DeepSeek 开源生产级推理引擎 DeepInfer，8 小时收获 6.3K 星
DeepSeek 把自家用在生产环境的推理引擎开源了，还兼容华为昇腾 950 芯片，看推理优化的人可以拆开研究。
DeepSeek 以 Apache 2.0 协议开源了支撑其生产服务的推理引擎 DeepInfer，8 小时内获得超过 6.3K 星。目前该引擎只支持 DeepSeek-V4.1-Flash 模型，且仅覆盖离线推理，不含 KV cache、prefill 与 decode 分离及负载均衡等服务化组件。核心是速度层，包括 DSpark——可提前起草 5 个 token 并用置信度估计调度验证，以及 bounded replay——只重算覆盖滑动注意力窗口的 140 token 后缀来压缩 prefill。配套的 deepseek-recipe 库能把 Chat Completions、Responses 和 Messages 请求转换成与官方 API 行为一致的 DeepSeek V4 和 V4.1 提示词。
DeepSeek just open-sourced DeepInfer, the inference engine behind its production services, under the Apache 2.0 license.
and it got 6.3K+ stars in 8 hours
The release runs only DeepSeek-V4.1-Flash, on NVIDIA SM100 GPUs or Huawei Ascend 950 chips, and covers offline inference rather than a complete serving stack.
It omits the parts that turn an engine into a service: a KV cache system, prefill and decode disaggregation, and load balancing across replicas.
What ships is the speed layer, including DSpark, which drafts up to 5 tokens ahead and uses confidence estimates to schedule their verification.
Bounded replay trims prefill by recomputing only a 140-token suffix sized to cover the sliding attention window.
A companion library, deepseek-recipe, will turn Chat Completions, Responses and Messages requests into DeepSeek V4 and V4.1 prompts that match the official API’s behaviour.
