pytorch VS tensorflow

[vmtmxmf5]

Pytorch 명령어

tf	torch
tf.Variable()	torch.FloatTensor() torch.LongTensor()
axis	dim
a.shape	a.shape a.size()
a.reshape()	a.view() a.reshape()
a[tf.newaxis, ...]	a[None, ...] a.unsqueeze(dim)
tf.concat([x, y], axis=)	torch.cat([x, y], dim=)
tf.zeros_like()	torch.zeros_like()

덮어쓰기 기능

a.mul_(2)

index_fill

tmp.indexfill(dim, index=torch.tensor([0,2]), val)

병합

1. torch.cat할 때 dim=0의 사이즈만 자유롭고 나머지는 같아야 함

2. cat vs stack

torch.cat([a.unsqueeze(0), b.unsqueeze(0), c.unsqueeze(0)])
torch.stack([a,b,c])

random uniform dist.

torch.empty(4, 2).uniform_(0,1)