The beginning training epoch.
Number of batches in a epoch.
Fit the model.
The evaluation metric, cannot be null
A string kvstore type: 'local' : multi-devices on a single machine, will automatically choose one from 'local_update_cpu', 'local_allreduce_cpu', and 'local_allreduce_device' 'dist_sync' : multi-machines with BSP 'dist_async' : multi-machines with partical asynchronous In default uses 'local', often no need to change for single machine.
A callback that is invoked at end of each epoch. This can be used to checkpoint model each epoch.
A callback that is invoked at end of each batch For print purpose
When not specified, default logger will be used.
The list of work load for different devices, in the same order as ctx
Run the prediction, always only use one device.
the number of batch to run. Go though all batches if set -1
The predicted value of the output. Note the network may have multiple outputs, thus it return an array of NDArray
Checkpoint the model checkpoint into file.
Prefix of model name.
will be saved for symbol.
will be saved for parameters.
FeedForward.load : the method to load the model back.
Serialize the model to Java byte array
serialized model bytes