Skip to content

Latest commit

 

History

History
70 lines (58 loc) · 2.54 KB

ResNeXt.md

File metadata and controls

70 lines (58 loc) · 2.54 KB

PyTorch ResNeXt (Image Classification)

This document describes evaluation of optimized checkpoints for ResNeXt

Environment Setup

Setup AI Model Efficiency Toolkit (AIMET)

Please install and setup AIMET before proceeding further. This model was tested with the torch_gpu variant of AIMET 1.24.

Additional Setup Dependencies

pip install torchvision==0.11.2 --no-deps

Obtain the Original Model for Comparison

  • Pytorch Torchvision hub instances of ResNeXt101_32x8d are used as reference FP32 models. These instances are optimized using AIMET to obtain quantized optimized checkpoints.

Experiment setup

git clone https://github.com/quic/aimet-model-zoo.git

export PYTHONPATH=$PYTHONPATH:<path to aimet_model_zoo_path>

Dataset

This evaluation was designed for the 2012 ImageNet Large Scale Visual Recognition Challenge (ILSVRC2012), which can be obtained from: http://www.image-net.org/
The dataset directory is expected to have 3 subdirectories: train, valid, and test (only the validation dataset is used for this evaluation). Each of the {train, valid, test} directories is then expected to have 1000 subdirectories, each containing the images from the 1000 classes present in the ILSVRC2012 dataset, such as in the example below:

  train/
  ├── n01440764
  │   ├── n01440764_10026.JPEG
  │   ├── n01440764_10027.JPEG
  │   ├── ......
  ├── ......
  val/
  ├── n01440764
  │   ├── ILSVRC2012_val_00000293.JPEG
  │   ├── ILSVRC2012_val_00002138.JPEG
  │   ├── ......
  ├── ......

Usage

To run evaluation with QuantSim in AIMET, use the following

python resnext_quanteval.py\
  --model-config <configuration to be tested> \
  --dataset-path <path to ImageNet dataset> \
  --use-cuda <whether to run on GPU or cpu>

Available model configurations are:

  • resnext101_w8a8

Quantization Configuration

The following configuration has been used for the above models for INT8 quantization:

  • Weight quantization: 8 bits, symmetric quantization
  • Bias parameters are not quantized
  • Activation quantization: 8 bits, asymmetric quantization
  • Model inputs are quantized
  • 2 images per class (2000 total images) from the calibration dataset were used for computing encodings
  • TF_enhanced was used as quantization scheme
  • Batch norm folding was applied to get the best W8A8 optimized checkpoint