorFlow GPU 2 0 当我安装它并导入包然后运行我的 CNN 模型时它工作正常但当我尝试运行训练模型时出现错误这是我的错误报告 Epoch 1 50 UnknownError Tr
如何解码 dtype=numpy.string_ 的 numpy 数组？

我需要使用 Python 3 解码按以下方式编码的字符串 gt gt gt s numpy asarray numpy string hello nworld gt gt gt s array b hello nworld dtype S1
本地设置的 Cython 编译器指令是否影响一个或所有函数？

我正在努力使用 Cython 加速一些 Python Numpy 代码并且对本地设置如定义的here http docs cython org en latest src reference compilation html在文档中

随机推荐

反光板导航SLAM（二）VEnus代码浅析

上一章简单介绍了VEnus中几个主要函数的作用 xff0c 这里详细展开看一下每个函数的具体思路 xff0c 通过研究具体的代码我们可以简单了解VEnus中对于反光柱定位的具体流程 1 IntensityExtraction Extract
TEB算法详解（TebLocalPlannerROS::computeVelocityCommands（3））

第一章主要分析了teb算法的准备条件 xff0c 包括获取当前位姿与速度对全局路径的裁减以获取局部路径等以及局部地图的获取等第二章主要讲述了在获取前置条件后 xff0c 如何根据前置条件进行位姿优化 xff0c teb的路径优化主要是调
TEB算法详解参数详解

teb算法的基本思路之前已经看完了 xff0c 今天主要看一下teb算法的参数配置文件 xff0c 分析一下每个配置参数的作用 xff1a teb的参数主要可以包含以下几个部分 xff1a 1 Trajectory Trajectory的参
ROS 如何用自己的电脑连接其他设备查看Rviz

Rviz的使用在ROS开发中是很常见的东西 xff0c 有时候为了方便我们需要查看某台设备的一些数据 xff0c 但是又没有一些直观的连接工具例如向日葵之类的这些工具是要基于屏幕的 xff0c 如果设备上没有屏幕是无法使用的那在只有ss
调试时正常，maven install 找不到符号

找不到符号很可能是引用的自己的库没有加载最新版把所有库都clean gt install一下或root项目下clean gt install
【springboot】org.aspectj.lang.ProceedingJoinPoint获取返回类型、参数名称/值等

1 参数值 span class token class name Object span span class token punctuation span span class token punctuation span args s
no opencv_java455 in java.library.path

直接把 opencv build java x64下的文件opencv java455 dll复制到Java JDK安装的bin目录下就解决了这个问题
openssl pem 生成公钥和私钥及文件

下载 xff1a http slproweb com products Win32OpenSSL html 首先得安装OpenSSL软件包openssl xff0c 安装了这个软件包之后 xff0c 我们可以做这些事情 xff1a o Cr
DebugDiag 2 收集Dump

DebugDiag简介 Debug Diagnostic Tool DebugDiag 是微软提供的工具 xff0c 可以用来追踪windows平台下的程序崩溃 xff0c 卡死 xff0c 内存泄漏等一些疑难问题的原因 xff0c 按照问
Stream List 的排序

1 list的正序 span class token class name List span span class token generics span class token punctuation lt span span clas
嵌入式WebSocket的实现

系统硬件采用STM32 43 DM9000AEP xff0c 具体实现步骤如下 xff0c 第一步建立TCP监听端口可以随意 xff0c 等待浏览器发出http请求 Chrome浏览器发出的请求第二步建立握手服务端需要解析最少4个头字段
spring boot 注册成 windows 服务

1 windows系统下将 34 java jar 34 注册成 windows 服务 demo bat 64 echo off d span class token builtin class name cd span d span cl
Golang UI

GUI库 Fyne 跨平台 xff0c 移动端相对复杂 https github com fyne io fyne Walk 精于 windows https github com lxn walk
解决golang编译提示dial tcp 172.217.160.113:443: connectex: A connection attempt failed

废话不多说 xff0c 直接开讲还多新手朋友在编译go的过程中是不是经常会遇到这样的报错 module Get https proxy golang org dial tcp 172 217 160 113 443 connectex A
更新后报错

更新后 xff0c 运行程序就报这个错 no required module provides package xxx go mod file not found in current directory or any parent dir
go: go.mod file not found in current directory or any parent directory； see ‘go help modules‘

go go mod file not found in current directory or any parent directory see go help modules 此时还需要初始化Go moudle go mod init
vc++ 6.0bug C1853

c documents and settingsstudent桌面musicplayermaindlg cpp 1 fatal error C1853 Debug MusicPlayer pch is not a precompiled h
人工智能知识点汇总

一 AI应用领域 AI目前主要的应用领域有3个方向 xff0c 包括 xff1a 计算机视觉语音交互自然语言处理 1 1 计算机视觉 xff08 CV xff09 计算机视觉是一门研究如何使机器看的科学 xff0c 就是指用摄影机和
Java17(291)之后 , 禁用了TLS1.1 , 使JDBC无法用SSL连接SqlServer怎么办,以下是解决办法

修改java security文件 1 找到jre的java security文件 2 打开java security并搜索 jdk tls disabledAlgorithms 61 3 删掉TLSv1 TLSv1 1 4 保存可以了
NumPy、Torch和Tensorflow 代码对比

深度学习在深入学习的基本单位上实施初级到高级操作 gt Excerpts 我习惯于为不同的问题创建新的深度学习架构 xff0c 但选择哪个框架 xff08 Keras Pytorch TensorFlow xff09 通常比较困难由于其