Timo Sämann

ENet. ENet: A Deep Neural Network Architecture for Real-Time Semantic Segmentation

caffe-segnet-cudnn5. This repository was a fork of BVLC/caffe and includes the upsample, bn, dense_image_data and softmax_with_loss (with class weighting) layers of caffe-segnet (https://github.com/alexgkendall/caffe-segnet) to run SegNet with cuDNN version 5.

caffe-enet. C++

cuda_mlp. Run a neural network without PyTorch, TensorRT, or Python — just pure CUDA.

mamba-llm-fused-kernel. This repository provides a minimal, self-contained example showing how LLM-based compilation (LLM-assisted kernel generation) can restructure computation in ways traditional compilers typically cannot.

distributed-training-overview.

precision-search-kernelbench. Python

ml-inference-optimization-taxonomy. ml-inference-optimization-taxonomy

mamba-scan-optimizations. This repository presents an LLM-guided optimization of Mamba’s selective scan, transforming a research-grade PyTorch reference implementation into progressively more optimized variants