Calculate accuracy
Manage multiple evaluation metrics.
Constructing a context which is used to specify the device and device type that will be utilized by the engine.
Custom evaluation metric that takes a NDArray function.
Base class for operators implemented in Scala
Base class for operator property class implemented in Scala.
class batch of data
DataIter object in mxnet.
pack of DataIter, use as Iterable class
Base class of all evaluation metrics
Symbolic Executor component of MXNet
WARNING: it is your responsibility to clear this object through dispose().
Symbolic Executor component of MXNet
WARNING: it is your responsibility to clear this object through dispose().
Symbol.bind : to create executor
Calculate the F1 score of a binary classification problem.
Class for reducing learning rate in factor
Class for reducing learning rate in factor
Assume the weight has been updated by n times, then the learning rate will be base_lr * factor^^(floor(n/step))
Model class of MXNet for training and predicting feedforward nets.
Model class of MXNet for training and predicting feedforward nets. This class is designed for a single-data single output supervised network.
Base class for Initializer.
Learning rate scheduler, which adaptively changes the learning rate based on the training progress.
Calculate Mean Absolute Error loss
Scala interface for read/write RecordIO data formmat with index.
Scala interface for read/write RecordIO data formmat with index. Support random access.
Scala interface for read/write RecordIO data format
Initialize the weight with mixed Initializer
Describe the model flow
Monitor outputs, weights, and gradients for debugging.
NDArray object in mxnet.
NDArray object in mxnet.
NDArray is basic ndarray/Tensor like data structure in mxnet.
NOTE: NDArray is stored in native memory. Use NDArray in a try-with-resources() construct
or a [[org.apache.mxnet.ResourceScope]] in a try-with-resource to have them
automatically disposed. You can explicitly control the lifetime of NDArray
by calling dispose manually. Failure to do this will result in leaking native memory.
NameManager to do automatic naming.
NameManager to do automatic naming. User can also inherit this object to change naming behavior.
Initialize the weight with normal(0, sigma)
Calculate perplexity.
Calculate Root Mean Squred Error loss
This class manages automatically releasing of org.apache.mxnet.NativeResource
s
This class allow you to write cuda kernel in Scala and call them with NDArray.
Shape of NDArray or other data
Sparse NDArray is the child class of NDArray designed to hold the Sparse format
Sparse NDArray is the child class of NDArray designed to hold the Sparse format
Currently, Rowsparse and CSR typed NDArray is supported. Most of the Operators
will convert Sparse NDArray to dense. Basic operators like add
will
have optimization for sparse operattions
Symbolic configuration API of mxnet.
Symbolic configuration API of mxnet.
WARNING: it is your responsibility to clear this object through dispose().
Calculate top k predictions accuracy
Initialize the weight with uniform [-scale, scale]
Initialize the weight with Xavier or similar initialization scheme.
(Since version 1.5.0) Please use ResourceScope instead
Callback functions that can be used to track various status during epoch.
IO iterators for loading training & validation data
Image API of Scala package enable OpenCV feature
Key value store interface of MXNet for parameter synchronization.
Layout definition of DataDesc N Batch size C channels H Height W Weight T sequence length undefined default value of Layout
NDArray Object extends from NDArrayBase for abstract function signatures Main code will be generated during compile time through Macros
NDArray Object extends from NDArrayBase for abstract function signatures Main code will be generated during compile time through Macros
typesafe NDArray API: NDArray.api._ Main code will be generated during compile time through Macros
typesafe NDArray API: NDArray.api._ Main code will be generated during compile time through Macros
typesafe NDArray random module: NDArray.random._ Main code will be generated during compile time through Macros
typesafe NDArray random module: NDArray.random._ Main code will be generated during compile time through Macros
NumpyScope object provides util functions for turning on/off NumPy compatibility and checking whether NumPy compatibility has been turned on/off.
NumpyScope object provides util functions for turning on/off NumPy compatibility and checking whether NumPy compatibility has been turned on/off. NumPy compatibility is introduced first to support zero-dim and zero-size tensors as in NumPy.
Random Number interface of mxnet.
Symbol Object extends from SymbolBase for abstract function signatures Main code will be generated during compile time through Macros
Symbol Object extends from SymbolBase for abstract function signatures Main code will be generated during compile time through Macros
typesafe Symbol API: Symbol.api._ Main code will be generated during compile time through Macros
typesafe Symbol API: Symbol.api._ Main code will be generated during compile time through Macros
typesafe Symbol random module: Symbol.random._ Main code will be generated during compile time through Macros
typesafe Symbol random module: Symbol.random._ Main code will be generated during compile time through Macros
A collector to store NDArrays.
A collector to store NDArrays.
It provides a scope, NDArrays allocated in the scope can either
val a = NDArray.array(Array(-1f, 0f, 1f, 2f, 3f, 4f), shape = Shape(2, 3)) val res = NDArrayCollector.auto().withScope { (NDArray.relu(a) + a).toArray }In the case above, the intermediate NDArrays (created by NDArray.relu and +) will be disposed automatically.
val collector = NDArrayCollector.manual() val res = collector.withScope { (NDArray.relu(a) + a).toArray } collector.foreach(_.dispose())For Java users:
NDArray a = NDArray.array(new float[]{-1f, 0f, 1f, 2f, 3f, 4f}, Shape.create(2, 3), Context.cpu(0)); float[] sliced = NDArrayCollector.auto().withScope( new scala.runtime.AbstractFunction0() { @Override public float[] apply() { a.slice(0, 1).toArray(); } });
(Since version 1.5.0) Please use ResourceScope instead
Base class for operator property class implemented in Scala. MXNET_CPU_WORKER_NTHREADS must be greater than 1 for custom op to work on CPU