You are viewing a plain text version of this content. The canonical link for it is here.
Posted to commits@mxnet.apache.org by GitBox <gi...@apache.org> on 2018/11/13 02:25:40 UTC

[GitHub] roywei opened a new issue #13243: CPP examples training acc does not increase

roywei opened a new issue #13243: CPP examples training acc does not increase
URL: https://github.com/apache/incubator-mxnet/issues/13243
 
 
   
   ## Description
   I found this problem in PR: https://github.com/apache/incubator-mxnet/pull/13185
   Following examples here:
   https://github.com/apache/incubator-mxnet/tree/master/cpp-package/example
   
   During traing, the acc actually does not change at all.
   
   Right now affected models:
   1. inception-bn
   2. resnet
   
   will continue to check other models
   
   
   ## Error Message:
   ```
   [02:17:51] src/io/iter_mnist.cc:110: MNISTIter: load 60000 images, shuffle=1, shape=(50,784)
   [02:17:52] src/io/iter_mnist.cc:110: MNISTIter: load 10000 images, shuffle=1, shape=(50,784)
   [02:17:52] resnet.cpp:199: Epoch: 0
   [02:18:06] resnet.cpp:219: Train Accuracy: 0.0987167
   [02:18:08] resnet.cpp:230: Validation Accuracy: 0.098
   [02:18:08] resnet.cpp:199: Epoch: 1
   [02:18:22] resnet.cpp:219: Train Accuracy: 0.0987167
   [02:18:24] resnet.cpp:230: Validation Accuracy: 0.098
   [02:18:24] resnet.cpp:199: Epoch: 2
   [02:18:38] resnet.cpp:219: Train Accuracy: 0.0987167
   [02:18:40] resnet.cpp:230: Validation Accuracy: 0.098
   [02:18:40] resnet.cpp:199: Epoch: 3
   [02:18:54] resnet.cpp:219: Train Accuracy: 0.0987167
   [02:18:56] resnet.cpp:230: Validation Accuracy: 0.098
   
   ```
   ## Minimum reproducible example
   (If you are using your own code, please provide a short script that reproduces the error. Otherwise, please provide link to the existing example.)
   
   ## Steps to reproduce
   follow:
   https://github.com/apache/incubator-mxnet/tree/master/cpp-package/example
   run
   ```
   ./resnet
   ```
   
   ## What have you tried to solve it?
   1. changing learning rate and weight decay does not help.
   2. changed to `sgd` optimizer instead of deprecated `ccsgd` does not help.
   3. used same setup in python version, python works fine acc increased every epoch, cpp does not.
   

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


With regards,
Apache Git Services