A node type was not returned. This is either because a gradient was not defined, or the input is independent of the output about torch-autograd

[alexbw]

Assignment via __newindex is not allowed in autograd functions:

So:

 join_mixture_result[{{},{i}}] = mixture_result

is illegal. We're thinking on ways of providing better error messages for this.

[jarmstrong2]

What would you recommend as an alternative to assigning values to an already defined tensor? I did try torch.cat as well but without success.

[fmassa]

@jarmstrong2 maybe you can use select/narrow + copy ?

[alexbw]

Why did torch.cat fail?

On Mon, Jan 4, 2016 at 5:16 PM, John Armstrong [email protected]
wrote:

What would you recommend as an alternative to assigning values to an
already defined tensor? I did try torch.cat as well but without success.

—
Reply to this email directly or view it on GitHub
#74 (comment)
.

[jarmstrong2]

it gave the same error, so it's a bit of a mystery to me also.