Overview
Following the Model Deployment process, you can quickly and easily send publish them in the model library. Below is the API Reference with details to follow the process, however, we recommend to use the Model Deployment page to do so.
Only Data Scientists can deploy models.
Steps
To deploy a model, start by creating it and adding the model container image and metadata. To finish, set a model picture, add tags, and update the model’s status as active.
Below is the list of steps to deploy a model.
| Step | Method | Endpoint | |
|---|---|---|---|
| 1 | Create a model | POST | /api/models |
| 2 | Add a name and summary | PATCH | /api/models/:modelId |
| 3 | Add the model container image | POST | /api/models/:modelId/versions/:version/container-image |
| 4 | Add requirements | PATCH | /api/models/:modelId/versions/:version |
| 5 | Add timeouts | PATCH | /api/models/:modelId/versions/:version |
| 6 | Add metrics | PATCH | /api/models/:modelId/versions/:version |
| 7 | Test: load the model | POST | /api/models/:modelId/versions/:version/load-process |
| 8 | Create inputs | PATCH | /api/models/:modelId/versions/:version |
| 9 | Create outputs | PATCH | /api/models/:modelId/versions/:version |
| 10 | Add a test file | POST | /api/models/:modelId/versions/:version/testInput |
| 11 | Test: run the model | POST | /api/models/:modelId/versions/:version/run-process |
| 12 | Add a description | PATCH | /api/models/:modelId/versions/:version |
| 13 | Add release notes | PATCH | /api/models/:modelId/versions/:version |
| 14 | Add transparency and bias reporting | PATCH | /api/models/:modelId/versions/:version |
| 15 | Add tags | PATCH | /api/models/:modelId |
| 16 | Add a model picture | POST | /api/models/:modelId/image |
| 17 | Deploy the model | PATCH | /api/models/:modelId/versions/:version |
The models object
{
"modelId": "...",
"latestVersion": "...",
"latestActiveVersion": "...",
"versions": [],
"author": "...",
"name": "...",
"description": "...",
"permalink": "...",
"features": [],
"isActive": false,
"isRecommended": false,
"isCommercial": false,
"tags": [],
"images": [],
"lastActiveDateTime": "...",
"createdByEmail": "...",
"createdByFullName": "...",
"visibility": {
"scope": "...",
"teams": []
}
}
| Parameter | Type | Description |
|---|---|---|
| modelId | string | A model identifier provided by Modzy. It contains 10 random alphanumeric characters. |
| latestVersion | string | The model’s most recent active version. This version is displayed by default in the Model Details page. |
| latestActiveVersion | string | The model’s most recent active version. This version is displayed by default in the Model Details page. |
| versions | array | An array that contains all the version objects. |
| author | string | The organization that created the model. |
| name | string | The model’s name. It may only contain alphanumeric, dash, and underscore characters. Be sure to use a name that describes the model’s functionality so that others can find the model easily. |
| description | string | The model details summary. It’s featured in the Model Library. |
| permalink | string | A link to the model. It combines the modelId, author, and name. |
| features | array | An array that contains all the feature objets. |
| isActive | boolean | Active models are available in the Marketplace and can execute inference on data. At least one active version is required for the model to remain active. |
| isRecommended | boolean | Adds a "recommended" badge to the model. It features in the Model Marketplace. Defaults to false. |
| isCommercial | boolean | The model’s commercial status. Commercial models can be purchased. |
| tags | array | An array that contains all the tag objects. |
| images | array | An array that contains all the image objects. |
| lastActiveDateTime | string | The last time the model was used in ISO8601 (YYYY-MM-DDThh:mm:ss.sTZD) format. |
| createdByEmail | string | The email of the user that deployed the model. |
| createdByFullName | string | The full name of the user that deployed the model. |
| visibility | object | An object that contains a model’s scope that defines who can see the model (ALL, ACCOUNT, TEAM, PRIVATE) and the model’s teams. |
Model picture
Models require a model picture set at the Model Deployment. This picture is resized by Modzy and set as the model’s thumbnail, card, and background pictures before the model is published. The model’s logo picture is inherited from the account’s picture and cannot be updated. The Model service has APIs to upload and get pictures.
Accepted files
Modzy supports JPG, JPEG, and PNG pictures with a maximum size of 5 MB. The minimum resolution is 1440 x 1024 px.
Resizing
Modzy resizes pictures if the width or the height dimensions vary to match the thumbnail, card, and background pictures. The resizing origin is the center of the image.
The pictures object
{
"url" : "...",
"caption" : "...",
"alt" : "...",
"relationType" : "..."
}
| Parameter | Type | Description |
|---|---|---|
| url | string | The model’s picture location. |
| caption | string | A picture’s description available to all users, set upon model deployment. Defaults to the model’s name. |
| alt | string | A picture’s alternate description for the search engine and screen readers. It matches the caption description. |
| relationType | string | It defines where and how the model images are sized and visible in the Model Marketplace. Values are set to thumbnail, background, and card for the model picture and logo for the company logo. |
Tags
Tags describe and filter models across the Model Library and are visible in the sidebar and search box. Modzy has three predefined tag categories: Task, Subject, and Input type and provides an option to create custom tags.
Predefined tags
Each category of predefined tags category answers a question:
| Category | Answers |
|---|---|
| Task | How does the model process data? |
| Subject | What content does the model process? |
| Input Type | What are the model’s accepted data types? |
Tags by category
| Subject | Task | Input Type |
|---|---|---|
| Language and Text | Label or Classify | Audio |
| Colors, Textures, and Patterns | Count | Binary |
| Computers and Software | Enhance or Preprocess | Image |
| Equipment and Machinery | Forecast or Estimate | Signal |
| Food | Locate or Detect | Time Series |
| Geography | Track | Text |
| Graph | ||
| Infrastructure and Buildings | ||
| Other | ||
| People and Activities | ||
| Satellites and Space | ||
| Topics, Ideas, and Sentiment | ||
| Vehicles and Transportation | ||
| Video, Images, and Illustrations | ||
| Weapons |
The tags object
"tags": [
{
"identifier": "count",
"name": "Count",
"dataType": "Task",
"isCategorical": true,
"images": []
}
]
| Parameter | Type | Description |
|---|---|---|
| identifier | string | A tag identifier provided by Modzy. It contains the tag’s name in lower case and it replaces spaces with underscores. |
| name | string | A tag name to filter and search for the model. |
| dataType | string | Defines a tag’s category. Values for categorical tags may be Task, Subject, and Input Type. For non-categorical tags the value is Tags. |
| isCategorical | boolean | A flag for predefined tags. If true, the tag is listed in the Model Marketplace’s navigation bar. If false, the tag is listed only in the Model Details page. |
| images | array | An array that contains the tag’s image objects. |