To check that this works, simply test the pretrained model commands for the pretrained models. I.e., try using the --load-pretrained flag for the pretrained model scripts. It should load the networks and produce good performance.
Essentially, we just change the URL reference to the pretrained model directory on AWS (done internally within the company). This PR uses that new URL reference so that we can reduce references to chainer.
To check that this works, simply test the pretrained model commands for the pretrained models. I.e., try using the
--load-pretrained
flag for the pretrained model scripts. It should load the networks and produce good performance.Essentially, we just change the URL reference to the pretrained model directory on AWS (done internally within the company). This PR uses that new URL reference so that we can reduce references to chainer.