Tag
vLLM announces native support for Google DeepMind's DiffusionGemma, a 26B discrete diffusion language model that generates 256-token blocks in parallel, enabling low-latency inference at 1200+ tok/s on a single H200.
The author details attempts to locally train a Qwen 3.6 27B autoregressive-to-diffusion model on an Nvidia 5090 GPU using qlora and modifications from open-dllm and d3LLM, facing VRAM constraints and hardware issues while exploring one-shot diffusion techniques.