[pytorch learning notes] tensor

Plate purpose ：

Systematization Pytorch The learning process , Only record dry goods

tensor

Tensor and ndarrays similar , Both can usually share the same underlying memory , This eliminates the need to replicate data .

import torch
import numpy as np

tensor initialization

import torch
import numpy as np

##  Initialization tensor 

#  Directly from data 
data = [[1,2], [3,4]]
x_data = torch.tensor(data)

# numpy  turn torch
np_array = np.array(data)
x_np = torch.from_numpy(np_array)

#  From the other tensor
x_ones = torch.ones_like(x_data)
x_rand = torch.rand_like(x_data,dtype=torch.float)

#  Use random or constant values 
shape = (2,3)
rand_tensor = torch.rand(shape)
ones_tensor = torch.ones(shape)
zeros_tensor = torch.zeros(shape)

# tensor attribute 
tensor = torch.rand(3,4)
print(tensor.shape)  #  dimension 
print(tensor.dtype)  #  data type 
print(tensor.device) # tensor Operation unit 

tensor operation

#  section 
tensor = torch.ones(4,4)
print(tensor[0]) #  first line 
print(tensor[:,0]) #  First column 
print(tensor[...,-1]) #  The last line 
tensor[:,1] = 0
print(tensor)
''' tensor([1., 1., 1., 1.]) tensor([1., 1., 1., 1.]) tensor([1., 1., 1., 1.]) tensor([[1., 0., 1., 1.], [1., 0., 1., 1.], [1., 0., 1., 1.], [1., 0., 1., 1.]]) '''
#  Splicing 
t1 = torch.cat([tensor,tensor,tensor],dim=1)  #  Stack by column 
print(t1)

#  Arithmetic operations 
y1 = tensor @ tensor.T  # @  Point multiplication  .T Transposition 
y2 = tensor.matmul(tensor.T)
y3 = torch.rand_like(y1)

torch.matmul(tensor, tensor.T,out=y3)  # tensor Point multiplication   Output y3

z1 = tensor * tensor
z2 = tensor.mul(tensor)
z3 = torch.rand_like(tensor)
torch.mul(tensor,tensor,out=z3)

# tensor Aggregate into a value 
agg = tensor.sum()# <class 'torch.Tensor'>
agg_item = agg.item()#<class 'float'>
print(type(agg_item))


#  Local operation 
tensor.add_(5)  # _ Indicates in situ operation , Directly modifying tensor, Can save memory , But there will be problems when calculating the derivative , Loss of raw data 
print(tensor)


# cpu and numpy Upper tensor Can share their underlying memory location 
t = torch.ones(5) # torch.tensor
n = t.numpy()# ndarray

t.add_(1)
print(t)# tensor([2., 2., 2., 2., 2.])
print(n)# [2. 2. 2. 2. 2.]