needs to be checked if PE model accepts batches (if so - maybe worth claiming use of batches of inputs)

needs to be checked if PE model accepts batches (if so - maybe worth claiming use of batches of inputs)

Originally posted by @PawelPeczek-Roboflow in https://github.com/roboflow/inference/pull/1350#discussion_r2138408406

looks like it does:

• https://github.com/roboflow/inference/blob/main/inference/core/entities/requests/perception_encoder.py#L45

Clip also does actually and block decalres input the same way:
data: Union[Selector(kind=[IMAGE_KIND, STRING_KIND]), str]

@PawelPeczek-Roboflow Is this the only thing that would need to be added?

@classmethod
def get_parameters_accepting_batches(cls) -> List[str]:
    return ["data"]

(above results in 'Batch' object has no attribute 'to_inference_format')

Start with inference/core/entities/requests/perception_encoder.py around line 45 and inspect how the PE model declares batch-capable parameters. Reproduce the reported 'Batch' object has no attribute 'to_inference_format' error, then verify that batched inputs are accepted and processed successfully. Confirm whether the required change is limited to the PE model declaration.