BigDL-Nano PyTorch Quantization with ONNXRuntime accelerator Quickstart

In this guide we will describe how to obtain a quantized model running inference in the ONNXRuntime engine with the APIs delivered by BigDL-Nano in 4 simple steps

Step 0: Prepare Environment

We recommend using conda to prepare the environment. Please refer to the install guide for more details.

conda create py37 python==3.7.10 setuptools==58.0.4
conda activate py37
# nightly bulit version
pip install --pre --upgrade bigdl-nano[pytorch]
# set env variables for your conda environment
source bigdl-nano-init

To quantize model using ONNXRuntime as backend, it is required to install Intel Neural Compressor, onnxruntime-extensions as a dependency of INC and some onnx packages as below

pip install neural-compress==1.11
pip install onnx onnxruntime onnxruntime-extensions

Step 1: Load the data

import torch
from torchvision.io import read_image
from torchvision import transforms
from torchvision.datasets import OxfordIIITPet
from torch.utils.data.dataloader import DataLoader

train_transform = transforms.Compose([transforms.Resize(256),
                                      transforms.RandomCrop(224),
                                      transforms.RandomHorizontalFlip(),
                                      transforms.ColorJitter(brightness=.5, hue=.3),
                                      transforms.ToTensor(),
                                      transforms.Normalize([0.485, 0.456, 0.406], [0.229, 0.224, 0.225])])
val_transform = transforms.Compose([transforms.Resize([224, 224]), transforms.ToTensor(), transforms.Normalize([0.485, 0.456, 0.406], [0.229, 0.224, 0.225])])
# Apply data augmentation to the tarin_dataset
train_dataset = OxfordIIITPet(root = ".", transform=train_transform)
val_dataset = OxfordIIITPet(root=".", transform=val_transform)
# obtain training indices that will be used for validation
indices = torch.randperm(len(train_dataset))
val_size = len(train_dataset) // 4
train_dataset = torch.utils.data.Subset(train_dataset, indices[:-val_size])
val_dataset = torch.utils.data.Subset(val_dataset, indices[-val_size:])

train_dataloader = DataLoader(train_dataset, batch_size=32)

Step 2: Prepare your Model

import torch
from torchvision.models import resnet18
from bigdl.nano.pytorch import Trainer
from torchmetrics import Accuracy
model_ft = resnet18(pretrained=True)
num_ftrs = model_ft.fc.in_features

# Here the size of each output sample is set to 37.
model_ft.fc = torch.nn.Linear(num_ftrs, 37)
loss_ft = torch.nn.CrossEntropyLoss()
optimizer_ft = torch.optim.SGD(model_ft.parameters(), lr=0.01, momentum=0.9, weight_decay=5e-4)

# Compile our model with loss function, optimizer.
model = Trainer.compile(model_ft, loss_ft, optimizer_ft, metrics=[Accuracy])
trainer = Trainer(max_epochs=5)
trainer.fit(model, train_dataloader=train_dataloader)

# Inference/Prediction
x = torch.stack([val_dataset[0][0], val_dataset[1][0]])
model_ft.eval()
y_hat = model_ft(x)
y_hat.argmax(dim=1)

Step 3: Quantization with ONNXRuntime accelerator

With the ONNXRuntime accelerator, Trainer.quantize() will return a model with compressed precision but running inference in the ONNXRuntime engine.

you can add quantization as below:

from torchmetrics.functional import accuracy
ort_q_model = trainer.quantize(model, accelerator='onnxruntime', calib_dataloader=train_dataloader, metric=accuracy)

# run simple prediction
y_hat = ort_q_model(x)
y_hat.argmax(dim=1)