#0: Add pre- and post- processing to DLA demo

[milank94]

Adding pre- and post-processing functions for the DLA demo.

The model task used in the demo is for image classification (as trained on the ImageNet dataset) following the guide from: https://github.com/ucbdrive/dla/blob/master/classify.py

@nvukobratTT Important note: The original demo used input shape of (1, 3, 384, 1280) but for the image classification task the input should be (1, 3, 224, 224) as defined by the trained data. All variants need to be tested.

For extracting the pre and post processing steps, these were the references used:

• https://github.com/tenstorrent/tt-buda-demos/blob/main/model_demos/cv_demos/densenet/pytorch_densenet.py#L16
• https://github.com/tenstorrent/tt-buda-demos/blob/main/model_demos/cv_demos/resnet/pytorch_resnet.py#L38
• https://github.com/ucbdrive/dla/blob/master/classify.py#L152

[nvukobratTT]

@ashokkumarkannan1 before we merge this PR, can you modify internal model tests and check out if they're passing successfully? If yes, please let me know and we can push these changes on both pybuda and tt-buda-demos repo :))

[ashokkumarkannan1]

Sure @nvukobratTT . I will look into it and get back.

[ashokkumarkannan1]

Hi @nvukobratTT , I have verified the shapes and its working fine locally.

[nvukobratTT]

Hi @nvukobratTT , I have verified the shapes, and it's working fine locally.

Great, thanks for confirmation! I'm going to approve and merge this change in track_pybuda branch, as we have followup changes that are waiting for it :))