lartpang/PyTorchTricks
A collection of practical PyTorch optimization tricks covering model design, inference acceleration, quantization, and memory optimization.
Not currently ranked — collecting fresh signals.
star history
This repository documents various techniques for improving PyTorch model training and inference. It covers topics like model design patterns, convolution-BN fusion for faster inference, model compression and quantization methods, memory-efficient training strategies such as mixed precision with apex, and accelerated data loading using tar format with IterableDataset. Content is organized by category with links to reference papers and resources.
Frequently asked
- What is lartpang/PyTorchTricks?
- A collection of practical PyTorch optimization tricks covering model design, inference acceleration, quantization, and memory optimization.
- Is PyTorchTricks open source?
- Yes — lartpang/PyTorchTricks is an open-source project tracked on heatdrop.
- How popular is PyTorchTricks?
- lartpang/PyTorchTricks has 1.2k stars on GitHub.
- Where can I find PyTorchTricks?
- lartpang/PyTorchTricks is on GitHub at https://github.com/lartpang/PyTorchTricks.