Builder method for creating LSTMs, as in the C++ code.
Represents a "dimension", which you should think of as the dimension of a tensor.
Represents a "dimension", which you should think of as the dimension of a tensor. Can only be constructed using factory methods in the companion object.
Represents an expression on the computation graph.
Represents an expression on the computation graph. Can only be constructed using the functions contained in the companion object.
For the most part, SWIG makes it so you just use Scala classes where you would use C++ pointers to class instances.
For the most part, SWIG makes it so you just use Scala classes where you would use C++ pointers to class instances. The exception is for pointers to primitives. The SWIG wrappers for those are not Scala-like at all, so here are wrapper that are easier to work with.
Builder method for creating GRUs, as in the C++ code.
Builder method for creating GRUs, as in the C++ code. For its public methods see edu.cmu.dynet.RnnBuilder.
New serialization, much less featureful than old serialization.
A (learnable) parameter of a model.
Specifies how to initialize a parameter.
Specifies how to initialize a parameter. Construct a particular initialization using the various factory methods on the companion object.
The underlying storage for a model parameter.
The underlying storage for a model parameter. You will never need to construct this yourself, but can get it back from edu.cmu.dynet.ParameterCollection.parametersList().
DyNet operates on Tensor
objects.
DyNet operates on Tensor
objects. You should never need to construct them directly;
however, here are some methods for getting values out of them.
Interface for edu.cmu.dynet.ParameterCollection trainers.
Interface for edu.cmu.dynet.ParameterCollection trainers. You want to use a specific subclass.
SWIG converts C++ unsigned
to Scala Long
The ComputationGraph object contains the singleton DyNet computation graph instance.
The ComputationGraph object contains the singleton DyNet computation graph instance. Any C++ instance method is instead implemented as a static function here.*
Factory for edu.cmu.dynet.Dim instances.
Contains methods for creating edu.cmu.dynet.Expressions.
Contains methods for creating edu.cmu.dynet.Expressions. There are several ways to create expressions:
* from explicit values (e.g. input
)
* randomly (e.g. randomNormal
)
* from edu.cmu.dynet.ParameterCollection parameters (e.g. parameter
)
* from other expressions (e.g. softmax
and pow
)
Contains methods for initializing DyNet
Helpers that don't fit anywhere else but that might make your life easier
Builder method for creating LSTMs, as in the C++ code. For its public methods see edu.cmu.dynet.RnnBuilder.