Skip to content
New issue

Have a question about this project? Sign up for a free GitHub account to open an issue and contact its maintainers and the community.

Sign up for GitHub

By clicking “Sign up for GitHub”, you agree to our terms of service and privacy statement. We’ll occasionally send you account related emails.

Already on GitHub? Sign in to your account

Add MathsProvider to template for dynamic model loading #152

Merged
merged 2 commits into from
Nov 5, 2024
Merged

Add MathsProvider to template for dynamic model loading #152

Changes from all commits
Commits
Show all changes
2 commits
Select commit Hold shift + click to select a range
3fe1f9f
Add MathsProvider to template for dynamic model loading
mikeoliphant Nov 4, 2024
42a2a2c
Add MathProvider to GRU layer creation and to createActivation.
mikeoliphant Nov 4, 2024
File filter

Filter by extension

Filter by extension
Conversations
Failed to load comments.
Loading
Jump to
Jump to file
Failed to load files.
Loading
Diff view
Diff view
30 changes: 15 additions & 15 deletions RTNeural/model_loader.h
View file
Open in desktop
Original file line number Diff line number Diff line change
Expand Up @@ -318,10 +318,10 @@ namespace json_parser
}

/** Creates a GRULayer from a json representation of the layer weights. */
template <typename T>
std::unique_ptr<GRULayer<T>> createGRU(int in_size, int out_size, const nlohmann::json& weights)
template <typename T, typename MathsProvider = DefaultMathsProvider>
std::unique_ptr<GRULayer<T, MathsProvider>> createGRU(int in_size, int out_size, const nlohmann::json& weights)
{
auto gru = std::make_unique<GRULayer<T>>(in_size, out_size);
auto gru = std::make_unique<GRULayer<T, MathsProvider>>(in_size, out_size);
loadGRU<T>(*gru.get(), weights);
return std::move(gru);
}
Expand Down Expand Up @@ -385,10 +385,10 @@ namespace json_parser
}

/** Creates a LSTMLayer from a json representation of the layer weights. */
template <typename T>
std::unique_ptr<LSTMLayer<T>> createLSTM(int in_size, int out_size, const nlohmann::json& weights)
template <typename T, typename MathsProvider = DefaultMathsProvider >
std::unique_ptr<LSTMLayer<T, MathsProvider>> createLSTM(int in_size, int out_size, const nlohmann::json& weights)
{
auto lstm = std::make_unique<LSTMLayer<T>>(in_size, out_size);
auto lstm = std::make_unique<LSTMLayer<T, MathsProvider>>(in_size, out_size);
loadLSTM<T>(*lstm.get(), weights);
return std::move(lstm);
}
Expand Down Expand Up @@ -560,24 +560,24 @@ namespace json_parser
}

/** Creates an activation layer of a given type. */
template <typename T>
template <typename T, typename MathsProvider = DefaultMathsProvider>
std::unique_ptr<Activation<T>>
createActivation(const std::string& activationType, int dims)
{
if(activationType == "tanh")
return std::make_unique<TanhActivation<T>>(dims);
return std::make_unique<TanhActivation<T, MathsProvider>>(dims);

if(activationType == "relu")
return std::make_unique<ReLuActivation<T>>(dims);

if(activationType == "sigmoid")
return std::make_unique<SigmoidActivation<T>>(dims);
return std::make_unique<SigmoidActivation<T, MathsProvider>>(dims);

if(activationType == "softmax")
return std::make_unique<SoftmaxActivation<T>>(dims);
return std::make_unique<SoftmaxActivation<T, MathsProvider>>(dims);

if(activationType == "elu")
return std::make_unique<ELuActivation<T>>(dims);
return std::make_unique<ELuActivation<T, MathsProvider>>(dims);

return {};
}
Expand All @@ -602,7 +602,7 @@ namespace json_parser
}

/** Creates a neural network model from a json stream. */
template <typename T>
template <typename T, typename MathsProvider = DefaultMathsProvider>
std::unique_ptr<Model<T>> parseJson(const nlohmann::json& parent, const bool debug = false)
{
auto shape = parent.at("in_shape");
Expand Down Expand Up @@ -639,7 +639,7 @@ namespace json_parser
if(!activationType.empty())
{
debug_print(" activation: " + activationType, debug);
auto activation = createActivation<T>(activationType, layerDims);
auto activation = createActivation<T, MathsProvider>(activationType, layerDims);
_model->addLayer(activation.release());
}
}
Expand Down Expand Up @@ -683,12 +683,12 @@ namespace json_parser
}
else if(type == "gru")
{
auto gru = createGRU<T>(model->getNextInSize(), layerDims, weights);
auto gru = createGRU<T, MathsProvider>(model->getNextInSize(), layerDims, weights);
model->addLayer(gru.release());
}
else if(type == "lstm")
{
auto lstm = createLSTM<T>(model->getNextInSize(), layerDims, weights);
auto lstm = createLSTM<T, MathsProvider>(model->getNextInSize(), layerDims, weights);
model->addLayer(lstm.release());
}
else if(type == "prelu")
Expand Down
Loading