Nick Weir

@bruesba, Yeah, that's strange. It's particularly weird that it doesn't throw an error until the `dec5` layer of the model, which is the first up-convolution after the encoder. If I...

@bruesba did you happen to figure out what was going on here? I want to try to track this down before releasing 0.2.0.

OK no worries, thanks for the response!

Hi @andresfchamorro, Thanks for letting us know! Can you provide any more context on your use case? (Is it an open dataset by any chance? What did you change in...

Hey @andresfchamorro and @bruesba, I _might_ have figured out what's going on here. __Short answer:__ inference using XD_XD's model requires the x and y dimensions of your data to be...

OK, sounds like we're narrowing down. That was our experience on debugging also (error in inference, not train). I wonder if this would _always_ be the case during training, or...

@KPGeo thanks for this, very good to know. We can make those changes unless you want to put a PR in.

Thanks for this @vincentsarago. The intention wasn't necessarily to re-project - the source file should already be in that CRS (in effect the source file _is_ a temp file) -...

@vincentsarago, have you ever thought about putting a conda package up for `rio-tiler`? I'm working on getting a conda build for this project and am trying to nail down the...

@vincentsarago thanks for the super prompt reply - I'm happy to help out with that, I've put together packages for conda-forge before.