Just a placeholder TODO

spatial_batchnorm_backward working. next up: train final good model

redid spatial_batchnorm_forward more elegantly

spatial_batchnorm_forward works now

3Layer network confirmed working now. Next up: spatial batchnorm

Got the 3 layer working. Need to validate one epoch gets me >40% accuracy. if it does, we're done

Skipping the naive backward impls, and marking as such via asserts that always fail.

Got naive max pool layer working

convolve over all channels at once

move the numpy.pad operation out of the inner loops

conv_forward_naive finished (super naive, all for loops, no matrix cleverness)

checkpoint. super naive conv fwd pass working

Finished FullyConnectedNets.ipynb

got 48.4% accuracy, using the middle learning rate, on the final exercise

Removed a debug message, all that's left now is the last bit "Train a good model!"