-
Hi,
Thank you for your work! I'm currently facing a challenge where I got a fixed transport cost matrix with shape n*n, while having many different marginal distributions $a$ and $b$ with shape [B,…
-
Essentially whenever I calculate the sinkhorn loss, pytorch's internal gradient tracking turns back on.
I believe this is to do with line 280 in geomloss/sinkhorn_divergence.py:
`torch.autograd.se…
-
Thank you for your work! It gives me a more clear understanding of OT.
But I am new to the Sinkhorn iteration algorithm, and as I understand it, the iteration formula is as follows:
 the authors talk about using [Sinkhorn distances](https://arxiv.org/pdf/1306.0895.pdf) to reduce complex…
-
Hi, I am trying to use `sinkhorn` matcher for batch sizes >8 (per device) as in the code. It (understandably) fails at #L62 since `bs * dim` exceeds sampling range available `10 * dim`.
```shell
V…
-
-
Recently came across a use case for this. Might be something worth implementing in the future.
https://arxiv.org/pdf/1910.12958.pdf
-
Hi,
First, thanks a lot for a great package that is very straightforward to use!
I am confused by some results I am getting where the Sinkhorn distance (with default parameters) is negative. Fro…