ZhennanQin commented on issue #12530: Implement mkldnn convolution fusion and
quantization.
URL: https://github.com/apache/incubator-mxnet/pull/12530#issuecomment-428053235
> Looks like some tests are failing
@eric-haibin-lin Fixed.
ZhennanQin commented on issue #12530: Implement mkldnn convolution fusion and
quantization.
URL: https://github.com/apache/incubator-mxnet/pull/12530#issuecomment-426123379
@marcoabreu Ping again... As you requested changes to this PR, can you
review again to list if any concern isn't
ZhennanQin commented on issue #12530: Implement mkldnn convolution fusion and
quantization.
URL: https://github.com/apache/incubator-mxnet/pull/12530#issuecomment-426122682
@zheng-da Based on my knowledge, merging this PR won't cause following CI
failed. The reason that travis timeout is
ZhennanQin commented on issue #12530: Implement mkldnn convolution fusion and
quantization.
URL: https://github.com/apache/incubator-mxnet/pull/12530#issuecomment-424637291
@marcoabreu Can you review again to see if the change resolve your concern?
Thanks.
ZhennanQin commented on issue #12530: Implement mkldnn convolution fusion and
quantization.
URL: https://github.com/apache/incubator-mxnet/pull/12530#issuecomment-424175202
@marcoabreu Because for original convolution op, the newly created params
doesn't be used. And for _sg_mkldnn_conv,
ZhennanQin commented on issue #12530: Implement mkldnn convolution fusion and
quantization.
URL: https://github.com/apache/incubator-mxnet/pull/12530#issuecomment-420857295
@zheng-da There's strong dependence between quantization and fusion because
they all changed mkldnn_conv.cc, which
