This guide provides detailed instructions for exporting PyTorch machine learning models to the ONNX format. It aims to simplify the process of converting models to ONNX for interoperability and deployment across different platforms and frameworks.
Before you begin, ensure you have the following:
- Python 3.6+
- pip
- Pytorch
- Basic understanding of machine learning models and frameworks
- ONNX and latest onnxruntime
- Netron.app for visualizing and analyzing ONNX models
To get started, you'll need to install the necessary libraries. You can install the required libraries by using:
pip install -r requirements.txt
Classification
- AlexNet
- ConvNeXt_base
- ConvNeXt_small
- ConvNeXt_large
- DenseNet121
- DenseNet161
- DenseNet169
- DenseNet201
- EfficientNet_BO
- EfficientNet_B1
- EfficientNet_B2
- EfficientNet_B3
- EfficientNet_B4
- EfficientNet_B5
- EfficientNet_B6
- EfficientNet_B7
- EfficientNet_V2_S
- EfficientNet_V2_M
- EfficientNet_V2_L
- GoogleNet
- Inception_V3
- MNASNet0_5
- MNASNet0_75
- MNASNet1_0
- MNASNet1_3
- MaxiVit_T
- MobileNet_V2
- MobileNet_V3_S
- MobileNet_V3_L
- ResNet18
- ResNet34
- ResNet50
- ResNet101
- ResNet152
- ShuffleNet
Object Detection Models
- FasterRCNN_V1 (fasterrcnn_resnet50_fpn)
- FasterRCNN_V2 (fasterrcnn_resnet50_fpn_v2)
- FasterRCNN_V3_L (fasterrcnn_mobilenet_v3_large_fpn)
- FasterRCNN_V3_L_320 (fasterrcnn_mobilenet_v3_large_320_fpn)
- YOLOV8n
- YOLOV8s
- YOLOV8m
- YOLOV8l
- YOLOV8x
Image Segmentation Models
- UNet
Natural Language Processing
- BERT
Find practical examples in the Examples directory.
This project is licensed under the MIT License.