marcoabreu commented on issue #10433: [MXNET-290] MKLDNN support for model
quantization
URL: https://github.com/apache/incubator-mxnet/pull/10433#issuecomment-396297785
I'd welcome it if you could add a method to the C-API that provides context
information about the compile options. E.g.
marcoabreu commented on issue #10433: [MXNET-290] MKLDNN support for model
quantization
URL: https://github.com/apache/incubator-mxnet/pull/10433#issuecomment-395128502
I think we should have the following flow:
- MKL-DNN enabled: MKL-OP -> Native-OP -> Fail
- MKL-DNN disabled:
marcoabreu commented on issue #10433: [MXNET-290] MKLDNN support for model
quantization
URL: https://github.com/apache/incubator-mxnet/pull/10433#issuecomment-395083922
Thanks for elaborating, seems like we encountered a problem. This sounds
very wrong to me. Like I said previously, we
marcoabreu commented on issue #10433: [MXNET-290] MKLDNN support for model
quantization
URL: https://github.com/apache/incubator-mxnet/pull/10433#issuecomment-394865292
I will review it tomorrow.
Xinyu Chen schrieb am Di., 5. Juni 2018, 23:20:
> @marcoabreu
marcoabreu commented on issue #10433: [MXNET-290] MKLDNN support for model
quantization
URL: https://github.com/apache/incubator-mxnet/pull/10433#issuecomment-382426955
@reminisce could you please review this PR?
This is an