Models available in Datrics platform
The current list of models that are available out of the box is:
Each of this base models can be trained using the pipeline and retrained using the deployed pipeline. To understand how to train a model via multiple pipelines or train it in 1 pipeline and use in the deployment, check the linked model page
Model Repository
Each model could be saved into the model repository, just click on model info and then “save model” button.
Models could be versioned.
All models appeared in the Models Repository. This repository is shared between all projects of the group. Models could be linked and deployed from there.
Save the model into the docker
Each model could be downloaded as JSON file (to run via datrics-model python library) or as a docker image.
Docker image could be saved automatically into the provided docker images registry that should be specified by the system administrator during the installation
Model docker deployment
To deploy Model, just run Docker container.
The container will have the endpoint
And send JSON with all features used to train model.
Example:
[ { "annual_inc_bins_higher_75k": 0, "annual_inc_bins_less_50k": 1, "dti_bins_higher_20": 1, "dti_bins_less_15": 0, "tot_cur_bal_bins_higher_12k": 0, "tot_cur_bal_bins_less_12k": 1, "fico_range_high_bins_(643.999, 684.0]": 1, "fico_range_high_bins_(709.0, 850.0]": 0, "lti_bins_higher_0.1": 0, "lti_bins_less_0.06": 1, "funded_amnt_inv_bins_higher_15k": 0, "funded_amnt_inv_bins_less_9k": 1, "home_ownership_MORTGAGE": 0, "home_ownership_RENT": 1, "verification_status_Not Verified": 1, "verification_status_Verified": 0, "purpose_credit_card": 0, "grade_A": 0, "grade_B": 1, "grade_D": 0, "grade_E": 0, "grade_F": 0, "grade_G": 0, "normalized_weight": 1 } ]
As a result you will receive
task_idUse this task_id to get the result:
https://yourmodelcontainer:8000/status/{task_id}
To check the health of the model use GET /health endpoint