Thomas Germer

I implemented heap-based token merging for my simplified TinyLlama implementation. Feel free to copy what you need. Here is the function: https://github.com/99991/SimpleTinyLlama/blob/main/tokenizer.py#L70 This function might be easier to understand after...

> The backends support a separate hatch color and collections store the hatch color as `self._hatch_color`. Thank you for the hint! It has helped me to find the following workaround...

> I have worked on this issue. Please review it I got the following error with your latest commit at the time, but it seems that new commits are still...

I was able to successfully run your code with this branch. It produced a satisfactory hatched bar plot. The legend is correct as well. 

Perhaps your disk write speed is too low? How fast is it? You can measure it with this python script: ```python # Measure write speed by repeatedly writing 1 MB...

I don't get exactly the same apple either, but it is very close. I can not tell the difference by eye.  Settings: * Prompt: `an apple, high quality` *...

The encoder has been trained to produce latents which do not upset the base model. More details in section 3 of paper: https://arxiv.org/pdf/2402.17113.pdf Since your model is fine-tuned from the...

We are happy to know that this project is useful to you! We have just published a new release: * https://pypi.org/project/PyMatting/ * https://github.com/pymatting/pymatting/releases/tag/v1.1.10 Personally, I found that shared matting was...

Hello Tatiana, the transparency issue becomes apparent when looking at the resulting alpha matte (computed with `lkm_radius=30` in this example). As you can see, the alpha mask for the skin...

The trimap must consist of three parts: Foreground (white, color value 1.0), background (black, color value 0.0), and unknown region (gray, color value 0.5). Your trimap is incorrect. Instead of...