Now anyone can train Imagenet in 18 minutes https://t.co/xQh0mYqELH
— Jeremy Howard (@jeremyphoward) August 10, 2018
The result of our collaboration of @fastdotai and @yaroslavvb from @DIU_x.
Now anyone can train Imagenet in 18 minutes https://t.co/xQh0mYqELH
— Jeremy Howard (@jeremyphoward) August 10, 2018
The result of our collaboration of @fastdotai and @yaroslavvb from @DIU_x.
New, record-breaking speed results on ImageNet. Previous record was 30 mins using a TPU-pod by Google; now down to 18 mins on AWS by https://t.co/ktYtgBpxGr team! https://t.co/bEIMHOBn6M
— Rachel Thomas (@math_rachel) August 10, 2018
ICYMI, below are the https://t.co/ktYtgBpxGr results from April & post about our techniques.
— Rachel Thomas (@math_rachel) August 10, 2018
Post with today's results here: https://t.co/7pdbY1fBCF https://t.co/w736LtRZaK
40% faster than Google’s record on their proprietary TPU Pod cluster. https://t.co/nAnaYmMbCB uses the same number of processing units as Google’s benchmark (128) and costs around $40 to run. https://t.co/SJKKAm0Tmb
— Oren Etzioni (@etzioni) August 11, 2018