A couple of pretrained models?

[317070]

I have Lasagne versions of vgg-16, vgg-19 and googlenet with their pretrained parameters: http://www.vlfeat.org/matconvnet/pretrained/ https://github.com/BVLC/caffe/tree/master/models/bvlc_googlenet

Would there be interest in adding these recipes? I reckon these recipes are more the microwave lasagne kind, but this might interest people too?

[f0k]

Would there be interest in adding these recipes?

I guess so, yes! In what form would these recipes be? Ideally, I guess they'd construct the network in Python code, download and read the original model files (e.g., in matlab format for the first web site) and use set_all_param_values() to set the parameters to the correct state. For demonstration purposes, they could compile a prediction function. Not sure where to put such recipes, though... they could live in the "Examples" category or be given their own (something like a model zoo, although model fridge or model freezer might be more appropriate here).

[dnouri]

I'm wondering if these should really be recipes. Maybe a small add-on library is better suited? Just a serving suggestion.

[ebenolson]

Since pretrained models are nearly all we have in Recipes so far, I think it's a pretty good fit :)

But maybe creating a new top level directory for models would make sense.

[benanne]

Yeah, I think we should also keep them here, this is where people will come looking for them. A separate subdirectory for pre-trained models is not a bad idea.

[f0k]

A separate subdirectory for pre-trained models is not a bad idea.

I'm still sad we didn't go for food-related directory names: starters (tutorials), mains (papers), sides (snippets), and now takeaway (premade models).

[benanne]

there's no guarantee that this (remarkably well working) analogy will keep being productive though :p