Nico Galoppo
Nico Galoppo
Note also that because convnet-benchmarks is using old-style prototxt, the NNPackConvolutionParameter message is not parsed correctly by caffe (e.g. to set algorithm: FFT_16x16)
If the reported timings are per image (not per batch), then I take my comments about not being reproduce back. However, it could be nice to add some notes to...
@Maratyszcza thanks for the link to the `cpu` branch of `convnet-benchmarks`. That's useful. Regarding the convolution algorithm: did you have to convert the prototxt to the new format to pick...
So this is for an Intel(R) Core(TM) i5-6600K CPU @ 3.50GHz. I'm getting pretty good results, in the lines of what your numbers are, except for conv2. I'm not sure...
@Maratyszcza Yes, from the `cpu` branch. I did convert those to the new prototxt format (using `tools/upgrade_net_proto_text` (so that I could change the convolution algorithm without rebuilding), but that shouldn't...
I'd be in favor of this or any other system to have platform-specific linking of files. :+1:
Makes sense, although unlikely for any _existing_ users of homeshick. While I don't disagree that semantic versioning is a useful feature, at this point in time I don't think it...
> I have the linux bashrc in a linux castle and the mac bashrc in a mac castle. The local bashrc is an option for me to have machine local...
> I support changing # to two #'s, just in case someone number files with #'s as pointed out above. I have another system for handling bashrc/bash_profile though, which works...
@iron8core Makes sense!