Terminology#
Model#
A trained ML model instance needs to be saved with BentoML API. A model can be pushed to and pulled from Yatai. See the BentoML documentation for a more detailed explanation of Model.
Model Registry#
The model registry is a hub for storing, versioning, and sharing models for collaboration. The relationship between model registry
and models
is analogous to Docker registry
and Docker images
.
Bento#
Bento 🍱 is a file archive with all the source code, models, data files and dependency configurations required for running a user-defined bentoml.Service, packaged into a standardized format. See the BentoML documentation for a more detailed explanation of Bento.
Bento Registry#
The bento registry is a hub for storing, versioning, and sharing Bento
for collaboration. The relationship between Bento registry
and Bentos
is analogous to Docker registry
and Docker images
.
BentoRequest CRD#
BentoRequest CRD is a Kubernetes Custom Resource Definition (CRD) added to the Kubernetes cluster by yatai-image-builder. Each BentoRequest CR will generate a Bento CR with the same name after the OCI image is built. The CRD describes Bento image build information and runners information. For a full list of the possible descriptive fields and an example CRD, see BentoRequest CRD.
Bento CRD#
Bento CRD is a Kubernetes Custom Resource Definition (CRD) added to the Kubernetes cluster by yatai-image-builder. Bento CRs are often generated through the BentoRequest CR, but you can create a Bento CR manually, and yatai-deployment relies on the Bento CR to get the Bento information. The CRD describes Bento image information and Bento runners information. For a full list of the possible descriptive fields and an example CRD, see Bento CRD.
BentoDeployment CRD#
BentoDeployment CRD is a Kubernetes Custom Resource Definition (CRD) added to the Kubernetes cluster by yatai-deployment. The CRD describes Bento deployment. For a full list of the possible descriptive fields and an example CRD, see BentoDeployment CRD.