Disable dynamic batching for a TRT model

[le-Greg]

Hi, i have a Triton server with 2 parts :

• A python backend
• A TensorRT model

The python code is basically this :
1- Receive HTTP request
2- Preprocess the input and batch the data
3- Send an internal request to the TensorRT model
4- Postprocess the output
5- Send the answer with HTTP

The python code already takes care of batching, and i don't want Triton's dynamic batching feature to batch multiple requests together. So can i disable dynamic batching for a TRT model, while keeping the model with a flexible batch dimension for the python code to use ?