Deploy EAS Inference Services for One-Click Model Serving - Platform for AI

Configure a service deployment configuration once during model registration, and it is automatically applied every time you deploy that model to Elastic Algorithm Service (EAS) with one-click deployment — no manual editing required.

Set a deployment configuration

When registering a new model, scroll to the Model Service Deployment Configuration section. Select Custom Configuration and paste your deployment configuration JSON into the text box.

Choose the configuration template that matches your deployment approach:

Custom container image

Use this when you want full control over the runtime environment by bringing your own container image.

{
  "containers": [
    {
      "image": "registry-vpc.cn-shanghai.aliyuncs.com/xxx/yyy:zzz",
      "env": [
        {
          "name": "var_name",
          "value": "var_value"
        }
      ],
      "command": "/data/eas/ENV/bin/python /data/eas/app.py",
      "port": 8000
    }
  ],
  "storage": [
    {
      "oss": {
        "readOnly": false
      },
      "properties": {
        "resource_type": "model"
      }
    }
  ]
}

For the complete list of EAS configuration parameters, see Custom images and Service deployment.

Preset processor

Use this when EAS provides a built-in processor that matches your model's framework and hardware requirements.

{
  "processor": "tensorflow_gpu_1.12"
}

For available processor names, see PMML processor.

Custom processor

Use this when you need a custom processor defined by your own service script.

{
  "processor_entry": "./service.py",
  "processor_type": "python",
  "processor_path": "http://eas-data.oss-cn-shanghai.aliyuncs.com/demo/service.py",
  "data_image": "registry.cn-shanghai.aliyuncs.com/eas-service/develop:latest"
}

For details on building a custom processor, see Custom processor.

Deploy a model with one-click deployment

Prerequisites

Currently, only models from Model Gallery support one-click deployment. One-click deployment does not support models registered from OSS or from custom training jobs.
For models registered from OSS or custom training, use custom deployment instead: mount your OSS model path to the container and configure the service manually.

Steps

On the Model Management page, click Deploy to EAS in the Operation column for the model, then follow the on-screen instructions. You are automatically redirected to the EAS service deployment page.
On the service deployment page, key parameters in the model service information area are prefilled from the deployment configuration you set during registration. Configure any remaining required parameters and start the deployment.
For details on each parameter, see Custom deployment.

What's next

Service deployment — full EAS deployment configuration reference
Custom deployment — deploy models from OSS or custom training