Pedro Cuenca

> Can you call the function `to_rgb` (or similar) and make it a noop when the format is already RGB? I didn't want to call it `to_rgb` because then it...

> Also can you check if it affects the `prepare_dataset` notebook and the wandb logging (may need to use a `jax.device_get` there). Sure, good point. But this was branched from...

I addressed a couple of @borisdayma's comments, but it turns out we **can't** use `tfio.experimental.color.lab_to_rgb` in jitted code because there's an implicit conversion to `numpy` [here](https://github.com/tensorflow/io/blob/v0.24.0/tensorflow_io/python/experimental/color_ops.py#L477). I checked the color...

@patrickvonplaten You are right! I stashed my changes to apply them in a new branch and messed up somehow. I have just pushed the correct version, I hope. Sorry about...

## Current status ### ☑️ Easy bugs * Set offset=1 (#4) * Iterate for the total number of steps (#6). ### ☑️ Scheduler state bug I created a [stateless implementation](https://github.com/patil-suraj/stable-diffusion-jax/blob/0d0e6c2/stable_diffusion_jax/scheduling_pndm.py#L234)...

After the sinusoidal fix (#7), the differences after the inference loop are in the order of `~8e-4` for the slice I print in one of my test images. This is...

> I think we should not create flloat64 timesteps 👍 > I'd actually merge everything directly into `diffusers` and make a bigger release (0.3.0) saying we now also support JAX...

Hi @smileBeda! We haven't submitted an iPhone/iPadOS version yet, so you'd need to compile it yourself for now. The reason was that it's harder to test on a variety of...

We are going to submit one soon, but you can also build it yourself using this repo.

Hi @Acephalia, thanks for reaching out! We'll submit an update to the App Store soon!