Optimize a ResNet50* Bfloat16 Model Package with PyTorch*
Published: 12/09/2020
Download Command
wget https://storage.googleapis.com/intel-optimized-tensorflow/models/v2_3_0/pytorch-resnet50-bfloat16-inference.tar.gz
Description
This document has instructions for running ResNet50* bfloat16 inference using Intel® Extension for PyTorch*.
Datasets
The ImageNet validation dataset is used when testing accuracy. The inference scripts use synthetic data, so no dataset is needed.
Download and extract the ImageNet2012 dataset from http://www.image-net.org/, then move validation images to labeled subfolders, using the valprep.sh shell script
The accuracy script looks for a folder named val
, so after running the data prep script, your folder structure should look something like this:
imagenet
└── val
├── ILSVRC2012_img_val.tar
├── n01440764
│ ├── ILSVRC2012_val_00000293.JPEG
│ ├── ILSVRC2012_val_00002138.JPEG
│ ├── ILSVRC2012_val_00003014.JPEG
│ ├── ILSVRC2012_val_00006697.JPEG
│ └── ...
└── ...
The folder that contains the val
directory should be set as the DATASET_DIR
when running accuracy (for example: export DATASET_DIR=/home//imagenet
).
Quick Start Scripts
Script name | Description |
---|---|
bf16_online_inference |
Runs online inference using synthetic data (batch_size=1). |
bf16_batch_inference |
Runs batch inference using synthetic data (batch_size=128). |
bf16_accuracy |
Measures the model accuracy (batch_size=128). |
Bare Metal
To run on bare metal, the following prerequisites must be installed in your environment:
- Python* 3
- Intel® Extension for PyTorch*
- Torchvision v0.6.1
- Numactl
Download and untar the model package and then run a quick start script.
# Optional: to run accuracy script
export DATASET_DIR=<path to the preprocessed imagenet dataset>
# Download and extract the model package
wget https://storage.googleapis.com/intel-optimized-tensorflow/models/v2_3_0/pytorch-resnet50-bfloat16-inference.tar.gz
tar -xzf pytorch-resnet50-bfloat16-inference.tar.gz
cd pytorch-resnet50-bfloat16-inference
bash quickstart/<script name>.sh
Documentation and Sources
Get Started
Main GitHub*
Readme
Release Notes
Get Started Guide
Code Sources
Report Issue
License Agreement
LEGAL NOTICE: By accessing, downloading or using this software and any required dependent software (the “Software Package”), you agree to the terms and conditions of the software license agreements for the Software Package, which may also include notices, disclaimers, or license terms for third party software included with the Software Package. Please refer to the license file for additional details.
Related Containers and Solutions
ResNet50* BFloat16 Inference TensorFlow* Container
Product and Performance Information
Performance varies by use, configuration and other factors. Learn more at www.Intel.com/PerformanceIndex.