What should be here?

[rwightman]

I think this could be an amazing resource once it gets going. I do have one rather broad an open ended question. What, exactly, is supposed to be published here, where is the bar intended to be in terms of admission?

Is the intent that anyone and everyone publishes their models here? 5 copies of EfficientNet, a ResNet-50 with better trained weights than Torchvision? Or is the intent to curate a list of the best, unique models from research community?

What is the intended difference between the 'research' class and the, as of yet unused, 'development' ... ResNet-50 from Torchvision is labeled as 'research', where is that line supposed to be?

Is there intended to be any standard on performance of model and the weights? e.g. must meet or exceed published paper results in top1 accuracy, mAP, etc.

[soumith]

Good Question.

We have clarity on most of the answer, and are refining some of it.

What should be in the research category?

• artifacts of a published or an arxiv paper (or something of a similar nature that serves a different audience -- such as ULMFit) that a large audience would need
• reproduces the published results (or better)
• aimed at researchers either trying to reproduce a baseline, or trying to build downstream research on top of the model (such as feature-extraction or fine-tuning) (and)
• aimed at researchers looking for a demo of the paper for subjective evaluation
• (debating) are not better off as libraries
  • there are classes of algorithms that are actually better served as libraries than models, because they often need modification for further research or they don't fit well as an isolated piece

Is the intent that anyone and everyone publishes their models here? 5 copies of EfficientNet, a ResNet-50 with better trained weights than Torchvision? Or is the intent to curate a list of the best, unique models from research community?

The initial intent is curation, while we build a better UI that has a voting and class system. So, popular and official models, mainly -- unless there's a case to be made otherwise. Once we have a UI that allows anyone and everyone to freely publish, while not hurting the trust of downstream users, I think the gates can open up.

[nikolz0]

Добрый день, Ошибка в Вашем примере

[NicolasHug]

@nikolz0 the numpy package seems to be missing, you can install it with pip install numpy or conda install numpy.

[nikolz0]

Спасибо за ответ. Проблема решилась обновлением пакета numpy.