DiffusionGemma
2 storiesEvery DiffusionGemma story collected by The AI Daily, 2 so far, newest first, refreshed hourly.
Related topics
A researcher shared DiffusionGemma, explaining how a diffusion language model generates text in parallel and implementing it from scratch in PyTorch, exploring non-autoregressive text generation.
OpenUI released OUI-1, fine-tuned on DiffusionGemma using a custom DSL OpenUI‑Lang instead of HTML/Markdown/React. Fine‑tuning can save context compared with prompting a general LLM to emit OpenUI‑Lang, but may bias the model toward that format. DiffusionGemma isn't supported by llama.cpp (so Ollama isn't an option); weights are on Hugging Face, and users wonder how to run it locally on a consumer GPU like an RTX 5090.
That is everything