SOTA low-bit LLM quantization (INT8/FP8/INT4/FP4/NF4) & sparsity; leading model compression techniques on TensorFlow, PyTorch, and ONNX Runtime

Must read research papers and links to tools and datasets that are related to using machine learning for compilers and systems optimisation

bpftune uses BPF to auto-tune Linux systems

Kernel Tuner

Machine Learning Framework for Operating Systems - Brings ML to Linux kernel

Stretching GPU performance for GEMMs and tensor contractions.

CLTune: An automatic OpenCL & CUDA kernel tuner

Autotune Module for Python "PyAutoTune"

Alchemy Cat —— 🔥Config System for SOTA

Phoebe

Benchmark scripts for TVM

ebpf profiler for jvm

Collective Knowledge crowd-tuning extension to let users crowdsource their experiments (using portable Collective Knowledge workflows) such as performance benchmarking, auto tuning and machine learning across diverse platforms with Linux, Windows, MacOS and Android provided by volunteers. Demo of DNN crowd-benchmarking and crowd-tuning:

K2vTune (A Workload-aware Configuration Tuning for RocksDB)

A Generic Distributed Auto-Tuning Infrastructure

A GPU benchmark suite for autotuners

Backoff uses an exponential backoff algorithm to backoff between retries with optional auto-tuning functionality.

Autotuner for Spark applications

This software package accompanies the paper "A Methodology for Comparing Auto-Tuning Optimization Algorithms" (https://doi.org/10.1016/j.future.2024.05.021), making the guidelines in the methodology easy to apply.

MATOG: CUDA Array Access Auto-Tuner