model registry
In plain wordsA central catalog where a team stores every version of its trained models, with records of how each was made and which version is approved for use.
A versioned, access-controlled store for trained model artifacts and their metadata that serves as the system of record between training and deployment. Each registered version links the weightsweights, n.: The learned numbers inside a neural network that set how strongly each part influences the next. The finished set of weights is, in effect, the trained model itself. or serialized model to its lineage — training-code commit, data snapshot, hyperparametershyperparameter, n.: A setting chosen by people before training, such as how fast a model learns or how big it is, as opposed to the internal values the model learns by itself from data., trackedexperiment tracking, n.: Keeping an automatic logbook of every model-training attempt — settings, data, code and results — so a team can compare attempts, repeat the good ones and see what changed. metrics — and to its signature (expected inputs and outputs) and software dependencies. Versions move through lifecycle stages or aliases such as candidate, staging, production and archived, often behind approval gates, so that deployment systems can fetch 'the current production model' by name and roll back by pointing to an earlier version. Registries support audit and governance by recording who approved what and when. Examples include the registries in MLflowMLflow, prop. n.: A free software platform that records machine-learning experiments, packages the resulting models and keeps track of which versions are ready to be used., Amazon SageMaker, Azure Machine Learning, Vertex AI and Weights & BiasesWeights & Biases, prop. n.: A company and online service that machine-learning teams use to log, chart and compare their training experiments, models and AI-app test results in shared dashboards.; for LLMlarge language model, n.: A computer program trained on enormous amounts of text to predict the words that come next. At vast scale, that simple skill lets it chat, answer questions, translate, summarize and write code. applications, prompts and provider model versions are increasingly registered alongside weights.
Promote version 14 in the model registry and the serving cluster will pick it up within five minutes.
Origin
Grew out of in-house machine-learning platforms of the mid-2010s and became a standard product feature in the late 2010s; MLflowMLflow, prop. n.: A free software platform that records machine-learning experiments, packages the resulting models and keeps track of which versions are ready to be used. added a Model Registry component in 2019, and the major cloud ML platforms offer equivalents.
Usage note
Not the same as a public model hub such as Hugging FaceHugging Face, prop. n.: A company whose website is the main public library for sharing AI models and datasets — often called the GitHub of machine learning — and which makes widely used free AI software.'s, which distributes models to outsiders; a registry governs an organization's own release process. A higher version number identifies a newer artifact, not necessarily a better one.
Related
- MLOpsMLOps, n.: The working habits and tools that keep machine-learning models dependable inside real products once the experiment is over: automated testing, release, monitoring and retraining; the factory and maintenance crew around a model, not the model itself.
- experiment trackingexperiment tracking, n.: Keeping an automatic logbook of every model-training attempt — settings, data, code and results — so a team can compare attempts, repeat the good ones and see what changed.
- model deploymentmodel deployment, n.: Putting a trained AI model to work in a real product so that people or other software can use it — and doing so carefully enough that the change can be undone if something goes wrong.
- MLflowMLflow, prop. n.: A free software platform that records machine-learning experiments, packages the resulting models and keeps track of which versions are ready to be used.
- model cardmodel card, n.: A short document published with an AI model explaining what it is for, how it was trained and tested, how well it works for different groups of people, and its known limits.
- machine learning pipelinemachine learning pipeline, n.: An automated chain of steps that takes raw data, prepares it, trains and tests a model and hands it over for use, so the whole process can be rerun the same way.
Mentioned in
experiment trackingexperiment tracking, n.: Keeping an automatic logbook of every model-training attempt — settings, data, code and results — so a team can compare attempts, repeat the good ones and see what changed., machine learning pipelinemachine learning pipeline, n.: An automated chain of steps that takes raw data, prepares it, trains and tests a model and hands it over for use, so the whole process can be rerun the same way., MLOpsMLOps, n.: The working habits and tools that keep machine-learning models dependable inside real products once the experiment is over: automated testing, release, monitoring and retraining; the factory and maintenance crew around a model, not the model itself., model deploymentmodel deployment, n.: Putting a trained AI model to work in a real product so that people or other software can use it — and doing so carefully enough that the change can be undone if something goes wrong., data versioningdata versioning, n.: Keeping a record of every version of the data used to train and test AI models, so that any past result can be traced back to exactly the data that produced it., MLflowMLflow, prop. n.: A free software platform that records machine-learning experiments, packages the resulting models and keeps track of which versions are ready to be used., model deprecationmodel deprecation, n.: When an AI company announces that an older model will stop being offered and then switches it off, so the apps and people relying on it must move to a newer one., model hubmodel hub, n.: An online library where trained AI models are published so that others can find, download and reuse them instead of training their own from scratch., Weights & BiasesWeights & Biases, prop. n.: A company and online service that machine-learning teams use to log, chart and compare their training experiments, models and AI-app test results in shared dashboards.