I've noticed that this pattern arises in other models as well. Can we put this post processing part inside utils folder e.g. ghostnet/utils/utils.py where we would have method something like this:

def post_processing(top_k, url, file_name):
    # Post processing
    probabilities = torch.nn.functional.softmax(output[0][0], dim=0)
    urllib.request.urlretrieve(url, file_name)

    with open(file_name, "r") as f:
        categories = [s.strip() for s in f.readlines()]
    topk_prob, topk_catid = torch.topk(probabilities, top_k)
    for i in range(topk_prob.size(0)):
        print(categories[topkcatid[i]], topk_prob[i].item())

Also let's do this for other models as well :)

Initialization of all mobilenet_vx models is almost the same, can we also put that function in .../mobilenet/utils/ folder so we can reuse it for other mobilenets as well.

Please pass the number of top k categories as a param to the function (where default is 5) because we might not always want the top 5 categories. Apply this change in other utils as well :)

Please rename the top5_prob and top5_catid to topk_prob and topk_catid in this and other files as well.

Pls rename back to variants to maintain the standard format

same here too

same here

Please cleanup downloaded imagenet_classes.txt file after successful post processing

Please cleanup downloaded image file after successful processing. Also pls update for other models too