Torch Model Archiver is a tool used for creating archives of trained neural net models that can be consumed by TorchServe inference.
Use the Torch Model Archiver CLI to start create a .mar file.
Torch Model Archiver is part of TS. However, you can install Torch Model Archiver stand alone.
Detailed documentation and examples are provided in the README.
pip install torch-model-archiver
You can check the latest source code as follows:
git clone https://github.com/pytorch/serve.git
If you use torchserve in a publication or project, please cite torchserve: https://github.com/pytorch/serve