README.md

DIEN training

Description

This document has instructions for running DIEN training using Intel-optimized TensorFlow.

Datasets

Use prepare_data.sh to get a subset of the Amazon book reviews data and process it. Or download and extract the preprocessed data files directly:

wget https://zenodo.org/record/3463683/files/data.tar.gz
wget https://zenodo.org/record/3463683/files/data1.tar.gz
wget https://zenodo.org/record/3463683/files/data2.tar.gz

tar -jxvf data.tar.gz
mv data/* .
tar -jxvf data1.tar.gz
mv data1/* .
tar -jxvf data2.tar.gz
mv data2/* .

Set the DATASET_DIR to point to the directory with the dataset files when running DIEN.

Quick Start Scripts

Script name	Description
training.sh	Runs training with a batch size of 128 for the specified precisions fp32, bfloat16 and bfloat32.

Run the model

• Follow the instructions to setup your bare metal environment on Linux systems. Ensure that you have a clone of the AI Reference Models Github repository.

  git clone https://github.com/IntelAI/models.git
  

• Install intel-tensorflow>=2.5.0 on your system.

Run on Linux

# cd to your AI Reference Models directory
cd models

# Set env vars
export DATASET_DIR=<path to the DIEN dataset>
export PRECISION=<set precision to fp32 or bfloat16 or bfloat32>
export OUTPUT_DIR=<path to the directory where log files will be written>

# Run a quickstart script
./quickstart/recommendation/tensorflow/dien/training/cpu/training.sh

Additional Resources

• To run more advanced use cases, see the instructions here for calling the launch_benchmark.py script directly.
• To run the model using docker, please see the DevCatalog workload container