Run Your First Custom Inference Workload

This quick start provides a step-by-step walkthrough for running and querying a custom inference workload.

An inference workload provides the setup and configuration needed to deploy your trained model for real-time or batch predictions. It includes specifications for the container image, data sets, network settings, and resource requests required to serve your models.

Prerequisites

Before you start, make sure:

• You have created a project or have one created for you.

• The project has an assigned quota of at least 1 GPU.

• Knative is properly installed by your administrator.

Step 1: Logging In

UI

Browse to the provided NVIDIA Run:ai user interface and log in with your credentials.

API

To use the API, you will need to obtain a token as shown in API authentication.

Step 2: Submitting an Inference Workload

UI

1. Go to Workload manager → Workloads.

2. Click +NEW WORKLOAD and select Inference

3. Within the new form, select the cluster and project

4. Enter a unique name for the workload. If the name already exists in the project, you will be requested to submit a different name.

5. Click CONTINUE

   In the next step:

6. Create an environment for your workload

   • Click +NEW ENVIRONMENT

   • Enter a name for the environment. The name must be unique.

   • Enter the Image URL - runai.jfrog.io/demo/example-triton-server

   • Set the inference serving endpoint to HTTP and the container port to 8000

   • Click CREATE ENVIRONMENT

   The newly created environment will be selected automatically

7. Select the ‘half-gpu’ compute resource for your workload (GPU devices: 1)

   • If ‘half-gpu’ is not displayed in the gallery, follow the below steps:

     • Click +NEW COMPUTE RESOURCE

     • Enter a name for the compute resource. The name must be unique.

     • Set GPU devices per pod - 1

     • Set GPU memory per device

       • Select % (of device) - Fraction of a GPU device’s memory

       • Set the memory Request - 50 (the workload will allocate 50% of the GPU memory)

     • Optional: set the CPU compute per pod - 0.1 cores (default)

     • Optional: set the CPU memory per pod - 100 MB (default)

     • Click CREATE COMPUTE RESOURCE

   The newly created compute resource will be selected automatically

8. Under Replica autoscaling:

   • Set a minimum of 1 replica and maximum of 2 replicas

   • Set the conditions for creating a new replica to Concurrency (Requests) and the value to 3

   • Set when the replicas should be automatically scaled down to zero to After 5 minutes of inactivity. When automatic scaling to zero is enabled, the minimum number of replicas set in the previous step, automatically changes to 0.

9. Click CREATE INFERENCE

This would start a triton inference server with a maximum of 2 instances, each instance consumes half a GPU.

API

Copy the following command to your terminal. Make sure to update the below parameters. For more details, see Inferences API:

curl -L 'https://<COMPANY-URL>/api/v1/workloads/inferences' \ 
-H 'Content-Type: application/json' \
-H 'Authorization: Bearer <TOKEN>' \
-d '{ 
    "name": "workload-name", 
    "useGivenNameAsPrefix": true,
    "projectId": "<PROJECT-ID>", 
    "clusterId": "<CLUSTER-UUID>", 
    "spec": {
        "image": "runai.jfrog.io/demo/example-triton-server",
        "servingPort": {
            "protocol": "http",
            "container": 8000
        },
        "autoscaling": {
            "minReplicas": 0,
            "maxReplicas": 2,
            "metric": "concurrency",
            "metricThreshold": 3,
            "scaleToZeroRetentionSeconds": 300
        },
        "compute": {
            "cpuCoreRequest": 0.1,
            "gpuRequestType": "portion",
            "cpuMemoryRequest": "100M",
            "gpuDevicesRequest": 1,
            "gpuPortionRequest": 0.5
        }
    }
}'      

• <COMPANY-URL> - The link to the NVIDIA Run:ai user interface

• <TOKEN> - The API access token obtained in Step 1

• <PROJECT-ID> - The ID of the Project the workload is running on. You can get the Project ID via the Get Projects API.

• <CLUSTER-UUID> - The unique identifier of the Cluster. You can get the Cluster UUID via the Get Clusters API.

Note

The above API snippet runs with NVIDIA Run:ai clusters of 2.18 and above only.

Step 3: Querying the Inference Server

In this step, you'll test the deployed model by sending a request to the inference server. To do this, you'll launch a general-purpose workload, typically a Training or Workspace workload, to run the Triton demo client. You'll first retrieve the workload address, which serves as the model’s inference serving endpoint. Then, use the client to send a sample request and verify that the model is responding correctly.

UI

1. Go to the Workload manager → Workloads.

2. Click COLUMNS and select Connections.

3. Select the link under the Connections column for the inference workload created in Step 2

4. In the Connections Associated with Workload form, copy the URL under the Address column

5. Click +NEW WORKLOAD and select Training

6. Select the cluster and project where the inference workload was created

7. Under Workload architecture, select Standard

8. Select Start from scratch to launch a new workload quickly

9. Enter a unique name for the workload. If the name already exists in the project, you will be requested to submit a different name.

10. Click CONTINUE

    In the next step:

11. Create an environment for your workload

    • Click +NEW ENVIRONMENT

    • Enter a name for the environment. The name must be unique.

    • Enter the Image URL - runai.jfrog.io/demo/example-triton-client

    • Set the runtime settings for the environment. Click +COMMAND & ARGUMENTS and add the following:

      • Enter the command: perf_analyzer

      • Enter the arguments: -m inception_graphdef -p 3600000 -u <INFERENCE-ENDPOINT>. Make sure to replace the inference endpoint with the Address you retrieved above.

    • Click CREATE ENVIRONMENT

    The newly created environment will be selected automatically

12. Select the ‘cpu-only’ compute resource for your workspace

    • If ‘cpu-only’ is not displayed in the gallery, follow the below steps:

      • Click +NEW COMPUTE RESOURCE

      • Enter a name for the compute resource. The name must be unique.

      • Set GPU devices per pod - 0

      • Set CPU compute per pod - 0.1 cores

      • Set the CPU memory per pod - 100 MB (default)

      • Click CREATE COMPUTE RESOURCE

    The newly created compute resource will be selected automatically

13. Click CREATE TRAINING

API

Copy the following command to your terminal. Make sure to update the below parameters according. For more details, see Trainings API:

curl -L 'https://<COMPANY-URL>/api/v1/workloads/trainings' \ 
-H 'Content-Type: application/json' \
-H 'Authorization: Bearer <TOKEN>' \ 
-d '{ 
    "name": "workload-name", 
    "projectId": "<PROJECT-ID>", 
    "clusterId": "<CLUSTER-UUID>",
    "spec": {  
        "image": "runai.jfrog.io/demo/example-triton-client",
        "command": "perf_analyzer",
        "args": "-m inception_graphdef -p 3600000 -u <INFERENCE-ENDPOINT>",
        "compute": {
            "cpuCoreRequest":0.1,
            "cpuMemoryRequest": "100M",
        }
    }
}   

• <COMPANY-URL> - The link to the NVIDIA Run:ai user interface

• <TOKEN> - The API access token obtained in Step 1

• <PROJECT-ID> - The ID of the Project the workload is running on. You can get the Project ID via the Get Projects API.

• <CLUSTER-UUID> - The unique identifier of the Cluster. You can get the Cluster UUID via the Get Clusters API.

• <INFERENCE-ENDPOINT> - You can get the inference endpoint from the urls parameter via the Get Workloads API.

Note

The above API snippet runs with NVIDIA Run:ai clusters of 2.18 and above only.

Next Steps

• Select the inference workload you created in Step 2 and go to the Metrics tab to see various GPU and inference metrics graphs rise.

• Manage and monitor your newly created workload using the Workloads table.