sparsezoo 0.1.1

Neural network model repository for highly sparse and sparse-quantized models with matching sparsification recipes

SparseZoo

Neural network model repository for highly sparse and sparse-quantized models with matching sparsification recipes

Overview

SparseZoo is a constantly-growing repository of highly sparse and sparse-quantized models with matching sparsification recipes for neural networks. It simplifies and accelerates your time-to-value in building performant deep learning models with a collection of inference-optimized models and recipes to prototype from.

Available via API and hosted in the cloud, the SparseZoo contains both baseline and models optimized to different degrees of inference performance vs. baseline loss recovery. Recipe-driven approaches built around sparsification algorithms allow you to take the models as given, transfer-learn from the models onto private datasets, or transfer the recipes to your architectures.

This repository contains the Python API code to handle the connection and authentication to the cloud.

Sparsification

Sparsification is the process of taking a trained deep learning model and removing redundant information from the overprecise and over-parameterized network resulting in a faster and smaller model. Techniques for sparsification are all encompassing including everything from inducing sparsity using pruning and quantization to enabling naturally occurring sparsity using activation sparsity or winograd/FFT. When implemented correctly, these techniques result in significantly more performant and smaller models with limited to no effect on the baseline metrics. For example, pruning plus quantization can give over 7x improvements in performance while recovering to nearly the same baseline accuracy.

The Deep Sparse product suite builds on top of sparsification enabling you to easily apply the techniques to your datasets and models using recipe-driven approaches. Recipes encode the directions for how to sparsify a model into a simple, easily editable format.

• Download a sparsification recipe and sparsified model from the SparseZoo.
• Alternatively, create a recipe for your model using Sparsify.
• Apply your recipe with only a few lines of code using SparseML.
• Finally, for GPU-level performance on CPUs, deploy your sparse-quantized model with the DeepSparse Engine.

Full Deep Sparse product flow:

Quick Tour

Each model in the SparseZoo has a specific stub that identifies it. The stubs are made up of the following structure:

DOMAIN/SUB_DOMAIN/ARCHITECTURE{-SUB_ARCHITECTURE}/FRAMEWORK/REPO/DATASET{-TRAINING_SCHEME}/SPARSE_NAME-SPARSE_CATEGORY-{SPARSE_TARGET}

The properties within each model stub are defined as the following:

Model Property	Definition	Examples
DOMAIN	The type of solution the model is architected and trained for	cv, nlp
SUB_DOMAIN	The sub type of solution the model is architected and trained for	classification, segmentation
ARCHITECTURE	The name of the guiding setup for the network's graph	resnet_v1, mobilenet_v1
SUB_ARCHITECTURE	(optional) The scaled version of the architecture such as width or depth	50, 101, 152
FRAMEWORK	The machine learning framework the model was defined and trained in	pytorch, tensorflow_v1
REPO	The model repository the model and baseline weights originated from	sparseml, torchvision
DATASET	The dataset the model was trained on	imagenet, cifar10
TRAINING_SCHEME	(optional) A description on how the model was trained	augmented, lower_lr
SPARSE_NAME	An overview of what was done to sparsify the model	base, pruned, quant (quantized), pruned_quant, arch (architecture modified)
SPARSE_CATEGORY	Descriptor on the degree to which the model is sparsified as compared with the baseline metric	none, conservative (100% baseline), moderate (>= 99% baseline), aggressive (< 99%)
SPARSE_TARGET	(optional) Descriptor for the target environment the model was sparsified for	disk, edge, deepsparse, gpu

The contents of each model are made up of the following:

• model.md: The model card containing metadata, descriptions, and information for the model.
• model.onnx: The ONNX representation of the model's graph.
• model.onnx.tar.gz: A compressed format for the ONNX file. Currently ONNX does not support sparse tensors and quantized sparse tensors well for compression.
• [FRAMEWORK]/model.[EXTENSION]: The native ML framework file(s) for the model in which it was originally trained. Such as PyTorch, Keras, TensorFlow V1
• recipes/original.[md|yaml]: The original sparsification recipe used to create the model.
• recipes/[NAME].[md|yaml]: Additional sparsification recipes that can be used with the model such as transfer learning.
• sample-originals: The original sample data without any preprocessing for use with the model.
• sample-inputs: The sample data after pre processing for use with the model.
• sample-outputs: The outputs after running the sample inputs through the model.
• sample-labels: The labels that classify the sample inputs.

Python APIS

The Python APIs respect this format enabling you to search and download models. Some code examples are given below.

Searching the Zoo

from sparsezoo import Zoo

models = Zoo.search_models(domain="cv", sub_domain="classification")
print(models)

Common Models

from sparsezoo.models.classification import resnet_50

model = resnet_50()
model.download()

print(model.onnx_file.downloaded_path())

Searching Optimized Versions

from sparsezoo import Zoo
from sparsezoo.models.classification import resnet_50

search_model = resnet_50()
optimized_models = Zoo.search_optimized_models(search_model)

print(optimized_models)

Console Scripts

In addition to the Python APIs, a console script entry point is installed with the package sparsezoo. This enables easy interaction straight from your console/terminal. Note, for some environments the console scripts cannot install properly. If this happens for your system and the sparsezoo command is not available, https://github.com/neuralmagic/sparsezoo/blob/main/scripts/sparsezoo.py may be used in its place.

sparsezoo -h

Searching

Search command help

sparsezoo search -h

Searching for all classification models in the computer vision domain

sparsezoo search --domain cv --sub-domain classification \
    --architecture resnet_v1 --sub-architecture 50

Searching for all ResNet-50 models

sparsezoo search --domain cv --sub-domain classification

Downloading

Download command help

sparsezoo download -h

Download ResNet-50 Model

sparsezoo download --domain cv --sub-domain classification \
    --architecture resnet_v1 --sub-architecture 50 \
    --framework pytorch --repo sparseml --dataset imagenet \
    --optim-name base --optim-category none

Download pruned and quantized ResNet-50 Model

sparsezoo download --domain cv --sub-domain classification \
    --architecture resnet_v1 --sub-architecture 50 \
    --framework pytorch --repo sparseml \
    --dataset imagenet --training-scheme augmented \
    --optim-name pruned_quant --optim-category aggressive

For a more in-depth read, check out SparseZoo documentation.

Installation

This repository is tested on Python 3.6+, and Linux/Debian systems. It is recommended to install in a virtual environment to keep your system in order.

Install with pip using:

pip install sparsezoo

Then if you would like to explore any of the scripts or notebooks clone the repository and install any additional dependencies as required.

Available Models and Recipes

A number of pre-trained baseline and sparsified models across domains and sub domains are available and constantly being added. For an up to date list, please consult the available models listing.

Resources and Learning More

• SparseZoo Documentation
• SparseML Documentation
• Sparsify Documentation
• DeepSparse Documentation
• Neural Magic Blog, Resources, Website

Contributing

We appreciate contributions to the code, examples, and documentation as well as bug reports and feature requests! Learn how here.

Join the Community

For user help or questions about SparseZoo, use our GitHub Discussions. Everyone is welcome!

You can get the latest news, webinar and event invites, research papers, and other ML Performance tidbits by subscribing to the Neural Magic community.

For more general questions about Neural Magic, please email us at [e-mail] or fill out this form.

License

The project is licensed under the Apache License Version 2.0.

Release History

Official builds are hosted on PyPi

• stable: sparsezoo
• nightly (dev): sparsezoo-nightly

Additionally, more information can be found via GitHub Releases.