DaG: Cross-Modal-Guided Token Compression Training-free token compression for long-context AV-LLM inference. ActTaylor: 结构化剪枝研究 Activation-aware structured pruning for efficient LLM deployment. ESP32-S3 + Qwen: Build Your Own Smart AI Device (Demo Video) Build a lightweight voice-interactive AI device with ESP32-S3 and Qwen. LiCO AI 平台 CV/NLP 训练、优化、部署的一体化平台工程实践。 Framing 视频分析平台 视频分帧、关键帧预测与后端服务集成。
