Can I export/download the model file created after completing the training with the credits?
This is good feedback, however, and something the team is actively considering once scoping out how this would work.
We do give access to an always-on/always-ready API endpoint; edge device deployment for devices such as iOS and NVIDIA Jetsons; and more deployment options for Enterprise servers and GPU via Docker.
Everything listed as “On Device” in parentheses here (including all Enterprise options) are available offline, without API.
Testing local deployment: Launch: Test Computer Vision Models Locally
For weights, you would need to train your own model with the model library (with your own GPU, a cloud-hosted GPU, or hosted GPU notebook such as Google Colab or AWS SageMaker Studio Labs):
- Full tutorials (written guides and video) for training are available here: GitHub - roboflow/notebooks: Set of Jupyter Notebooks linked to Roboflow Blogpost and used in our YouTube videos.
- Testing local deployment: Launch: Test Computer Vision Models Locally
Do we have any limit on the number of times I can use the model for a local deployment? i.e Is there a limit on No. of times I can call model.predict()?
Hi @joseva_vacha, for deployment to devices, there is a device limit of 5 free devices for deployment.
These are the posted device limits on the Pricing page for Public workspaces, for example:
If you ever receive an error on the device count (i.e the system reading as you’ve deployed to 5 total devices, when you haven’t) please do email us, and I’m happy to get the setting reconfigured.
If you use regular Hosted API Inference, there are 1,000 free inference credits provided each month for projects in your workspace.
Additionally, feel free to Apply for Research Credits, or to become a Blog Contributor, for access to higher device limits without charge: Contribute to the Computer Vision Community
We’ll get your form reviewed, and set the device limits for your project based on what is approved.
If you are just working on a short-term personal project, we’re also happy to add higher limits for tagging us in a social media post about the project, or including a link to the project in your GitHub repository.