README.md

Keras Frontend

Installing Dependencies

Importing Conda Environment:

conda env create -f keras_environment.yml --name ${KERAS_ENV_NAME}

Note: pip version MUST be > 19.3

Activating Conda Environment:

conda activate ${KERAS_ENV_NAME}

Building and Installing Frontend:

python setup.py build

python setup.py install

Running Benchmaks

Benchmarks under ./src/

List of benchmarks and the expected accuracies:

Benchmark	Accuracy
AlexNet-CIFAR10	79.16
AlexNet2-CIFAR10	85.10
AlexNet-ImageNet	56.23
LeNet-MNIST	99.11
MobileNet-CIFAR10	82.40
ResNet18-CIFAR10	89.52
ResNet50-ImageNet	74.50
VGG16-CIFAR10	89.42
VGG16-CIFAR100	66.20
VGG16-ImageNet	72.50

Activate conda environment (above) before running benchmarks

Synopsys

python src/${BENCH_NAME}.py  [hpvm_reload|keras_reload]  [frontend] 

Parameters:

hpvm_reload : Reloads HPVM weights (format used in model_params found here: [ADD link to Google Drive]) from directory specified in Benchmark constructor.

keras_reload: Reloads weights in Keras .h5 file format

frontend: Invokes the HPVM frontend and dumps weights in directory specified in constructor

Building New Benchmarks

All benchmarks inherit from the commom parent Benchmark class. Each benchmark overrides virtual functions for building the model, training, inference, and data preprocessing.

def buildModel(self): returns a keras model

def data_preprocess(self): returns X_train, y_train, X_test, y_test, X_tuner, and y_tuner data — in that order; this data will be directly used later for training and inference

def trainModel(self, model, X_train, y_train, X_test, y_test): returns a trained keras model