savvaki

@spillerrec Are there any updates on this issue?

@spillerrec Do you perhaps have the tf code lying around that you used to generate the onnx models that you posted above?

I created an "equivalent" operation to `tf.image.extract_patches` using an identity kernel with a convolution. The code seemingly works for batches, as well as different patch sizes. In addition, multiple input...

> I'll check it out and report. Thanks for the pointer. I know this is an old issue but did you manage to sort this out? I have the same...

I can confirm that the resumed runs behave as @abaisero mentioned. Multiple runs with the same resume ID are stored in the wandb folder which all need to be synced...