Popular GitHub Repositories for Flops

Image Restoration Toolbox (PyTorch). Training and testing codes for DPIR, USRNet, DnCNN, FFDNet, SRMD, DPSR, BSRGAN, SwinIR

Flops counter for convolutional networks in pytorch framework

The calflops is designed to calculate FLOPs、MACs and Parameters in all various neural networks, such as Linear、 CNN、 RNN、 GCN、Transformer(Bert、LlaMA etc Large Language Model)

Pax is a Jax-based machine learning framework for training large scale models. Pax allows for advanced and fully configurable experimentation and parallelization, and has demonstrated industry leading model flop utilization rates.

AI and Memory Wall

llm theoretical performance analysis tools and support params, flops, memory and latency analysis.

Count number of parameters / MACs / FLOPS for ONNX models.

Your one stop CLI for ONNX model analysis.

Calculating FLOPs of Pre-trained Models in NLP

WeightUsageAnalyzer is a lightweight toolkit for analyzing the effective use of weights in neural networks. It provides quantitative metrics (entropy, coverage, FLOPs) and visualizations to identify redundant parameters and guide model simplification, compatible with both PyTorch and TensorFlow.

WeightUsageAnalyzer is a lightweight toolkit for analyzing the effective use of weights in neural networks. It provides quantitative metrics (entropy, coverage, FLOPs) and visualizations to identify redundant parameters and guide model simplification, compatible with both PyTorch and TensorFlow.

Neural Architecture Search (NAS) on a UNet neural network (SR3) to optimize for minimal FLOPs and inference latency; while working in a denoising diffusion generative framework (DDPM)

A general and accurate RAM / MACs / FLOPs profiler for ONNX models (Currently work-in-progress)

RSRC Calculator is a practical tool designed to evaluate the efficiency of AI models in the post-scaling era: Recursive Self-Referential Compression (RSRC), this tool computes training efficiency metrics by analyzing factors such as training FLOPs, energy consumption, and model architecture details.

This project demonstrates the impact of model design choices on both energy consumption and economic cost. It analyzes the weight importance within a neural network, estimates the total FLOPs required for inference, and explores how quantization and pruning affect resource efficiency.