ASR using Diffusion Language Model

Optimized SM121 vLLM container and benchmark report for nvidia/diffusiongemma-26B-A4B-it-NVFP4

FastMCP fleet MCP server for diffusion LMs (dLLM). DiffusionGemma on Goliath RTX 4090 — batch inference, HLE-shaped reasoning, ~200–400 tok/s. Doc phase; llama-diffusion-cli sidecar next. Complements local-llm-mcp.

DiffusionGemma node pack for ComfyUI, built on MCP backbone for fully-agentic consumption — discrete diffusion text generation with per-step canvas snapshots, commit heatmaps, and structured trace data. Watch meaning crystallize out of noise.

Native WinUI 3 control panel for running local llama.cpp and DiffusionGemma backends with model management, Hugging Face downloads, runtime tuning, logs, and resource monitoring.

Matrix-style logit conditioning for DiffusionGemma's llama.cpp denoiser

Local DiffusionGemma coding agent for Windows, WSL2, and 16 GB NVIDIA GPUs

llama.cpp fork with experimental DiffusionGemma full-GPU and CUDA fusion support

Docker-Compose template to self-host Google DiffusionGemma 26B on an NVIDIA GPU host via llama.cpp