Efficient ML

This repository also contains my learning on advanced efficient ML.

Papers to read

Deep compression

EIE: Efficient Inference Engine on Compressed Deep Neural Network

Learning both Weights and Connections for Efficient Neural Networks

Papers I'm reading

A White Paper on Neural Network Quantization

A Perceptually-Motivated Approach for Low-Complexity, Real-Time Enhancement of Fullband Speech

Programming advice for me :

Use sanity checks : Python : assert isinstance(prune_ratio, (float, list)), C++ : static_assert(std::is_floating_point<T>::value, "T must be floating point");, take a look at def channel_prune(model: nn.Module,prune_ratio: Union[List, float]) -> nn.Module in lab1.ipynb
deepcopy ;)# GPU-MODE