Techniques for dynamic machine learning integration
Various embodiments are generally directed to techniques for dynamically integrating ML functionality into computing systems, such as a content services platform (CSP), for instance. Many embodiments include ML integrated into a CSP and using production content as corpora (e.g., training and/or evaluation data). Some embodiments are particularly directed to generating and updating data for training and evaluating machine learning (ML) models, then making identified ML models available in various target environments. For example, embodiments may provide automatic, or semi-automatic, updating and deploying of ML models for making inferences, such as inferring labels for data in a content repository of a CSP.
1. An apparatus, the apparatus comprising:
a processor; and
a memory comprising instructions that when executed by the processor cause the processor to:
obtain source data from one or more data sources, the source data comprising a plurality of data items;
produce training data based on the source data, wherein production of the training data includes removal of at least one of the plurality of data items from the source data based on one or more data metrics associated with a machine learning framework;
provide the training data to the machine learning framework to generate a model version;
receive the model version from the machine learning framework;
deploy a docker image, with the model version, based on a status setting corresponding to the model version;
deploy the docker image to a target environment to deploy the model version to the target environment;
register the model version with an instance of a production repository to associate the model version to the target environment;
generate a log of activity associated with the training data via an audit system that provides indications to distinguish content of machine language models from manual input, wherein the log of activity is associated with deploying the model version to the target environment based on the status setting corresponding to the model version, and with integrating the production repository associated with the model version to the target environment; and
restore the target environment and the one or more data sources to a previous state based on the log of activity and a historical index.
2. The apparatus of claim 1 , wherein the target environment includes a content services platform.
3. The apparatus of claim 2 , wherein the target environment includes a content repository of the content services platform.
4. The apparatus of claim 1 , the memory comprising instructions that when executed by the processor cause the processor to generate the log of activity associated with one or more of:
obtaining the source data from the one or more data sources,
producing the training data based on the source data,
providing the training data to the machine learning framework, and
receiving the model version from the machine learning framework.
5. The apparatus of claim 1 , wherein each data item includes data and metadata corresponding to the data.
6. The apparatus of claim 1 , wherein production of the training data includes transformation of at least one of the plurality of data items in the source data from a format incompatible with the machine learning framework into a format compatible with the machine learning framework.
7. The apparatus of claim 1 , wherein the data metric comprises one or more of a file type, a file size threshold, a quality threshold, and a reliability indicator.
8. The apparatus of claim 1 , wherein production of the training data includes normalizing the source data.
9. The apparatus of claim 1 , wherein the log of activity associated with the training data via the audit system enables actions from a prediction to be backtracked to the previous state.
10. At least one non-transitory computer-readable medium comprising a set of instructions that, in response to being executed by a processor circuit, cause the processor circuit to:
obtain source data from one or more data sources, the source data comprising a plurality of data items;
produce training data based on the source data, wherein production of the training data includes removal of at least one of the plurality of data items from the source data based on one or more data metrics associated with a machine learning framework;
provide the training data to the machine learning framework to generate a model version;
receive the model version from the machine learning framework;
deploy a docker image, with the model version, based on a status setting corresponding to the model version;
deploy the docker image to a target environment to deploy the model version to the target environment;
register the model version with an instance of a production repository to associate the model version to the target environment;
generate a log of activity associated with the training data that provides indications to distinguish content of machine language modes from manual input; wherein the log of activity is associated with deploying the model version to the target environment based on the status setting corresponding to the model version, and with integrating the production repository associated with the model version to the target environment; and
restore the target environment and the one or more data sources to a previous state based on the log of activity and a historical index.
11. The at least one non-transitory computer-readable medium of claim 10 , wherein the target environment includes a content services platform.
12. The at least one non-transitory computer-readable medium of claim 11 , wherein the target environment includes a content repository of the content services platform.
13. The at least one non-transitory computer-readable medium of claim 10 , comprising instructions that, in response to being executed by the processor circuit cause the processor circuit to generate the log of activity associated with one or more of obtaining the source data from the one or more data sources, producing the training data based on the source data, providing the training data to the machine learning framework, receiving the model version from the machine learning framework, deploying the model version to the target environment based on the status setting corresponding to the model version, and integrating the production repository associated with the model version to the target environment.
14. The at least one non-transitory computer-readable medium of claim 10 , wherein each data item includes data and metadata corresponding to the data.
15. The at least one non-transitory computer-readable medium of claim 10 , wherein production of the training data includes transformation of at least one of the plurality of data items in the source data from a format incompatible with the machine learning framework into a format compatible with the machine learning framework.
16. The at least one non-transitory computer-readable medium of claim 10 , wherein the log of activity associated with the training data enables actions from a prediction to be backtracked to the previous state.
17. A computer-implemented method, comprising:
obtaining source data from one or more data sources, the source data comprising a plurality of data items;
producing training data based on the source data, wherein production of the training data includes removal of at least one of the plurality of data items from the source data based on one or more data metrics associated with a machine learning framework;
providing the training data to the machine learning framework to generate a model version;
receiving the model version from the machine learning framework;
deploying a docker image, with the model version, based on a status setting corresponding to the model version;
deploying the docker image to a target environment to deploy the model version to the target environment;
registering the model version with an instance of a production repository to associate the model version to the target environment;
generate a log of activity associated with the training data that provides indications to distinguish content of machine language modes from manual input, wherein the log of activity is associated with deploying the model version to the target environment based on the status setting corresponding to the model version, and with integrating the production repository associated with the model version to the target environment; and
restoring the target environment and the one or more data sources to a previous state based on the log of activity and a historical index.
18. The computer-implemented method of claim 17 , comprising generating the log of activity associated with one or more of obtaining the source data from the one or more data sources, producing the training data based on the source data, providing the training data to the machine learning framework, receiving the model version from the machine learning framework, deploying the model version to the target environment based on the status setting corresponding to the model version, and integrating the production repository associated with the model version to the target environment.
19. The computer-implemented method of claim 17 , wherein production of the training data includes transforming at least one of the plurality of data items in the source data from a format incompatible with the machine learning framework into a format compatible with the machine learning framework.
20. The computer-implemented method of claim 17 , wherein the log of activity associated with the training data enables actions from a prediction to be backtracked to the previous state.