Deep Interest Evolution Network for Click-Through Rate Prediction with OpenVINO backend

This repo introduce how to run Deep Interest Evoluation Network (DIEN) with OpenVINO backend.

Please refer to orignal repo for details: https://github.com/alibaba/ai-matrix/tree/master/macro_benchmark/DIEN_TF2

prepare data

method 1

You can get the data from amazon website and process it using the script

sh prepare_data.sh

method 2 (recommended)

Because getting and processing the data is time consuming，so we had processed it and upload it for you. You can unzip it to use directly.

sh prepare_dataset.sh

When you see the files below, you can do the next work. 
- cat_voc.pkl 
- mid_voc.pkl 
- uid_voc.pkl 
- local_train_splitByUser 
- local_test_splitByUser 
- reviews-info
- item-info

Setup Python Environment

pip install openvino openvino-dev[tensorflow]

Convert original Tensorflow model to OpenVINO FP32 IR

mo --input_meta_graph dnn_best_model_trained/ckpt_noshuffDIEN3.meta \
   --input "Inputs/mid_his_batch_ph[-1,-1],Inputs/cat_his_batch_ph[-1,-1],Inputs/uid_batch_ph[-1],Inputs/mid_batch_ph[-1],Inputs/cat_batch_ph[-1],Inputs/mask[-1,-1],Inputs/seq_len_ph[-1]" \
   --output dien/fcn/Softmax --model_name DIEN -o openvino/FP32 --compress_to_fp16=False

Run the Benchmark

Run the Benchmark with Tensorflow backend

./infer.sh tensorflow

Run the Benchmark with OpenVINO backend using FP32 inference precision

./infer.sh openvino f32

Run the Benchmark with OpenVINO backend using BF16 inference precision

./infer.sh openvino bf16

Please note, Xeon native supports BF16 infer precision since 4th Generation Intel® Xeon® Scalable Processors. Run BF16 on legacy Xeon platform may lead to performance degradation.

sammysun0711 / dien_openvino_demo