CODEWITHSUNDEEP
luatorch
Kevin Bullaughey
Kevin Bullaughey

Reputation: 2576

What are the most idiomatic ways to combine tensors in torch?

I'm confronted with concatenating three tensors together so that 3 px1 tensors become one 3px1 tensor.

The most succinct I could come up with was:

torch.Tensor{v2:totable(),v4:totable(),v6:totable()}:view(3*p,1)

Are there ways to do this without converting to tables and back to tensors? It seems like there should be a generic way of concatenating tensors along some specified dimension assuming they have compatible shapes.

I can see how it would be possible to write such a function, does one not exist?

Upvotes: 10

Views: 5833

Answers (1)

smhx
smhx

Reputation: 2266

a = torch.randn(3,1)
b = torch.randn(3,1)
c = torch.randn(3,1)

d = torch.cat(a,b,1):cat(c,1)

print(d)

Upvotes: 17

Related Questions