Skip to content

Latest commit

 

History

History
97 lines (74 loc) · 9.25 KB

README.md

File metadata and controls

97 lines (74 loc) · 9.25 KB

Image Classification Sample

This sample demonstrates a DL model compression in case of an image-classification problem. The sample consists of basic steps such as DL model initialization, dataset preparation, training loop over epochs, training and validation steps. The sample receives a configuration file where the training schedule, hyper-parameters, and compression settings are defined.

Features

  • Torchvision models (ResNets, VGG, Inception, etc.) and datasets (ImageNet, CIFAR 10, CIFAR 100) support
  • Custom models support
  • Configuration file examples for sparsity, quantization, and quantization with sparsity
  • Export to ONNX that is supported by the OpenVINO™ toolkit
  • DataParallel and DistributedDataParallel modes
  • Tensorboard-compatible output

Quantize FP32 Pretrained Model

This scenario demonstrates quantization with fine-tuning of MobileNet v2 on the ImageNet dataset.

Dataset Preparation

To prepare the ImageNet dataset, refer to the following tutorial.

Run Classification Sample

  • If you did not install the package, add the repository root folder to the PYTHONPATH environment variable.
  • Go to the examples/classification folder.
  • Run the following command to start compression with fine-tuning on GPUs: 
    python main.py -m train --config configs/quantization/mobilenet_v2_imagenet_int8.json --data /data/imagenet/ --log-dir=../../results/quantization/mobilenet_v2_int8/
    It may take a few epochs to get the baseline accuracy results.
  • Use the --multiprocessing-distributed flag to run in the distributed mode.
  • Use the --resume flag with the path to a previously saved model to resume training.

Validate Your Model Checkpoint

To estimate the test scores of your model checkpoint, use the following command:

python main.py -m test --config=configs/quantization/mobilenet_v2_imagenet_int8.json --resume <path_to_trained_model_checkpoint>

To validate an FP32 model checkpoint, make sure the compression algorithm settings are empty in the configuration file or pretrained=True is set.

Export Compressed Model

To export trained model to the ONNX format, use the following command:

python main.py -m test --config=configs/quantization/mobilenet_v2_imagenet_int8.json --resume=../../results/quantization/mobilenet_v2_int8/6/checkpoints/epoch_1.pth --to-onnx=../../results/mobilenet_v2_int8.onnx

Export to OpenVINO™ Intermediate Representation (IR)

To export a model to the OpenVINO IR and run it using the Intel® Deep Learning Deployment Toolkit, refer to this tutorial.

Results for quantization

Model Compression algorithm Dataset PyTorch compressed accuracy Config path PyTorch checkpoint
ResNet-50 None ImageNet 76.13 examples/classification/configs/quantization/resnet50_imagenet.json -
ResNet-50 INT8 ImageNet 76.05 examples/classification/configs/quantization/resnet50_imagenet_int8.json Link
ResNet-50 Mixed, 44.8% INT8 / 55.2% INT4 ImageNet 76.3 examples/classification/configs/quantization/resnet50_imagenet_mixed_int_hawq.json Link
ResNet-50 INT8 + Sparsity 61% (RB) ImageNet 75.22 examples/classification/configs/sparsity_quantization/resnet50_imagenet_rb_sparsity_int8.json Link
ResNet-50 INT8 + Sparsity 50% (RB) ImageNet 75.60 examples/classification/configs/sparsity_quantization/resnet50_imagenet_rb_sparsity50_int8.json Link
Inception V3 None ImageNet 77.32 examples/classification/configs/quantization/inception_v3_imagenet.json -
Inception V3 INT8 ImageNet 76.96 examples/classification/configs/quantization/inception_v3_imagenet_int8.json Link
Inception V3 INT8 + Sparsity 61% (RB) ImageNet 77.02 examples/classification/configs/sparsity_quantization/inception_v3_imagenet_rb_sparsity_int8.json Link
MobileNet V2 None ImageNet 71.81 examples/classification/configs/quantization/mobilenet_v2_imagenet.json Link
MobileNet V2 INT8 ImageNet 71.34 examples/classification/configs/quantization/mobilenet_v2_imagenet_int8.json Link
MobileNet V2 Mixed, 46.6% INT8 / 53.4% INT4 ImageNet 70.89 examples/classification/configs/quantization/mobilenet_v2_imagenet_mixed_int_hawq.json Link
MobileNet V2 INT8 + Sparsity 52% (RB) ImageNet 70.99 examples/classification/configs/sparsity_quantization/mobilenet_v2_imagenet_rb_sparsity_int8.json Link
SqueezeNet V1.1 None ImageNet 58.18 examples/classification/configs/quantization/squeezenet1_1_imagenet.json -
SqueezeNet V1.1 INT8 ImageNet 58.02 examples/classification/configs/quantization/squeezenet1_1_imagenet_int8.json Link
SqueezeNet V1.1 Mixed, 54.7% INT8 / 45.3% INT4 ImageNet 58.85 examples/classification/configs/quantization/squeezenet1_1_imagenet_mixed_int_hawq.json Link

Binarization

As an example of NNCF convolution binarization capabilities, you may use the configs in examples/classification/configs/binarization to binarize ResNet18. Use the same steps/command line parameters as for quantization (for best results, specify --pretrained), except for the actual binarization config path.

Results for binarization

Model Compression algorithm Dataset PyTorch compressed accuracy Config path PyTorch Checkpoint
ResNet-18 None ImageNet 69.76 examples/classification/configs/binarization/resnet18_imagenet.json -
ResNet-18 XNOR (weights), scale/threshold (activations) ImageNet 61.59 examples/classification/configs/binarization/resnet18_imagenet_binarization_xnor.json Link
ResNet-18 DoReFa (weights), scale/threshold (activations) ImageNet 61.56 examples/classification/configs/binarization/resnet18_imagenet_binarization_dorefa.json Link

Results for filter pruning

Model Compression algorithm Dataset PyTorch compressed accuracy Config path PyTorch Checkpoint
ResNet-50 None ImageNet 76.13 examples/classification/configs/quantization/resnet50_imagenet.json -
ResNet-50 Filter pruning, 30%, magnitude criterion ImageNet 75.7 examples/classification/configs/pruning/resnet50_pruning_magnitude.json Link
ResNet-50 Filter pruning, 30%, geometric median criterion ImageNet 75.7 examples/classification/configs/pruning/resnet50_pruning_geometric_median.json Link
ResNet-18 None ImageNet 69.76 examples/classification/configs/binarization/resnet18_imagenet.json -
ResNet-18 Filter pruning, 30%, magnitude criterion ImageNet 68.73 examples/classification/configs/pruning/resnet18_pruning_magnitude.json Link
ResNet-18 Filter pruning, 30%, geometric median criterion ImageNet 68.97 examples/classification/configs/pruning/resnet18_pruning_geometric_median.json Link
ResNet-34 None ImageNet 73.31 examples/classification/configs/pruning/resnet34_imagenet.json -
ResNet-34 Filter pruning, 30%, magnitude criterion ImageNet 72.54 examples/classification/configs/pruning/resnet34_pruning_magnitude.json Link
ResNet-34 Filter pruning, 30%, geometric median criterion ImageNet 72.62 examples/classification/configs/pruning/resnet34_pruning_geometric_median.json Link