org.bytedeco.pytorch

Class AdaptiveLogSoftmaxWithLossImpl

java.lang.Object

org.bytedeco.javacpp.Pointer

org.bytedeco.pytorch.Module

org.bytedeco.pytorch.AdaptiveLogSoftmaxWithLossImplCloneable

org.bytedeco.pytorch.AdaptiveLogSoftmaxWithLossImpl

All Implemented Interfaces:

AutoCloseable

@Namespace(value="torch::nn")
 @NoOffset
 @Properties(inherit=torch.class)
public class AdaptiveLogSoftmaxWithLossImpl
extends AdaptiveLogSoftmaxWithLossImplCloneable

Efficient softmax approximation as described in Efficient softmax approximation for GPUs_ by Edouard Grave, Armand Joulin, Moustapha Cissé, David Grangier, and Hervé Jégou. See https://pytorch.org/docs/master/nn.html#torch.nn.AdaptiveLogSoftmaxWithLoss to learn about the exact behavior of this module. See the documentation for torch::nn::AdaptiveLogSoftmaxWithLossOptions class to learn what constructor arguments are supported for this module. Example:

  AdaptiveLogSoftmaxWithLoss model(AdaptiveLogSoftmaxWithLossOptions(8, 10,
  {4, 8}).div_value(2.).head_bias(true));
  

Nested Class Summary

Nested classes/interfaces inherited from class org.bytedeco.javacpp.Pointer

Pointer.CustomDeallocator, Pointer.Deallocator, Pointer.NativeDeallocator, Pointer.ReferenceCounter

Field Summary

Fields inherited from class org.bytedeco.javacpp.Pointer

address, capacity, limit, position

Constructor Summary

Constructors

Constructor and Description
AdaptiveLogSoftmaxWithLossImpl(AdaptiveLogSoftmaxWithLossOptions options_)
AdaptiveLogSoftmaxWithLossImpl(long in_features, long n_classes, LongVector cutoffs)
AdaptiveLogSoftmaxWithLossImpl(Module pointer) Downcast constructor.
AdaptiveLogSoftmaxWithLossImpl(Pointer p) Pointer cast constructor.

Method Summary

Modifier and Type	Method and Description
Tensor	_get_full_log_prob(Tensor input, Tensor head_output) Given input tensor, and output of head, computes the log of the full distribution
LongVector	cutoffs() Cutoffs used to assign targets to their buckets.
AdaptiveLogSoftmaxWithLossImpl	cutoffs(LongVector setter)
ASMoutput	forward(Tensor input, Tensor target)
long	head_size() Output size of head classifier
AdaptiveLogSoftmaxWithLossImpl	head_size(long setter)
Tensor	log_prob(Tensor input) Computes log probabilities for all n_classes
long	n_clusters() Number of clusters
AdaptiveLogSoftmaxWithLossImpl	n_clusters(long setter)
AdaptiveLogSoftmaxWithLossOptions	options() The options with which this Module was constructed
AdaptiveLogSoftmaxWithLossImpl	options(AdaptiveLogSoftmaxWithLossOptions setter)
Tensor	predict(Tensor input) This is equivalent to log_pob(input).argmax(1) but is more efficient in some cases
void	pretty_print(Pointer stream) Pretty prints the AdaptiveLogSoftmaxWithLoss module into the given stream.
void	reset_parameters()
void	reset() reset() must perform initialization of all members with reference semantics, most importantly parameters, buffers and submodules.
long	shortlist_size()
AdaptiveLogSoftmaxWithLossImpl	shortlist_size(long setter)

Methods inherited from class org.bytedeco.pytorch.AdaptiveLogSoftmaxWithLossImplCloneable

asModule, asModule, clone, clone

Methods inherited from class org.bytedeco.pytorch.Module

apply, apply, apply, apply, apply, apply, apply, apply, buffers, buffers, children, eval, is_serializable, is_training, load, modules, modules, name, named_buffers, named_buffers, named_children, named_modules, named_modules, named_modules, named_parameters, named_parameters, parameters, parameters, register_buffer, register_buffer, register_module, register_module, register_parameter, register_parameter, register_parameter, register_parameter, save, shiftLeft, to, to, to, train, unregister_module, unregister_module, zero_grad

Methods inherited from class org.bytedeco.javacpp.Pointer

address, asBuffer, asByteBuffer, availablePhysicalBytes, calloc, capacity, capacity, close, deallocate, deallocate, deallocateReferences, deallocator, deallocator, equals, fill, formatBytes, free, getDirectBufferAddress, getPointer, getPointer, getPointer, getPointer, hashCode, interruptDeallocatorThread, isNull, isNull, limit, limit, malloc, maxBytes, maxPhysicalBytes, memchr, memcmp, memcpy, memmove, memset, offsetAddress, offsetof, offsetof, parseBytes, physicalBytes, physicalBytesInaccurate, position, position, put, realloc, referenceCount, releaseReference, retainReference, setNull, sizeof, sizeof, toString, totalBytes, totalCount, totalPhysicalBytes, withDeallocator, zero

Methods inherited from class java.lang.Object

finalize, getClass, notify, notifyAll, wait, wait, wait

Constructor Detail

AdaptiveLogSoftmaxWithLossImpl

public AdaptiveLogSoftmaxWithLossImpl(Pointer p)

Pointer cast constructor. Invokes Pointer(Pointer).

AdaptiveLogSoftmaxWithLossImpl

public AdaptiveLogSoftmaxWithLossImpl(Module pointer)

Downcast constructor.

AdaptiveLogSoftmaxWithLossImpl

public AdaptiveLogSoftmaxWithLossImpl(@Cast(value="int64_t")
                                      long in_features,
                                      @Cast(value="int64_t")
                                      long n_classes,
                                      @ByVal @Cast(value="std::vector<int64_t>*")
                                      LongVector cutoffs)

AdaptiveLogSoftmaxWithLossImpl

public AdaptiveLogSoftmaxWithLossImpl(@ByVal
                                      AdaptiveLogSoftmaxWithLossOptions options_)

Method Detail

forward

@ByVal
public ASMoutput forward(@Const @ByRef
                                Tensor input,
                                @Const @ByRef
                                Tensor target)

reset

public void reset()

Description copied from class: AdaptiveLogSoftmaxWithLossImplCloneable

reset() must perform initialization of all members with reference semantics, most importantly parameters, buffers and submodules.

Overrides:

reset in class AdaptiveLogSoftmaxWithLossImplCloneable

reset_parameters

public void reset_parameters()

pretty_print

public void pretty_print(@Cast(value="std::ostream*") @ByRef
                         Pointer stream)

Pretty prints the AdaptiveLogSoftmaxWithLoss module into the given stream.

Overrides:

pretty_print in class Module

_get_full_log_prob

@ByVal
public Tensor _get_full_log_prob(@Const @ByRef
                                        Tensor input,
                                        @Const @ByRef
                                        Tensor head_output)

Given input tensor, and output of head, computes the log of the full distribution

log_prob

@ByVal
public Tensor log_prob(@Const @ByRef
                              Tensor input)

Computes log probabilities for all n_classes

predict

@ByVal
public Tensor predict(@Const @ByRef
                             Tensor input)

This is equivalent to log_pob(input).argmax(1) but is more efficient in some cases

options

@ByRef
public AdaptiveLogSoftmaxWithLossOptions options()

The options with which this Module was constructed

options

public AdaptiveLogSoftmaxWithLossImpl options(AdaptiveLogSoftmaxWithLossOptions setter)

cutoffs

@ByRef
 @Cast(value="std::vector<int64_t>*")
public LongVector cutoffs()

Cutoffs used to assign targets to their buckets. It should be an ordered Sequence of integers sorted in the increasing order

cutoffs

public AdaptiveLogSoftmaxWithLossImpl cutoffs(LongVector setter)

shortlist_size

@Cast(value="int64_t")
public long shortlist_size()

shortlist_size

public AdaptiveLogSoftmaxWithLossImpl shortlist_size(long setter)

n_clusters

@Cast(value="int64_t")
public long n_clusters()

Number of clusters

n_clusters

public AdaptiveLogSoftmaxWithLossImpl n_clusters(long setter)

head_size

@Cast(value="int64_t")
public long head_size()

Output size of head classifier

head_size

public AdaptiveLogSoftmaxWithLossImpl head_size(long setter)