Just my scrappy cheat sheet for looking up things in pytorch that I know in numpy - until I find a better cheatsheet.
Even though there are a lot of similarities in syntax between the two, occasionally something is different that throws me off.
| Numpy | PyTorch | Notes |
|---|---|---|
np.empty((2, 2)) |
torch.empty(5, 3) |
empty array |
np.random.rand(3,2) |
torch.rand(5, 3) |
random |
np.zeros((5,3)) |
torch.zeros(5, 3) |
zeros |
np.array([5.3, 3]) |
torch.tensor([5.3, 3]) |
from list |
np.random.randn(*a.shape) |
torch.randn_like(a) |
|
np.arange(16) |
torch.range(0,15) |
array starting from 0 ending at 15 (inclusive) |
| Numpy | PyTorch | Notes |
|---|---|---|
x+y |
x+y y.add_(x) torch.add(x,y) |
addition |
np.dot(x,y) np.matmul(x,y) |
torch.mm(x,y) x.mm(y) |
matrix multiplication |
x*y |
x*y |
element-wise multiplication |
np.max(x) |
torch.max(x) |
|
np.argmax(x) |
torch.argmax(x) |
|
x**2 |
x**2 |
Element-wise powers |
| Numpy | PyTorch | Notes |
|---|---|---|
x.T np.transpose(x) |
torch.transpose(x, 0, 1) torch.transpose(x, 1, 0) |
transpose |
a = a.reshape(-1, 2) |
a = a.view(-1,2) |
reshape array to have two columns and however as many rows |
np.concatenate([a, b]) |
torch.cat([a,b]) |
concatenate list of arrays/tensors |
A lot more summarized here but need to re-organize: https://jhui.github.io/2018/02/09/PyTorch-Basic-operations/