Show HN: An automated news site Distribute Software

Category: ai-ml

Tags: inference-framework, llm-serving, gpu-acceleration, deepseek, high-performance

Score: 7.8/10 (Innovation: 7, Technical: 9, Documentation: 7, Utility: 8)

DeepInfer is DeepSeek's open-source offline inference framework for their V4.1-Flash model, featuring optimized kernels, advanced parallelism (Attention DP, MoE EP+EPLB), asynchronous scheduling, deterministic sampling, and structured outputs. It targets high-performance production deployment on NVIDIA SM100 GPUs and Huawei Ascend NPUs, with impressive technical depth though incomplete for full service (no KV cache or disaggregation). The HN submission itself appears mismatched (title/link about an automated news site), suggesting poor presentation.

Target audience: ML engineers, inference infrastructure engineers, backend devs working on LLM deployment at scale

Repository: https://distribute.software/ · Python · Apache-2.0 · 20224 stars

View on Hacker News