Cellulose is happy to launch PyTorch support today! With our newly released Python SDK, you’ll be able to export and upload your PyTorch models to the dashboard. The best part? It’s available to all users for free!
Uploading your PyTorch models to the dashboard is simple. First install the Python SDK, add the Cellulose decorator to your PyTorch nn.Module, and then call export by providing an instance to the model and an input tensor. We’ll handle the rest!
The only requirement is that the PyTorch models have to be TorchScript-able and exportable to ONNX so they can be uploaded. Don’t worry as the SDK does that automatically for you.
Here are some resources with more detailed instructions for you to get started.
We’re currently talking to early users, and some have already expressed interest in other model formats that better fit their workflow, such as TorchScript (JIT IR) or FX graph (FX IR). Fear not, we plan to add support soon.
We can’t wait to see what you’ll build next with Cellulose :)
Follow us on LinkedIn and/or Twitter to stay up-to-date on the latest announcements. Questions? Feel free to reach out to support@cellulose.ai!
