Ben Johnson

Yeah I think I have a reasonable path forward for this sketched out here: https://github.com/cfld/python_essentials Will be updating the build system and integrating better w/ essentials over the next ~...

@cameronshinn I pushed a version that works w/ the `master` branch of `essentials` here: https://github.com/cfld/python_essentials I think it's pretty reasonable, but there are a few lingering things: - [ ]...

Also -- right now a bunch of stuff in `CMakeLists` is hardcoded for my machine -- let me know if you need help getting it running on your machine +...

To follow up -- I ran ``` python linear.py --model_path results/0.0078125_128_128_cifar10_model_50.pth ``` which gives me a Top-1 accuracy of 78.5, so I think that supports my guess above. But --...

For number 2, the point was that you're using pooling w/ stride > 1 in the fixed architecture, but a combination of `_factorized_reduction` and pooling w/ stride = 1 in...

Alright thanks -- I'll give that a shot, but I won't tell anyone... I'm not sure whether Gaussian would be a good approximation in my case -- since there are...

I think the fix is `if contains(cp1, cp2, old[:depth]): continue` Likewise, I think should also change ``` if contains(act, cp1, new): continue ``` to ``` if contains(act, cp1, new[:depth -...

No I actually ended up using `sloth` instead: http://sloth.readthedocs.io/en/latest/first_steps.html I didn't actually have to do the annotation online or in a browser, so that was all I needed. Adding custom...

Looks like the code wants the initial example to be positive, so changing `seed_doc` to something in `np.where(rel_info['GPRO'].values)` (e.g. `1235`) appears to fix the error.