jonatan1626 commented on issue #17159: Performance regression from 1.4.1 to 
1.5.1
URL: 
https://github.com/apache/incubator-mxnet/issues/17159#issuecomment-568782727
 
 
   @apeforest so I think it's using Apache MXNet already? When running the 
script this is printed:
   ```
   EIA context not available, trying GPU...
   GPU not available, trying CPU...
   @@@@@@@@@@@@@@@@@@@@@@@@@@@@@@@@@@@@@@@@@@@@@
   starting test for Model: resnext-50 as float32 with batch size 1
   ```
   
   I reran the benchmark with the above script on c5.18xlarge:
   
   ```
   export OMP_NUM_THREADS=28
   export KMP_AFFINITY=granularity=fine,noduplicates,compact,1,0
   numactl --physcpubind=0-27 --membind=0 python gluon_resnext50.py --iteration 
100000
   ```
   
   **1.4.1**
   Install: `pip install mxnet-mkl==1.4.1`
   Output:
   ```
   Model: resnext-50
        ----Latency----
        Model Partition Time: 0
        Model Load Time: 91.37535095214844
        First Inference: 117.7060604095459
        p99: 16.603946685791016
        p90: 16.51310920715332
        p50: 16.464948654174805
        Avg: 16.465788638039882
        StdDev: 0.1372975856980321
   ```
   
   **1.5**
   Install: `pip install mxnet-mkl==1.5.1`
   Output:
   ```
   Model: resnext-50
        ----Latency----
        Model Partition Time: 0
        Model Load Time: 109.04812812805176
        First Inference: 147.76873588562012
        p99: 33.95581245422363
        p90: 32.033681869506836
        p50: 29.485464096069336
        Avg: 29.450167442438495
        StdDev: 2.0437403857263194
   ```
   
   **1.6**
   Install: `pip install mxnet-mkl==1.6.0b20191025`
   Output:
   ```
   Model: resnext-50
        ----Latency----
        Model Partition Time: 0
        Model Load Time: 110.43977737426758
        First Inference: 139.35613632202148
        p99: 18.41902732849121
        p90: 18.24951171875
        p50: 18.1272029876709
        Avg: 18.141887197731947
        StdDev: 0.13818348880145034
   ```

----------------------------------------------------------------
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