1
0
mirror of https://github.com/openshift/openshift-docs.git synced 2026-02-05 12:46:18 +01:00
Files
openshift-docs/modules/microshift-rhoai-query-model.adoc
2025-10-07 17:43:41 +00:00

78 lines
2.0 KiB
Plaintext

// Module included in the following assemblies:
//
// * microshift_ai/microshift-rhoai.adoc
:_mod-docs-content-type: PROCEDURE
[id="microshift-rhoai-query-model_{context}"]
= Querying your AI model
Make an inference request against the AI model server that is using the `ovms-resnet50` model.
.Prerequisites
* {microshift-short} is running.
* You configured the model-serving runtime.
* You uploaded your AI model to {microshift-short}.
.Procedure
* Make an inference request against the model server that is using the `ovms-resnet50` model by running the following command:
+
[source,terminal]
----
$ curl \
--data-binary "@./request.json" \
--header "Inference-Header-Content-Length: ${HEADER_LEN}" \
"${DOMAIN}/v2/models/ovms-resnet50/infer" \
--connect-to "${DOMAIN}::${IP}:" > response.json
----
+
.Example inferencing output, saved to a `response.json`
+
[source,json]
----
{
"model_name": "ovms-resnet50",
"model_version": "1",
"outputs": [{
"name": "1463",
"shape": [1, 1000],
"datatype": "FP32",
"data": [ ....... ] <1>
}]
}
----
<1> The contents of `.outputs[0].data` were omitted from the example for brevity.
.Verification
. To determine the model's prediction, get the index of the highest element in the `.outputs[0].data` to determine the model's predicted value by using the following Python script:
+
[source,python]
----
import json
with open('response.json') as f:
response = json.load(f)
data = response["outputs"][0]["data"]
argmax = data.index(max(data))
print(argmax)
----
+
.Example output
[source,text]
----
309 <1>
----
<1> In this example, the element labeled `309` is the model's response.
. Validate the output against link:https://github.com/openvinotoolkit/model_server/blob/main/client/common/resnet_input_images.txt[resnet's input data], for example:
+
[source,text]
----
../../../../demos/common/static/images/bee.jpeg 309
----
.Next steps
* Optional. Query the AI model using other images available in the resnet input data.