Export Formats
We currently provide the following options for artifact export on Nexus:
We don't support exporting CoreML and TensorRT for all models. Please contact us if you have any questions!
Model Export Type | Description |
---|---|
Generate TensorFlow Model | The regular TensorFlow model can be further edited and altered because it is still in its original form, so if the size of the model is not an issue for application, then one should opt for the TensorFlow model. |
Generate LiteRT (TFLite) Model | LiteRT (TensorFlow Lite) is designed to be a more lightweight, more efficient model that can be run on smaller devices like IoT devices or other mobile devices. Supports Post-Training Quantization. |
Generate ONNX Model | ONNX is utilized as an interim representation of your model that enables users to switch between different environments. |
Generate PyTorch Model | PyTorch is well supported on major cloud platforms, providing frictionless development and easy scaling. |
Generate CoreML Model | CoreML is designed to be a more lightweight, more efficient model that can be run on Apple iOS devices like iPhones and iPads. Supports Post-Training Quantization. |
Generate TensorRT Model | TensorRT is designed to be optimized specifically for NVIDIA GPUs or NVIDIA-based edge devices at the kernel level. Supports Post-Training Quantization . |
Common Questions
Are all YOLOv8 model export formats supported?
Datature does support Ultralytics YOLOv8 models on the platform, and as such, any YOLOv8 models trained on Datature's Nexus can be exported as PyTorch models, and using Ultralytics API, they can be converted into any of those formats listed in the link.
Updated 29 days ago