IvyGongoogle commented on issue #14159: [Feature Request] Support fp16 for c++ 
api
URL: 
https://github.com/apache/incubator-mxnet/issues/14159#issuecomment-485865319
 
 
   > @IvyGongoogle Are your inputs and weights in fp16. Your change should work 
to run fp16 inference. What batch size are you using and what is the model ? 
For smaller batch sizes you may not see a big speedup. Also what hardware are 
you running it on ?
   
   
   yes, my inputs and weights is in fp16 and batch_size is 16/32, but after my 
change above, the inference using `src/c_api/c_predict_api.cc` can run, but the 
speed using fp16 still is same with fp32, or slower.  However, as you know, 
when we run `example/image-classification/benchmark_score.py`, using fp16 is 
twice faster than fp32, so I doubt that why python and C++ inference api using 
fp16 have so big difference?

----------------------------------------------------------------
This is an automated message from the Apache Git Service.
To respond to the message, please log on to GitHub and use the
URL above to go to the specific comment.
 
For queries about this service, please contact Infrastructure at:
[email protected]


With regards,
Apache Git Services

Reply via email to