Requesting Open dLLM support for llama.cpp

[kannony]

I'd like to request llama.cpp support for Open dLLM. 🙏

Open dLLM is a fully open source diffusion model that offers efficiency and speed. It's similar to LLaDA and Dream, although it seems to have some efficiency advantages. The 0.5 d-coder (derived from Open dLLM) is only 0.5B parameters and yet it manages to run neck and neck with 7B and 8B models. I think it could be a good LLM for mobile platforms or edge computing.

It's also worth noting that It also includes pretraining, evaluation, inference, and checkpoints, making it the most open diffusion model (according to its dev).

Open d-coder is serived from Open dLLM:

https://github.com/pengzhangzhi/Open-dLLM

There's a Reddit post with video of the model in action here:

https://www.reddit.com/r/learnmachinelearning/comments/1otwo7f/opendllm_open_diffusion_large_language_models/