For Data Scientists
Project owners will select templated analysis and machine learning tasks developed by data scientists for use in their project. This section outlines the necessary steps to set up an AI model, provision tasks utilising them and manage model usage requests.
Uploading a Model
There are two points of entry for uploading a custom model to Bitfount. In Bitfount Desktop and Hub click the “Create Model” button in the top right of the “Models” page:
- Name the model and add a description.
- Select if the model is “Public” or “Private”. Private models require usage approval once a task utilising them has been connected to a project. Public models can be freely utilised by any Bitfount user.
- If you have the model code file on-hand, you can upload it by clicking the “Upload File” button and selecting the appropriate file. Bitfount supports the upload of .py files for saving a model.
- If using a pre-trained model, you can also upload the model weights by clicking the “Add” button and selecting the appropriate file.
- If you’d prefer to copy and paste your model for saving, Bitfount provides a text function which allows you to do so in the black box.
- When you’re ready, click the “Create Model” button at the bottom of the page. If you make a mistake or no longer wish to save your model, click the “Discard Changes” button at the bottom of the page.
You can also upload a model via the SDK. This can be completed by specifying the relevant information when you upload the task template containing the model for the first time. Please see Custom Models and Task Templates for more information.
Managing Models
Once a model is saved, it can be updated, downloaded or deleted as desired. If your model is private, you will also need to manage model access requests.
Editing & Versioning
Bitfount supports model versioning in the Desktop App, Hub and SDK. To update a model simply click “Edit” make any changes required in the code editor in the UI, confirm by selecting “Save model”. The model version number will automatically update, and you can view historic versions by selecting the “Version” dropdown. Please note that tasks utilising the model will continue to run the version they were set up with until they have been updated. A new project will need to be created if you an existing project owner is looking to utilise the new version.
Downloading
Navigate to the desired model’s card, scroll to the bottom of the page, and click the “Download” button to download a .py file of your model code.
Deleting
Navigate to the desired model’s card, and click the “Delete Model” button in the top right corner of the page. You will be prompted to confirm you’d like to delete the model.
Usage approval
If you have flagged set your model as being private, you will be notified via email when a task containing the model has been selected for use in a project. At this stage participants cannot run tasks utilising the model until usage has been approved. Review all usage requests by selecting the model card, navigating to the ”Projects” tab, clicking “Approve usage” or “Decline” as appropriate.
Projects do not currently support selective model choice when running tasks. Running a task will execute all models or queries provisioned to the given project within the task template.
Managing Tasks
Tasks are defined and created using the SDK. Please refer to the Task Templates to review the process for actioning this.
Updating
To update a task, such as the model version it uses, you must make the desired changes to the task template, and task YAML files. Once complete re-run the upload script to update.
Any projects using the task before an update was made will not consume the new version. If an existing project needs to run the new task, currently a new project must first be created.
Need Help?
If you have any questions after reviewing our Guides and Tutorials, visit our FAQs. If you can't find the answer you are looking for or would like to discuss anything further, please contact us at support@bitfount.com. We're here to help!