Mila
Deep Neural Network Library
Loading...
Searching...
No Matches
Mila::Dnn::GptTransformer< TDeviceType, TPrecision > Member List

This is the complete list of members for Mila::Dnn::GptTransformer< TDeviceType, TPrecision >, including all inherited members.

addComponent(ComponentPtr component)Mila::Dnn::CompositeComponent< TDeviceType, TPrecision >inline
backward(const TokenIndexType &input, const TensorType &output_grad) override (defined in Mila::Dnn::GptTransformer< TDeviceType, TPrecision >)Mila::Dnn::GptTransformer< TDeviceType, TPrecision >inline
Mila::Dnn::LanguageNetwork::backward(const TokenIndexType &input, const TensorType &output_grad)=0Mila::Dnn::LanguageNetwork< TDeviceType, TPrecision >pure virtual
build(const BuildContext &context) finalMila::Dnn::Component< TDeviceType, TPrecision >inlinevirtual
build_context_Mila::Dnn::Component< TDeviceType, TPrecision >protected
buildSequenceLengthFromArchive(const ModelArchive &archive)Mila::Dnn::GptTransformer< TDeviceType, TPrecision >inlinestatic
childCount() const noexceptMila::Dnn::CompositeComponent< TDeviceType, TPrecision >inline
clearComponents()Mila::Dnn::CompositeComponent< TDeviceType, TPrecision >inline
Component(const std::string &name)Mila::Dnn::Component< TDeviceType, TPrecision >inlineexplicit
ComponentBase typedef (defined in Mila::Dnn::CompositeComponent< TDeviceType, TPrecision >)Mila::Dnn::CompositeComponent< TDeviceType, TPrecision >
ComponentPtr typedef (defined in Mila::Dnn::GptTransformer< TDeviceType, TPrecision >)Mila::Dnn::GptTransformer< TDeviceType, TPrecision >
CompositeBase typedef (defined in Mila::Dnn::Network< TDeviceType, TPrecision >)Mila::Dnn::Network< TDeviceType, TPrecision >
CompositeComponent(const std::string &name)Mila::Dnn::CompositeComponent< TDeviceType, TPrecision >inlineexplicit
CompositeComponent(const CompositeComponent &)=delete (defined in Mila::Dnn::CompositeComponent< TDeviceType, TPrecision >)Mila::Dnn::CompositeComponent< TDeviceType, TPrecision >
CompositeComponent(CompositeComponent &&) noexcept=default (defined in Mila::Dnn::CompositeComponent< TDeviceType, TPrecision >)Mila::Dnn::CompositeComponent< TDeviceType, TPrecision >
configFromArchive(const ModelArchive &archive)Mila::Dnn::GptTransformer< TDeviceType, TPrecision >inlinestatic
createOptimizer(const TConfig &config)Mila::Dnn::Network< TDeviceType, TPrecision >inline
decode(const TokenIndexType &input, dim_t position) overrideMila::Dnn::GptTransformer< TDeviceType, TPrecision >inline
Mila::Dnn::LanguageNetwork::decode(const TokenIndexType &input, dim_t position)=0Mila::Dnn::LanguageNetwork< TDeviceType, TPrecision >pure virtual
EncoderType typedef (defined in Mila::Dnn::GptTransformer< TDeviceType, TPrecision >)Mila::Dnn::GptTransformer< TDeviceType, TPrecision >
findComponent(const std::string &path) constMila::Dnn::CompositeComponent< TDeviceType, TPrecision >inline
forward(const TokenIndexType &input) override (defined in Mila::Dnn::GptTransformer< TDeviceType, TPrecision >)Mila::Dnn::GptTransformer< TDeviceType, TPrecision >inline
Mila::Dnn::LanguageNetwork::forward(const TokenIndexType &input)=0Mila::Dnn::LanguageNetwork< TDeviceType, TPrecision >pure virtual
fromPretrained(const std::filesystem::path &model_path, std::size_t batch_size, std::size_t seq_length, DeviceId device_id=DeviceId{ TDeviceType, 0 }, bool strict=true) (defined in Mila::Dnn::GptTransformer< TDeviceType, TPrecision >)Mila::Dnn::GptTransformer< TDeviceType, TPrecision >inlinestatic
getComponent(const std::string &name) constMila::Dnn::CompositeComponent< TDeviceType, TPrecision >inline
getComponentAs(const std::string &name) constMila::Dnn::CompositeComponent< TDeviceType, TPrecision >inlineprotected
getComponents() constMila::Dnn::CompositeComponent< TDeviceType, TPrecision >inline
getDeviceId() const noexcept overrideMila::Dnn::Network< TDeviceType, TPrecision >inlinevirtual
getDeviceType()Mila::Dnn::Component< TDeviceType, TPrecision >inlinestatic
getExecutionContext() const (defined in Mila::Dnn::GptTransformer< TDeviceType, TPrecision >)Mila::Dnn::GptTransformer< TDeviceType, TPrecision >inline
getGradients() const overrideMila::Dnn::CompositeComponent< TDeviceType, TPrecision >inlinevirtual
getMemoryStats() const overrideMila::Dnn::GptTransformer< TDeviceType, TPrecision >inlinevirtual
getModelType() const (defined in Mila::Dnn::GptTransformer< TDeviceType, TPrecision >)Mila::Dnn::GptTransformer< TDeviceType, TPrecision >inline
getName() constMila::Dnn::Component< TDeviceType, TPrecision >inline
getParameterNames() constMila::Dnn::Component< TDeviceType, TPrecision >inlinevirtual
getParameters() const overrideMila::Dnn::CompositeComponent< TDeviceType, TPrecision >inlinevirtual
getPrecision() noexceptMila::Dnn::Component< TDeviceType, TPrecision >inlinestatic
getRequiredMemory(const BuildContext &context) constMila::Dnn::Component< TDeviceType, TPrecision >inlinevirtual
getTrainingMode() const noexceptMila::Dnn::Component< TDeviceType, TPrecision >inline
getType() const overrideMila::Dnn::Network< TDeviceType, TPrecision >inlinevirtual
GptTransformer(const std::string &name, const GptConfig &config, DeviceId device_id)Mila::Dnn::GptTransformer< TDeviceType, TPrecision >inlineexplicit
hasChildren() const noexceptMila::Dnn::CompositeComponent< TDeviceType, TPrecision >inline
hasComponent(const std::string &name) constMila::Dnn::CompositeComponent< TDeviceType, TPrecision >inline
hasExecutionContext() const noexceptMila::Dnn::Component< TDeviceType, TPrecision >inlineprotected
HostStagingMemoryResource typedefMila::Dnn::Component< TDeviceType, TPrecision >protected
isBuilt() const finalMila::Dnn::Component< TDeviceType, TPrecision >inlinevirtual
LanguageNetwork(const std::string &name) (defined in Mila::Dnn::LanguageNetwork< TDeviceType, TPrecision >)Mila::Dnn::LanguageNetwork< TDeviceType, TPrecision >inlineexplicit
LayerNormType typedef (defined in Mila::Dnn::GptTransformer< TDeviceType, TPrecision >)Mila::Dnn::GptTransformer< TDeviceType, TPrecision >
LinearType typedef (defined in Mila::Dnn::GptTransformer< TDeviceType, TPrecision >)Mila::Dnn::GptTransformer< TDeviceType, TPrecision >
load(ModelArchive &archive, SerializationMode mode)Mila::Dnn::Network< TDeviceType, TPrecision >inline
load_(ModelArchive &archive, SerializationMode) overrideMila::Dnn::GptTransformer< TDeviceType, TPrecision >inlineprotectedvirtual
loadParameter(const std::string &, const Serialization::ITensorBlob &)Mila::Dnn::Component< TDeviceType, TPrecision >inlinevirtual
loadParameterFromBlob(const std::string &param_name, const Serialization::ITensorBlob &blob, Tensor< TParameterPrecision, TMemoryResource > &target, const shape_t &expected_shape)Mila::Dnn::Component< TDeviceType, TPrecision >inlineprotected
loadParameters(PretrainedModelReader &reader, bool strict)Mila::Dnn::GptTransformer< TDeviceType, TPrecision >inline
MR typedef (defined in Mila::Dnn::GptTransformer< TDeviceType, TPrecision >)Mila::Dnn::GptTransformer< TDeviceType, TPrecision >
Network(const std::string &name)Mila::Dnn::Network< TDeviceType, TPrecision >inlineexplicit
NetworkBase typedef (defined in Mila::Dnn::GptTransformer< TDeviceType, TPrecision >)Mila::Dnn::GptTransformer< TDeviceType, TPrecision >
onBuilding(const BuildContext &context) overrideMila::Dnn::GptTransformer< TDeviceType, TPrecision >inlineprotectedvirtual
onExecutionContextSet() overrideMila::Dnn::CompositeComponent< TDeviceType, TPrecision >inlineprotectedvirtual
onTrainingModeChanging(TrainingMode training_mode) overrideMila::Dnn::GptTransformer< TDeviceType, TPrecision >inlineprotectedvirtual
operator=(const CompositeComponent &)=delete (defined in Mila::Dnn::CompositeComponent< TDeviceType, TPrecision >)Mila::Dnn::CompositeComponent< TDeviceType, TPrecision >
operator=(CompositeComponent &&) noexcept=default (defined in Mila::Dnn::CompositeComponent< TDeviceType, TPrecision >)Mila::Dnn::CompositeComponent< TDeviceType, TPrecision >
optimize()Mila::Dnn::CompositeComponent< TDeviceType, TPrecision >inlineprotectedvirtual
parameterCount() const overrideMila::Dnn::CompositeComponent< TDeviceType, TPrecision >inlinevirtual
prefill(const TokenIndexType &input) overrideMila::Dnn::GptTransformer< TDeviceType, TPrecision >inline
Mila::Dnn::LanguageNetwork::prefill(const TokenIndexType &input)=0Mila::Dnn::LanguageNetwork< TDeviceType, TPrecision >pure virtual
prefillFrom(const TokenIndexType &input, dim_t start_offset)Mila::Dnn::LanguageNetwork< TDeviceType, TPrecision >inlinevirtual
removeComponent(const std::string &name)Mila::Dnn::CompositeComponent< TDeviceType, TPrecision >inline
requireSerializableParameters() const overrideMila::Dnn::CompositeComponent< TDeviceType, TPrecision >inlineprotectedvirtual
rewindKvCache(dim_t position)Mila::Dnn::LanguageNetwork< TDeviceType, TPrecision >inlinevirtual
save(ModelArchive &archive, SerializationMode mode) constMila::Dnn::Network< TDeviceType, TPrecision >inline
save_(ModelArchive &archive, SerializationMode) const overrideMila::Dnn::GptTransformer< TDeviceType, TPrecision >inlineprotectedvirtual
saveFlatTensors(Serialization::SafeTensorsWriter &writer, const std::string &prefix, Serialization::TensorSavePass pass) const overrideMila::Dnn::CompositeComponent< TDeviceType, TPrecision >inlinevirtual
saveParameterToArchive(ModelArchive &archive, const std::string &parameter_name, const Tensor< TParameterPrecision, TMemoryResource > &parameter) constMila::Dnn::Component< TDeviceType, TPrecision >inlineprotected
saveParameterToWriter(Serialization::SafeTensorsWriter &writer, const std::string &flat_name, const Tensor< TParameterPrecision, TMemoryResource > &parameter, Serialization::TensorSavePass pass) constMila::Dnn::Component< TDeviceType, TPrecision >inlineprotected
setExecutionContext(IExecutionContext *context)Mila::Dnn::Component< TDeviceType, TPrecision >inlineprotected
setStageProbe(StageProbe probe)Mila::Dnn::LanguageNetwork< TDeviceType, TPrecision >inlinevirtual
setTrainingMode(TrainingMode mode)Mila::Dnn::Component< TDeviceType, TPrecision >inline
StageProbe typedefMila::Dnn::LanguageNetwork< TDeviceType, TPrecision >
synchronize() overrideMila::Dnn::Network< TDeviceType, TPrecision >inlinevirtual
TensorType typedef (defined in Mila::Dnn::GptTransformer< TDeviceType, TPrecision >)Mila::Dnn::GptTransformer< TDeviceType, TPrecision >
TokenIndexType typedef (defined in Mila::Dnn::GptTransformer< TDeviceType, TPrecision >)Mila::Dnn::GptTransformer< TDeviceType, TPrecision >
toString() const overrideMila::Dnn::GptTransformer< TDeviceType, TPrecision >inlinevirtual
TransformerBlockType typedef (defined in Mila::Dnn::GptTransformer< TDeviceType, TPrecision >)Mila::Dnn::GptTransformer< TDeviceType, TPrecision >
tryFindComponent(const std::string &path) constMila::Dnn::CompositeComponent< TDeviceType, TPrecision >inline
verifyArchitectureCompatibility(const PretrainedMetadata &metadata)Mila::Dnn::Network< TDeviceType, TPrecision >inlineprotected
zeroGradients() overrideMila::Dnn::GptTransformer< TDeviceType, TPrecision >inlinevirtual
~Component()=default (defined in Mila::Dnn::Component< TDeviceType, TPrecision >)Mila::Dnn::Component< TDeviceType, TPrecision >virtual
~CompositeComponent()=default (defined in Mila::Dnn::CompositeComponent< TDeviceType, TPrecision >)Mila::Dnn::CompositeComponent< TDeviceType, TPrecision >virtual
~GptTransformer() override=default (defined in Mila::Dnn::GptTransformer< TDeviceType, TPrecision >)Mila::Dnn::GptTransformer< TDeviceType, TPrecision >
~LanguageNetwork() override=default (defined in Mila::Dnn::LanguageNetwork< TDeviceType, TPrecision >)Mila::Dnn::LanguageNetwork< TDeviceType, TPrecision >
~Network() override=default (defined in Mila::Dnn::Network< TDeviceType, TPrecision >)Mila::Dnn::Network< TDeviceType, TPrecision >