pengzhao-intel commented on issue #14713: [WIP]MKLDNN RNN Inference Integration(fp32 LSTM and vRNN with tanh and relu) URL: https://github.com/apache/incubator-mxnet/pull/14713#issuecomment-493816826 > I re-evaluated the performance with MKLDNN commit 41bee20d7eb4a67feeeeb8d597b3598994eb1959 on local Skylake-8180 with 1 Sockets and 28 cores. Here is the result. > > ### Layer=1 bidirectional = False > API MKLDNN = 0 (sample/sec) MKLDNN = 1(sample/sec) speedup > FusedLSTM 289 682 2.36x > FusedvRNN with tanh 733 1408 1.43x > FusedvRNN with relu 1074 1375 1.28x > ### Layer=5 bidirectional = True > API MKLDNN = 0 (sample/sec) MKLDNN = 1(sample/sec) speedup > FusedLSTM 28 65 2.32x > FusedvRNN with tanh 70 149 2.13x > FusedvRNN with relu 94 147 1.56x Looks good! Could you summarize all combinations covered by this PR and test it?
---------------------------------------------------------------- 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
