Hi

I ran into some error regarding applying Batch Prediction on Vertex-AI which I haven't resolved this issue so far.
In my case I have 2 models, I built the MLP model using Tenforflow 11 and GradientBoost Decision Forest on Tenforflow.

I cannot use batch prediction that use BigQuery as a source to feed into the TensorFlow model to make prediction*
BigQuery is a simple format like row x column (tabular data) but TensorFlow is more complex like data in an array or nested array.

It seems like Batch Prediction is unable to convert data from Bigquery to be proper format for the TensorFlow model.

This is an error in each row

Post request fails. Cannot get predictions. Error: Exceeded retries: Non-OK result 400 ({
"error": "instances is a plain list, but expecting list of objects as multiple input tensors required as per tensorinfo_map"
}) from server, retry=3, ellapsed=0.01s.

I saw in Batch Prediction Youtube and batch_prediction.ipynb
As code below, what kind of framework did you use to build the model?

# Get a model that will make a batch prediction
model_id = 'projects/268076997885/locations/europe-west1/models/8895049068707840000'
model = aiplatform.Model(model_id)

This is valid format that I create batch job prediction choosing Google Cloud storage as input via ConsoleUI
I have to create incident_inputs_tensor_batch.jsonl as input in Google Cloud Storage to enable me to put this file into the model to make a prediction successfully using Bach Prediction.

{"sla": ["24x7 6Hrs Resolution Time"], "product_type": ["Software"], "brand": ["SAPB1"], "service_type": ["Incident"], "incident_type": ["Software"], "open_to_close_hour": [0.15], "response_to_resolved_hour": [0.133333333]}
{"sla": ["24x7 4Hrs Response Time"], "product_type": ["Server"], "brand": ["Cisco"], "service_type": ["Incident"], "incident_type": ["Fan Failure"], "open_to_close_hour": [2137.5], "response_to_resolved_hour": [2016.883333]}
{"sla": ["24x7 4Hrs Resolution Time"], "product_type": ["Firewall"], "brand": ["CheckPoint"], "service_type": ["Request"], "incident_type": ["General Incident"], "open_to_close_hour": [24.48333333], "response_to_resolved_hour": [24.01666667]}

But this is what Google Document as like Get batch predictions from a custom trained model mentions that it is valid format from Bigquery source to All other containers except pytorch-containers , But it is different from Tensorflow dataset

[ [1.0,3.0,"cat1"], [2.0,4.0,"cat2"] ]

This link is the main reference that I use for Batch Prediction
https://cloud.google.com/vertex-ai/docs/predictions/get-batch-predictions#bigquery

The below are step to set batch prediction on console.

data input on Bigquery

I highligh as green to indicate features and label.

these are steps to create job.