Wait for model readiness in test script

Author: sd109

charts/azimuth-llm/templates/test/end-to-end.yml

@@ -24,7 +24,6 @@ spec:
         - http://{{ .Values.ui.service.name }}.{{ .Release.Namespace }}.svc
         {{- end }}
       restartPolicy: Never
-  # Allow plenty of retries since downloading
-  # model weights can take a long time.
-  backoffLimit: 20
+  # Handle retries within gradio-test-client script
+  backoffLimit: 1
 {{- end -}}

web-apps/chat/gradio-client-test.py

@@ -1,7 +1,21 @@
 import sys
+import time
+
 from gradio_client import Client
 
 gradio_host = sys.argv[1]
-client = Client(gradio_host)
-result = client.predict("Hi", api_name="/chat")
-print(result)
+
+retries = 60
+for n in range(1, retries+1):
+    try:
+        client = Client(gradio_host)
+        result = client.predict("Hi", api_name="/chat")
+        print(result)
+        break
+    except Exception as err:
+        msg = f"Attempt {n} / {retries} encounter error: {err}"
+        if n < retries:
+            print(msg, "- waiting 10 seconds before retrying")
+            time.sleep(10)
+        else:
+            print(msg, "- no more retries left")