swamidass

I am working out some similar functionality in the anytensor library. Would implementing the jraph approach in anytensor solve your problem? https://github.com/swamidass/anytensor

Masking and padding only come up I jraph when you are trying to keep the batches a consistent size. The basic strategy is using segment-id vectors to do segment operations....

These are the dimensions of the source image, ``` {"@context":"http://iiif.io/api/image/3/context.json","id":"https://iiif.trustedkidney.com/iiif/3/test/test1.svs","type":"ImageService3","protocol":"http://iiif.io/api/image","profile":"level2","width":12600,"height":21604,"sizes":[{"width":12600,"height":21604},{"width":6300,"height":10802},{"width":3150,"height":5401},{"width":1575,"height":2700},{"width":787,"height":1350},{"width":393,"height":675},{"width":196,"height":337},{"width":98,"height":168}],"tiles":[{"width":512,"height":512,"scaleFactors":[1,2,4,8,16,32,64,128]}],"extraFormats":["jpeg","tif","tiff","webp"],"extraQualities":["color","gray","bitonal"],"extraFeatures":["canonicalLinkHeader","mirroring","profileLinkHeader","rotationArbitrary","sizeByDistortedWh","sizeByForcedWh","sizeByWhListed","sizeUpscaling"]} ``` More information here too. Seems that regenerating the pyramid (not ideal) using vips resolves this problem, but creates another....

To your other question, currently the links are live, so you can test it yourself if you want. If it is a rounding issue, it will probably require a more...

And here is some more information, I verified that my OSD implementation works just fine on other people's IIIF servers (not sure their implementations). Seems that the errors in OSC...

I can share the misbehaving test1.svs with you here, but I'd rather not do so publicly. May I message you privately?

let me note, as well, that there is already a segmented sinkhorn solver: https://ott-jax.readthedocs.io/en/latest/_autosummary/ott.tools.segment_sinkhorn.segment_sinkhorn.html?highlight=segment#ott.tools.segment_sinkhorn.segment_sinkhorn So possibly this is a matter of creating a front end API function that makes use...

So, the pipeline that works fine in Keras 2: - dataset is provided in ragged tensors, which allows fit shuffle and other sorts of alignment issue to work. - First...

Thanks for confirming the bug. Perhaps we'll get more help if we can create a mimimal project that reproduces the error? How hard would that be to do?

Thanks! Is there an alternate to this package?