imagenet

Reproducible ImageNet training with Ignite

In this example, we provide script and tools to perform reproducible experiments on training neural networks on ImageNet dataset.

Features:

• Distributed training with native automatic mixed precision
• Experiments tracking with ClearML

Model	Training Top-1 Accuracy	Training Top-5 Accuracy	Test Top-1 Accuracy	Test Top-5 Accuracy
ResNet-50	78%	92%	77%	94%

Experiment	Model	Training Top-1 Accuracy	Training Top-5 Accuracy	Test Top-1 Accuracy	Test Top-5 Accuracy	ClearML Link
configs/???.py

Setup

pip install -r requirements.txt

Docker

For docker users, you can use the following images to run the example:

docker pull pytorchignite/vision:latest

and install other requirements as suggested above

Usage

Please, export the DATASET_PATH environment variable for the ImageNet dataset.

export DATASET_PATH=/path/to/imagenet
# e.g. export DATASET_PATH=/data/ where "train", "val", "meta.bin" are located

Training

Single GPU

• Adjust batch size for your GPU type in the configuration file: configs/baseline_resnet50.py or configs/baseline_resnet50.py

Run the following command:

CUDA_VISIBLE_DEVICES=0 python -u main.py training configs/baseline_resnet50.py

Multiple GPUs

• Adjust total batch size for your GPUs in the configuration file: configs/baseline_resnet50.py or configs/baseline_resnet50.py

OMP_NUM_THREADS=1 torchrun --nproc_per_node=2 main.py training configs/baseline_resnet50.py

Acknowledgements

Trainings were done using credits provided by trainml.ai platform.