Pytorch 04: Random Tensors
Using torch.rand() and passing size() parameter
torch.rand is a function in PyTorch used to generate a tensor filled with random numbers sampled from a uniform distribution over the interval [0, 1). It is useful when you need random values for initialization or stochastic processes (discussed later) in machine learning and deep learning applications.
Let’s create a matrix using size() parameter.
# creating a random tensor of size --> (3, 3)
random_matrix = torch.rand(size = (3, 3))
random_matrix, random_matrix.dtype, random_matrix.size()
(tensor([[0.6532, 0.0823, 0.4880],
[0.3133, 0.5036, 0.1455],
[0.7320, 0.8213, 0.5246]]),
torch.float32,
torch.Size([3, 3]))
or we can create a tensor :
# creating a random tensor of size --> (1, 3, 3)
random_tensor = torch.rand(size = (1, 3, 3))
random_tensor, random_tensor.dtype, random_tensor.size()
(tensor([[[0.2552, 0.3850, 0.0569],
[0.3506, 0.9072, 0.4009],
[0.9173, 0.9203, 0.3054]]]),
torch.float32,
torch.Size([1, 3, 3]))
You can also generate a random tensor in the common image size of (244, 244, 3) ([height, width, colour_channels]).
and, torch.float32 is just the default data type for tensors.