Getting started
🔧
Connect to data
⏰
Schedule Runs
🤝
Share and collaborate
Extras
🔐
Security & Permissions
APP & USER MONITORING
Learn more
🔗
Links
Save model
Saving your ML models is what takes Dolphyn from a place of analysis to a machine learning power mill. 🚅
Saving a model
There are 3 main steps to take in order to save a model to your workspace.
1. Add the save model block to your page
Open the
/ menu and search for 'save model' or type /model to find the save model block.2. Select the model variable
Saving a model requires a python variable that represents the machine learning model. Some common examples can be found in libraries such as Prophet, Scikit-Learn, Tensorflow, and others.
Once you add the
Save Model block to your page, the first thing you'll need to do is set the source variable. This is the python variable that represents the machine learning model.
3. Name and choose the serialization framework
Once you've selected the
source variable, click on the Save as dropdown. This will open a popup where you'll need to do 2 things:- 1.Give your model a name
- 2.Select the serialization framework. Pick the framework that corresponds with the library that you used to build the model.
Supported frameworks
Don't see a framework you need listed here? Get in touch at [email protected] and we'll see if we can put it on the fast track.
At this time, Dolphyn supports the following serialization frameworks out of the box:
- ScikitLearn
- Prophet
- XGBoost
- Tensorflow
- PyTorch
Where does a saved model go?
The saved model is available in the
Models database in your workspace. Learn more about how the workspace database organization works. From there you'll be able to access the model, add metadata, share a link to it with collaborators, and download it.You can navigate there by going to the workspace sidebar and clicking on
Models.
Auto-versioning models
Versioning models is currently in private beta. Get in touch at [email protected] if you'd like early access or to learn more.
One last thing, deploy your model as an API
You may want to serve this model over the internet to enable other apps and websites to use it for real-time predictions. In the next section, we'll cover the Create API block, where you can one-click deploy models as APIs.
Last modified 2mo ago