Thomas Sajot

New version working much better but there is still some lag due to the canvas.put_image_data function (I believe). The new version supports rectangle images. The zoom box supports mouse drag....

I was thinking about it. One way to do it is to modify the `AbstractAnnotationCanvas` and wrap the methods `on_click`, `on_drag` and `on_release` to modify the (x, y) coordinates of...

Good catch. I am not so happy about the approach yet. Modifying the Abstract class would make it have more functionality that only gets used in the zoomed setting. But...

A new [Pull Request](https://github.com/janfreyberg/ipyannotations/pull/13) has been created. There is a notebook with a working example for Poly, box and Point annotations. The zoom is functional, but the translation is not...

The shapes are custom classes. You may be interested by the [Shapely](https://shapely.readthedocs.io/en/latest/manual.html) library which handles geometries and a wide variety of functionality on top of it. A useful one is...

It would also include the example `3rdparty` directory and other structures mentioned in the documentation.

Either that or create multiple example-python repos. `example-python-minimal`, `example-python-monorepo`, `example-python-polyglot`, ...

This seems like a good way to go.

Thank you for your reply. As you mentioned, `register_buffer` might be the correct way to go. The goal of the cache is to append historical elements alongside the current tensor.

Reading further the [Pytorch doc](https://pytorch.org/docs/stable/amp.html#autocasting): > [autocast](https://pytorch.org/docs/stable/amp.html#torch.autocast) should wrap only the forward pass(es) of your network, including the loss computation(s). Backward passes under autocast are not recommended. Backward ops run...