[TenorRT] Predict return bad request 400

[peacealex63]

Hi sir,
I deployed TensorRT on k8s by the following:
trt_model = Model(
name="testtensorrt",
platform=ModelFramework.TensorRT,
local_folder="",
uri="gs://seldon-models/triton/tf_cifar10",
description="An TensorRT model",
protocol=KFServingV2Protocol(),
)
from tempo.serve.metadata import KubernetesOptions
from tempo.seldon.k8s import SeldonCoreOptions
runtime_options = SeldonCoreOptions(
k8s_options=KubernetesOptions(
namespace="production",
authSecretName="minio-secret"
)
)
remote_model = deploy(trt_model, options=runtime_options)

And it deploy Tensorrt image successful, but when I want to do predict by the input:
https://github.com/SeldonIO/seldon-core/blob/master/notebooks/triton_examples.ipynb

payload = (
'{"inputs":[{"name":"input_1","datatype":"FP32","shape":[1, 32, 32, 3],"data":'
+ f"{test_example}"
+ "}]}"
)
print(remote_model.predict(payload))
And it returns error:
raise HTTPError(http_error_msg, response=self)
requests.exceptions.HTTPError: 400 Client Error: Bad Request for url: http://172.18.255.1/seldon/production/testtensorrt/v2/models/testtensorrt/infer

After this I set my payload to only numpy array type, like this:
np_test_example = np.array(test_example)
print(remote_model.predict(np_test_example))

But got the same error, Is my input wrong?
Thanks.