An optimized Softmax approximation framework for NPU-accelerated LLM inference. It features a non-uniform segmentation strategy and variable-degree polynomial approximation optimized via Particle Swarm Optimization (PSO).

edge-ai llm-inference npu-acceleration qualcomm-qnn softmax-approximation
1 Open Issue Need Help Last updated: Jun 30, 2026

Open Issues Need Help

View All on GitHub
enhancement help wanted awaiting review

An optimized Softmax approximation framework for NPU-accelerated LLM inference. It features a non-uniform segmentation strategy and variable-degree polynomial approximation optimized via Particle Swarm Optimization (PSO).

#edge-ai#llm-inference#npu-acceleration#qualcomm-qnn#softmax-approximation