Skip to content
View original post on X: vLLMOfficial· 34/100AI score34/100

Inferact releases DSpark speculator for Muse Spark Open

AISummary

Inferact trained a DSpark speculator for Muse Spark Open, using block diffusion to propose up to seven draft tokens in one pass. The recipe supports CUDA and ROCm, and the draft model is published on Hugging Face.

Post on XView on X
vLLMVerified on X
@vllm_project

Part of a thread · earlier post

Inferact trained a DSpark speculator for Muse Spark Open.

Block diffusion proposes up to seven draft tokens in one pass. Adaptive verification uses confidence-based scheduling, while sliding-window attention keeps the drafting path lightweight at long context.

The recipe supports CUDA and ROCm. Draft model:
https://huggingface.co/Inferact/Muse-Spark-open-DSpark

3/6

Source: vLLM · x.comPublished · added here

Topics

Open-source ecosystemDeployment & engineeringData & trainingProduct updates