January 12 2023
Last updated
Was this helpful?
Last updated
Was this helpful?
Introducing Markov Projects - a place where you can keep all your models, experiments and evaluations tracked in one place
You will now be able to associate all your experiments, evaluations and models within a project.
You can view your projects, and add new projects from the new tab in the sidebar.
You can register your models with MarkovML now. Models are used to consolidate your experiments and evaluations (earlier called Model Recordings) so that you can view your model training result and analyse its performance in one place.
You can manage your projects and models from the MarkovML SDK. You can find the instructions to do so in our documentation here.
You can now see a preview of the registered dataset from the dataset details screen.
You can now download chart data from any chart within MarkovML. This helps users understand the data that is used to create the chart. It can be downloaded in CSV format, so further processing can be done by the users themselves.
During dataset registration, all drop downs (like credentials, features, targets, data-family) are searchable.
For text based datasets, a new analysis is now supported that extracts normalized "skills" from the dataset. For example, from a resume dataset, the analyzer can extract hard skills like:
web-developer
python
and soft skills such as
hard-worker
The name Model Recordings has been renamed to Model Evaluations (or just Evaluations). This change can be found throughout the app, as well as in the docs. The methods in the SDK to create a model recording has been deprecated, and a new method to create an Evaluation has been added. You can find more about creating an evaluation here.
You can now use the MarkovML SDK to create email alerts. This can help you get notified over email when you like. From example, you can set your email alert after a long running model training process is complete.
Bug fixes and stability improvements
