kerasmxnet-run.log

kerasmxnet-run.log

Using MXNet backend
Hyper parameters: {'lr': '0.01', 'batch_size': '256', 'epochs': '10', 'gpus': '2'}
Input parameters: {'validation': {'S3DistributionType': 'FullyReplicated', 'TrainingInputMode': 'File', 'RecordWrapperType': 'None'}, 'training': {'S3DistributionType': 'FullyReplicated', 'TrainingInputMode': 'File', 'RecordWrapperType': 'None'}}
Files loaded
x_train shape: (60000, 1, 28, 28)
60000 train samples
10000 test samples
Train on 60000 samples, validate on 10000 samples
Epoch 1/10
/usr/local/lib/python3.5/dist-packages/mxnet/module/bucketing_module.py:408: UserWarning: Optimizer created manually outside Module but rescale_grad is not normalized to 1.0/batch_size/num_workers (1.0 vs. 0.00390625). Is this intended?
  force_init=force_init)
[17:43:09] src/operator/nn/./cudnn/./cudnn_algoreg-inl.h:107: Running performance tests to find the best convolution algorithm, this can take a while... (setting env variable MXNET_CUDNN_AUTOTUNE_DEFAULT to 0 to disable)

  256/60000 [..............................] - ETA: 35:17 - loss: 2.3051 - acc: 0.0938
 2048/60000 [>.............................] - ETA: 4:18 - loss: 1.8998 - acc: 0.3926 

<output removed>

56320/60000 [===========================>..] - ETA: 0s - loss: 0.0284 - acc: 0.9905
58624/60000 [============================>.] - ETA: 0s - loss: 0.0284 - acc: 0.9905
60000/60000 [==============================] - 2s 26us/step - loss: 0.0283 - acc: 0.9905 - val_loss: 0.0257 - val_acc: 0.9916

Test loss: 0.025707698954685294
Test accuracy: 0.9916
Saved Keras model
MXNet Backend: Successfully exported the model as MXNet model!
MXNet symbol file -  /opt/ml/model/mnist-cnn-10-symbol.json
MXNet params file -  /opt/ml/model/mnist-cnn-10-0000.params

Model input data_names and data_shapes are: 
data_names :  ['/conv2d_1_input1']
data_shapes :  [DataDesc[/conv2d_1_input1,(256, 1, 28, 28),float32,NCHW]]

Note: In the above data_shapes, the first dimension represent the batch_size used for model training. 
You can change the batch_size for binding the module based on your inference batch_size.
Saved MXNet model

===== Job Complete =====
Billable seconds: 121