NVIDIA Clara Train 4.1
1.0

Bring your own Writer

Writers in AIAA is the last stage, it will write the result and send it back to the client. You can bring your own writer just make sure it implements this method: __call__(self, data) and it returns a tuple of (output_filename, result_json).

Below is an example of custom writer:

Copy
Copied!
            

class CustomWriter: def __init__( self, image='pred', label_names=None ): self.image = image self.label_names = label_names def __call__(self, data): result = [] for label in data[self.image]: result.append(self.label_names[int(label)]) # in this writer it does not save things into a file # it just return the prediction result as json return None, {'prediction': result}

Assume you saved this custom transform in a file called custom_writer.py. For AIAA to pick up, you need to copy that file inside <AIAA workspace>/lib folder.

Then you can use it in config_aiaa.json as the following:

Copy
Copied!
            

{ "name": "custom_writer.CustomWriter", "args": { "image": "pred", "label_names": [ "label-0", "label-1", "label-2", "label-3", "label-4" ] } }

Note

Although AIAA supports only one writer, you can still return multiple results by packing them together using custom writer and unpack on the client side.

© Copyright 2020, NVIDIA. Last updated on Feb 2, 2023.