"model_file_name": model_file_name,
"check_status_only": False,
"model_class_name": model_class_name,
"model_class_file": model_class_file,
"endpoint_url": s3_url,
"access_key_id": s3_username,
"secret_access_key": s3_password,
"training_results_bucket": bucket_name,
"training_id": model_id,
}
# Deploy model with Knative route.
if cleanup:
if local_cluster_deployment:
# Using K8s api
metrics = run_safe(formData, "DELETE")
else:
response = requests.delete(kfserving_url, params=formData)
metrics = response.json()
print("Successfully cleanup old deployments")
else:
if local_cluster_deployment:
# Using K8s api
metrics = run_safe(formData, "POST")
else:
response = requests.post(kfserving_url, json=formData)
metrics = response.json()
# Print out the necessary endpoints and debugging outputs.
metrics[
'Prediction_Host'] = deployment_name + "." + namespace + "." + knative_custom_domain
f.close()
model_id = args.model_id
deployment_name = args.deployment_name
model_class_name = args.model_class_name
model_class_file = args.model_class_file
serving_image = args.serving_image
formData = {
"public_ip": seldon_ip,
"aws_endpoint_url": s3_url,
"aws_access_key_id": s3_access_key_id,
"aws_secret_access_key": s3_secret_access_key,
"training_results_bucket": bucket_name,
"model_file_name": "model.pt",
"deployment_name": deployment_name,
"training_id": model_id,
"container_image": serving_image,
"check_status_only": False,
"model_class_name": model_class_name,
"model_class_file": model_class_file
}
metrics = run_safe(formData, "POST")
print(metrics)
with open('/tmp/deployment_result.txt', "w") as report:
report.write(json.dumps(metrics))
print('\nThe Model is running at ' + metrics['deployment_url'])
