Aapo Kyrola b5c053b1c4 fix fp16 issues with resnet trainer ... Summary: My commit bab5bc broke things wiht fp16 compute, as i had tested it only with the null-input, that actually produced fp32 data (even dtype was given as float16). Also, I had confused the concepts of "float16 compute" and fp16 data. Issue #1408. This fixes those issues, tested with both Volta and M40 GPUs. Basically restored much of the previous code and fixed the null input to do FloatToHalf. Reviewed By: pietern Differential Revision: D6211849 fbshipit-source-id: 5b41cffdd605f61a438a4c34c56972ede9eee28e		2017-11-01 13:30:08 -07:00
..
seq2seq	simplify beam search code	2017-10-31 16:59:20 -07:00
__sym_init__.py	Re-license to Apache	2017-09-28 16:22:00 -07:00
download.py	Re-license to Apache	2017-09-28 16:22:00 -07:00
resnet.py	fix fp16 issues with resnet trainer	2017-11-01 13:30:08 -07:00
resnet_test.py	Re-license to Apache	2017-09-28 16:22:00 -07:00