NeuralAmpModelerCore

https://neuralampmodelercore.readthedocs.io/en/latest/
Admin
A high-performance C++ library for running neural network-based audio processing models, supporting...
Tokens:19,807
Snippets:462
Trust Score:9.6
Update:4 months ago
Show doc for...
Tokens:
Raw
### instance

Source: https://neuralampmodelercore.readthedocs.io/en/latest/api/get_dsp.html

Static function to get the singleton instance of ConfigParserRegistry.

```C++
static inline ConfigParserRegistry &instance()
{
    // Static function to get the singleton instance of ConfigParserRegistry.
}
```

--------------------------------

### DSP Object Creation from Configuration JSON

Source: https://neuralampmodelercore.readthedocs.io/en/latest/api/get_dsp.html

Demonstrates how to get a DSP object from a provided configuration JSON object, with an option to also retrieve the configuration data.

```C++
std::unique_ptr<DSP> get_dsp(const nlohmann::json &config, dspData &returnedConfig)
std::unique_ptr<DSP> get_dsp(const nlohmann::json &config)
```

--------------------------------

### Activation processing (gated mode example)

Source: https://neuralampmodelercore.readthedocs.io/en/latest/_sources/wavenet_walkthrough.rst.txt

This C++ code snippet demonstrates the activation processing logic when the gating mode is set to GATED. It applies a gating activation and optionally a post-activation FiLM.

```cpp
if (this->_gating_mode == GatingMode::GATED) {
    auto input_block = this->_z.leftCols(num_frames);
    auto output_block = this->_z.topRows(bottleneck).leftCols(num_frames);
    this->_gating_activation->apply(input_block, output_block);
    if (this->_activation_post_film) {
        this->_activation_post_film->Process(this->_z.topRows(bottleneck), condition, num_frames);
        this->_z.topRows(bottleneck).leftCols(num_frames).noalias() =
            this->_activation_post_film->GetOutput().leftCols(num_frames);
    }
}
```

--------------------------------

### Helper Registration

Source: https://neuralampmodelercore.readthedocs.io/en/latest/api/get_dsp.html

Registration helper for factories (external code compatibility). Wraps a FactoryFunction into a ConfigParserRegistry entry via FactoryConfig. Use this to register external architectures. Create a static instance to automatically register a factory when the program starts.

```cpp
struct Helper
{
    inline Helper(const std::string &name, FactoryFunction factory) 
    // Parameters:
    //   name – Architecture name
    //   factory – Factory function
}
```

--------------------------------

### Process Input to Output (Legacy)

Source: https://neuralampmodelercore.readthedocs.io/en/latest/api/conv1d.html

A legacy method for processing input and writing to an output matrix. It allows specifying start indices and the number of columns to process, maintaining compatibility with older versions.

```cpp
void process_(const Eigen::MatrixXf &input, Eigen::MatrixXf &output, const long i_start, const long ncols, const long j_start) const
```

--------------------------------

### GetOutput (const)

Source: https://neuralampmodelercore.readthedocs.io/en/latest/api/get_dsp.html

Get the entire internal output buffer (const version).

```C++
inline const Eigen::MatrixXf &GetOutput() const
{
    // Get the entire internal output buffer (const version)
    // Returns:
    // Const reference to the output buffer
}
```

--------------------------------

### DSP Object Creation from dspData Struct

Source: https://neuralampmodelercore.readthedocs.io/en/latest/api/get_dsp.html

Provides a way to get a DSP object directly from a dspData structure containing the model's configuration and weights.

```C++
std::unique_ptr<DSP> get_dsp(dspData &conf)
```

--------------------------------

### Get Sample Rate from .nam File

Source: https://neuralampmodelercore.readthedocs.io/en/latest/api/get_dsp.html

Retrieves the sample rate from a .nam file's JSON object.

```C++
double get_sample_rate_from_nam_file(const nlohmann::json &j)
```

--------------------------------

### BatchNorm Process Method

Source: https://neuralampmodelercore.readthedocs.io/en/latest/api/convnet.html

Processes input in-place within specified start and end indices.

```cpp
void process_(Eigen::MatrixXf &input, const long i_start, const long i_end) const

```

--------------------------------

### GetOutput (non-const)

Source: https://neuralampmodelercore.readthedocs.io/en/latest/api/get_dsp.html

Get the entire internal output buffer. This is intended for internal wiring between layers; callers should treat the buffer as pre-allocated storage and only consider the first num_frames columns valid for a given processing call. Slice with .leftCols(num_frames) as needed.

```C++
inline Eigen::MatrixXf &GetOutput()
{
    // Get the entire internal output buffer.
    // This is intended for internal wiring between layers; callers should treat the buffer as pre-allocated storage and only consider the first num_frames columns valid for a given processing call. Slice with .leftCols(num_frames) as needed.
    // Returns:
    // Reference to the output buffer
}
```

--------------------------------

### DSP Object Creation from Directory (Legacy)

Source: https://neuralampmodelercore.readthedocs.io/en/latest/api/get_dsp.html

Illustrates the legacy method for loading DSP models from a directory structure.

```C++
std::unique_ptr<DSP> get_dsp_legacy(const std::filesystem::path dirname)
```

--------------------------------

### Get Kernel Size

Source: https://neuralampmodelercore.readthedocs.io/en/latest/api/conv1d.html

Returns the kernel size of the convolution layer.

```cpp
long get_kernel_size() const
```

--------------------------------

### DSP Object Creation from .nam File

Source: https://neuralampmodelercore.readthedocs.io/en/latest/api/get_dsp.html

Shows how to load a DSP object from a .nam file, with an option to store the model's configuration data.

```C++
std::unique_ptr<DSP> get_dsp(const std::filesystem::path config_filename)
std::unique_ptr<DSP> get_dsp(const std::filesystem::path config_filename, dspData &returnedConfig)
```

--------------------------------

### RingBuffer GetChannels

Source: https://neuralampmodelercore.readthedocs.io/en/latest/api/ring_buffer.html

Get the number of channels (rows) configured for the RingBuffer.

```C++
inline int GetChannels() const

```

--------------------------------

### RingBuffer GetMaxBufferSize

Source: https://neuralampmodelercore.readthedocs.io/en/latest/api/ring_buffer.html

Get the maximum buffer size that was set during initialization.

```C++
inline int GetMaxBufferSize() const

```

--------------------------------

### Head scaling and output

Source: https://neuralampmodelercore.readthedocs.io/en/latest/_sources/wavenet_walkthrough.rst.txt

Scales the final head output from the last LayerArray and writes it to the output buffers.

```cpp
Eigen::MatrixXf& final_head = this->_layer_arrays.back().GetHeadOutputs();
// Apply head scale and write to output buffers
// (implementation details in wavenet.cpp)
```

--------------------------------

### Conv1x1 GetOutput Function

Source: https://neuralampmodelercore.readthedocs.io/en/latest/api/dsp.html

Gets the internal output buffer of the Conv1x1 layer.

```C++
inline Eigen::MatrixXf &GetOutput()
```

```C++
inline const Eigen::MatrixXf &GetOutput() const
```

--------------------------------

### ConfigParserHelper

Source: https://neuralampmodelercore.readthedocs.io/en/latest/api/get_dsp.html

Auto-registration helper for config parsers. Create a static instance to register a config parser at program startup.

```C++
#include <model_config.h>
Auto-registration helper for config parsers.
Create a static instance to register a config parser at program startup.
```

--------------------------------

### Get Dilation Factor

Source: https://neuralampmodelercore.readthedocs.io/en/latest/api/conv1d.html

Returns the dilation factor currently set for the Conv1D layer.

```cpp
inline int get_dilation() const
```

--------------------------------

### BlendingActivation Constructor

Source: https://neuralampmodelercore.readthedocs.io/en/latest/api/get_dsp.html

Constructor for BlendingActivation.

```cpp
inline BlendingActivation(activations::Activation::Ptr input_act, activations::Activation::Ptr blend_act, int input_channels = 1)
```

--------------------------------

### Input mixin processing

Source: https://neuralampmodelercore.readthedocs.io/en/latest/_sources/wavenet_walkthrough.rst.txt

This code snippet shows the processing steps for the input mixin, which handles the conditioning input separately and adds it to the convolution output, including optional pre-FiLM, input mixin convolution, and optional post-FiLM.

```cpp
if (this->_input_mixin_pre_film) {
    this->_input_mixin_pre_film->Process(condition, condition, num_frames);
    this->_input_mixin.process_(this->_input_mixin_pre_film->GetOutput(), num_frames);
} else {
    this->_input_mixin.process_(condition, num_frames);
}
if (this->_input_mixin_post_film) {
    Eigen::MatrixXf& input_mixin_output = this->_input_mixin.GetOutput();
    this->_input_mixin_post_film->Process_(input_mixin_output, condition, num_frames);
}
```

--------------------------------

### Get Output Channels

Source: https://neuralampmodelercore.readthedocs.io/en/latest/api/conv1d.html

Returns the number of output channels configured for the Conv1D layer.

```cpp
long get_out_channels() const
```

--------------------------------

### ConvNetConfig create Method

Source: https://neuralampmodelercore.readthedocs.io/en/latest/api/convnet.html

Constructs a DSP object from the ConvNet configuration.

```cpp
virtual std::unique_ptr<DSP> create(std::vector<float> weights, double sampleRate) override
```

--------------------------------

### Get Input Channels

Source: https://neuralampmodelercore.readthedocs.io/en/latest/api/conv1d.html

Returns the number of input channels configured for the Conv1D layer.

```cpp
long get_in_channels() const
```

--------------------------------

### GatingActivation Constructor

Source: https://neuralampmodelercore.readthedocs.io/en/latest/api/get_dsp.html

Constructor for GatingActivation.

```cpp
inline GatingActivation(activations::Activation::Ptr input_act, activations::Activation::Ptr gating_act, int input_channels = 1)
// Parameters:
//   input_act – Activation function for input channels
//   gating_act – Activation function for gating channels
//   input_channels – Number of input channels (default: 1)
//   gating_channels – Number of gating channels (default: 1)
```

--------------------------------

### Sum and pre-activation FiLM

Source: https://neuralampmodelercore.readthedocs.io/en/latest/_sources/wavenet_walkthrough.rst.txt

This code snippet demonstrates the summation of the convolution output and input mixin output, followed by an optional pre-activation FiLM modulation.

```cpp
this->_z.leftCols(num_frames).noalias() =

```

--------------------------------

### Get Number of Weights

Source: https://neuralampmodelercore.readthedocs.io/en/latest/api/conv1d.html

Calculates and returns the total number of weight parameters required for the Conv1D layer based on its configuration.

```cpp
long get_num_weights() const
```

--------------------------------

### Input convolution processing

Source: https://neuralampmodelercore.readthedocs.io/en/latest/_sources/wavenet_walkthrough.rst.txt

This code snippet illustrates the processing steps for the input convolution within a WaveNet layer, including optional pre-FiLM, dilated convolution, and optional post-FiLM.

```cpp
if (this->_conv_pre_film) {
    this->_conv_pre_film->Process(input, condition, num_frames);
    this->_conv.Process(this->_conv_pre_film->GetOutput(), num_frames);
} else {
    this->_conv.Process(input, num_frames);
}
if (this->_conv_post_film) {
    Eigen::MatrixXf& conv_output = this->_conv.GetOutput();
    this->_conv_post_film->Process_(conv_output, condition, num_frames);
}
```

--------------------------------

### Get Output Buffer (Const Version)

Source: https://neuralampmodelercore.readthedocs.io/en/latest/api/conv1d.html

Retrieves a const reference to the entire internal output buffer. This version is used when the object is not modifiable.

```cpp
inline const Eigen::MatrixXf &GetOutput() const
```

--------------------------------

### Conv1D Constructor (Default)

Source: https://neuralampmodelercore.readthedocs.io/en/latest/api/get_dsp.html

Default constructor. Initializes with dilation=1 and groups=1. Use set_size_() to configure.

```C++
inline Conv1D()
{
    // Default constructor.
    // Initializes with dilation=1 and groups=1. Use set_size_() to configure.
}
```

--------------------------------

### set_size_and_weights_

Source: https://neuralampmodelercore.readthedocs.io/en/latest/api/get_dsp.html

Set size and weights in one call.

```C++
void set_size_and_weights_(const int in_channels, const int out_channels, const int kernel_size, const int _dilation, const bool do_bias, const int groups, std::vector<float>::iterator &weights)
{
    // Set size and weights in one call.
    // Parameters:
    //  * **in_channels** – Number of input channels
    //  * **out_channels** – Number of output channels
    //  * **kernel_size** – Size of the convolution kernel
    //  * **_dilation** – Dilation factor for the convolution
    //  * **do_bias** – Whether to use bias
    //  * **groups** – Number of groups for grouped convolution
    //  * **weights** – Iterator to the weights vector. Will be advanced as weights are consumed.
}
```

--------------------------------

### Head 1x1 processing

Source: https://neuralampmodelercore.readthedocs.io/en/latest/_sources/wavenet_walkthrough.rst.txt

This C++ code snippet illustrates the processing performed by a head 1x1 convolution if it is configured. This step is optional and handles the skip connection.

```cpp
if (this->_head1x1) {
    this->_head1x1->process_(this->_z.topRows(bottleneck).leftCols(num_frames), num_frames);
    if (this->_head1x1_post_film) {
        Eigen::MatrixXf& head1x1_output = this->_head1x1->GetOutput();
        this->_head1x1_post_film->Process_(head1x1_output, condition, num_frames);
    }
    this->_output_head.leftCols(num_frames).noalias() =
        this->_head1x1->GetOutput().leftCols(num_frames);
}
```

--------------------------------

### Get Output Buffer

Source: https://neuralampmodelercore.readthedocs.io/en/latest/api/conv1d.html

Retrieves a reference to the entire internal output buffer. This is primarily for internal layer wiring. Callers should treat it as pre-allocated storage and use .leftCols(num_frames) to access valid data for a given processing call.

```cpp
inline Eigen::MatrixXf &GetOutput()
```

--------------------------------

### Condition processing

Source: https://neuralampmodelercore.readthedocs.io/en/latest/_sources/wavenet_walkthrough.rst.txt

Processes the input through a condition DSP if provided, otherwise uses the input directly as the condition.

```cpp
void WaveNet::_process_condition(const int num_frames) {
    if (this->_condition_dsp != nullptr) {
        // Process input through condition DSP
        this->_condition_dsp->process(/* input */, /* output */, num_frames);
        // Copy output to condition buffer
    } else {
        // Use input directly as condition
        this->_condition_output = this->_condition_input;
    }
}
```

--------------------------------

### Linear Config Creator

Source: https://neuralampmodelercore.readthedocs.io/en/latest/api/get_dsp.html

Config parser for ConfigParserRegistry.

```cpp
std::unique_ptr<ModelConfig> create_config(const nlohmann::json &config, double sampleRate)
// Parameters:
//   * **config** – JSON configuration object
//   * **sampleRate** – Expected sample rate in Hz
// Returns:
// unique_ptr<ModelConfig> wrapping a LinearConfig
```

--------------------------------

### ConfigParserRegistry

Source: https://neuralampmodelercore.readthedocs.io/en/latest/api/get_dsp.html

Singleton registry mapping architecture names to config parser functions. Both built-in and external architectures register here. There is one construction path for all architectures.

```C++
#include <model_config.h>
Singleton registry mapping architecture names to config parser functions.
Both built-in and external architectures register here. There is one construction path for all architectures.
```

--------------------------------

### Version Support Checking

Source: https://neuralampmodelercore.readthedocs.io/en/latest/api/get_dsp.html

Functions for checking and registering version support for model configurations.

```C++
void verify_config_version(const std::string version)
void register_version_support_checker(std::shared_ptr<const IVersionSupportChecker> checker)
Supported is_version_supported(const std::string version)
```

--------------------------------

### BlendingActivation Apply

Source: https://neuralampmodelercore.readthedocs.io/en/latest/api/get_dsp.html

Apply blending activation to input matrix.

```cpp
template<typename InputDerived, typename OutputDerived>
inline void apply(const Eigen::MatrixBase<InputDerived> &input, Eigen::MatrixBase<OutputDerived> &output)
```

--------------------------------

### Process

Source: https://neuralampmodelercore.readthedocs.io/en/latest/api/get_dsp.html

Process input and write to internal output buffer.

```C++
void Process(const Eigen::MatrixXf &input, const int num_frames)
{
    // Process input and write to internal output buffer.
    // Parameters:
    //  * **input** – Input matrix (channels x num_frames)
    //  * **num_frames** – Number of frames to process
}
```

--------------------------------

### Input rechanneling

Source: https://neuralampmodelercore.readthedocs.io/en/latest/_sources/wavenet_walkthrough.rst.txt

The input is first projected (rechanneled) to match the layer channel count.

```cpp
this->_rechannel.process_(layer_inputs, num_frames);
Eigen::MatrixXf& rechannel_output = _rechannel.GetOutput();
```

--------------------------------

### ModelConfig::create Function

Source: https://neuralampmodelercore.readthedocs.io/en/latest/api/get_dsp.html

Abstract function to construct a DSP object from configuration.

```cpp
virtual std::unique_ptr<DSP> create(std::vector<float> weights, double sampleRate) = 0

    
Construct a DSP object from this configuration.

Parameters:

  * **weights** – Model weights (taken by value to allow move for WaveNet)
  * **sampleRate** – Expected sample rate in Hz

Returns:

Unique pointer to a DSP object
```

--------------------------------

### Linear Config Parser

Source: https://neuralampmodelercore.readthedocs.io/en/latest/api/get_dsp.html

Parse Linear configuration from JSON.

```cpp
LinearConfig parse_config_json(const nlohmann::json &config)
// Parameters:
// **config** – JSON configuration object
// Returns:
// LinearConfig
```

--------------------------------

### Conv1D Constructor (Parameterized)

Source: https://neuralampmodelercore.readthedocs.io/en/latest/api/get_dsp.html

Constructor for Conv1D layer.

```C++
inline Conv1D(const int in_channels, const int out_channels, const int kernel_size, const int bias, const int dilation, const int groups = 1)
{
    // Constructor.
    // Parameters:
    //  * **in_channels** – Number of input channels
    //  * **out_channels** – Number of output channels
    //  * **kernel_size** – Size of the convolution kernel
    //  * **bias** – Whether to use bias (1 for true, 0 for false)
    //  * **dilation** – Dilation factor for the convolution
    //  * **groups** – Number of groups for grouped convolution (default: 1)
}
```

--------------------------------

### Layer processing loop

Source: https://neuralampmodelercore.readthedocs.io/en/latest/_sources/wavenet_walkthrough.rst.txt

Each layer processes the output of the previous layer, accumulating head outputs.

```cpp
for (size_t i = 0; i < this->_layers.size(); i++) {
    if (i == 0) {
        // First layer consumes the rechannel output buffer
        this->_layers[i].Process(rechannel_output, condition, num_frames);
    } else {
        // Subsequent layers consume the previous layer's output
        Eigen::MatrixXf& prev_output = this->_layers[i - 1].GetOutputNextLayer();
        this->_layers[i].Process(prev_output, condition, num_frames);
    }
    
    // Accumulate head output from this layer
    this->_head_inputs.leftCols(num_frames).noalias() += 
        this->_layers[i].GetOutputHead().leftCols(num_frames);
}
```

--------------------------------

### BatchNorm Default Constructor

Source: https://neuralampmodelercore.readthedocs.io/en/latest/api/convnet.html

Default constructor for BatchNorm.

```cpp
inline BatchNorm()

```

--------------------------------

### FiLM Constructor

Source: https://neuralampmodelercore.readthedocs.io/en/latest/api/get_dsp.html

Constructor for the FiLM class, which applies per-channel scaling and optional shifting.

```cpp
inline FiLM(const int condition_dim, const int input_dim, const bool shift, const int groups = 1)
```

--------------------------------

### LayerArray processing

Source: https://neuralampmodelercore.readthedocs.io/en/latest/_sources/wavenet_walkthrough.rst.txt

Processes LayerArrays sequentially, with the first processing the input and subsequent ones processing the previous output.

```cpp
// First layer array
this->_layer_arrays[0].Process(input, condition, num_frames);

// Subsequent layer arrays
for (size_t i = 1; i < this->_layer_arrays.size(); i++) {

```

--------------------------------

### Version Class

Source: https://neuralampmodelercore.readthedocs.io/en/latest/api/get_dsp.html

Represents a version with major, minor, and patch components.

```C++
#include <get_dsp.h>

class Version {
public:
    inline Version(int major, int minor, int patch);
    inline std::string toString() const;
    inline bool operator>(const Version &other) const;
    inline bool operator<(const Version &other) const;
    int major;
    int minor;
    int patch;
};
```

--------------------------------

### Residual and skip connections

Source: https://neuralampmodelercore.readthedocs.io/en/latest/_sources/wavenet_walkthrough.rst.txt

This C++ code snippet implements the final step of computing the residual and skip connections. It calculates the output for the next layer and the output for the head.

```cpp
// Store output to next layer (residual connection)
this->_output_next_layer.leftCols(num_frames).noalias() =
    input.leftCols(num_frames) + _1x1.GetOutput().leftCols(num_frames);

// Store output to head (skip connection)
if (this->_head1x1) {
    this->_output_head.leftCols(num_frames).noalias() =
        this->_head1x1->GetOutput().leftCols(num_frames);
} else {
    this->_output_head.leftCols(num_frames).noalias() =
        this->_z.topRows(bottleneck).leftCols(num_frames);
}
```

--------------------------------

### ConvNet Constructor

Source: https://neuralampmodelercore.readthedocs.io/en/latest/api/convnet.html

Constructor for the ConvNet class, initializing various parameters for the convolutional neural network.

```cpp
ConvNet(const int in_channels, const int out_channels, const int channels, const std::vector<int> &dilations, const bool batchnorm, const activations::ActivationConfig &activation_config, std::vector<float> &weights, const double expected_sample_rate = -1.0, const int groups = 1)
```

--------------------------------

### Activation Functions

Source: https://neuralampmodelercore.readthedocs.io/en/latest/api/get_dsp.html

Various activation functions.

```C++
inline float relu(float x);
inline float sigmoid(float x);
inline float hard_tanh(float x);
inline float leaky_hardtanh(float x, float min_val, float max_val, float min_slope, float max_slope);
inline float fast_tanh(const float x);
inline float fast_sigmoid(const float x);
inline float leaky_relu(float x, float negative_slope);
inline float leaky_relu(float x);
inline float swish(float x);
inline float hardswish(float x);
inline float softsign(float x);
```

--------------------------------

### 1x1 convolution

Source: https://neuralampmodelercore.readthedocs.io/en/latest/_sources/wavenet_walkthrough.rst.txt

This C++ code snippet shows the processing step involving a 1x1 convolution, which reduces the bottleneck channels back to the layer channel count. It also includes optional post-FiLM processing.

```cpp
_1x1.process_(this->_z.topRows(bottleneck), num_frames);
if (this->_1x1_post_film) {
    Eigen::MatrixXf& _1x1_output = this->_1x1.GetOutput();
    this->_1x1_post_film->Process_(_1x1_output, condition, num_frames);
}
```

--------------------------------

### BatchNorm Constructor with Weights

Source: https://neuralampmodelercore.readthedocs.io/en/latest/api/convnet.html

Constructor for BatchNorm that takes dimension and an iterator to weights. The weights iterator will be advanced as weights are consumed.

```cpp
BatchNorm(const int dim, std::vector<float>::iterator &weights)

```

--------------------------------

### IVersionSupportChecker::support Function

Source: https://neuralampmodelercore.readthedocs.io/en/latest/api/get_dsp.html

Pure virtual function to check version support.

```cpp
virtual Supported support(const std::string &version) const = 0
```

--------------------------------

### Linear Constructor

Source: https://neuralampmodelercore.readthedocs.io/en/latest/api/get_dsp.html

Constructor for the Linear class, implementing a simple linear convolution.

```cpp
Linear(const int in_channels, const int out_channels, const int receptive_field, const bool _bias, const std::vector<float> &weights, const double expected_sample_rate = -1.0)

    
Constructor.

Parameters:

  * **in_channels** – Number of input channels
  * **out_channels** – Number of output channels
  * **receptive_field** – Size of the impulse response
  * **_bias** – Whether to use bias
  * **weights** – Model weights (impulse response coefficients)
  * **expected_sample_rate** – Expected sample rate in Hz (-1.0 if unknown)
```

--------------------------------

### ConvNetBlock Process Method (Legacy)

Source: https://neuralampmodelercore.readthedocs.io/en/latest/api/convnet.html

Processes input using indices for compatibility with older versions.

```cpp
void process_(const Eigen::MatrixXf &input, Eigen::MatrixXf &output, const long i_start, const long i_end)
```

--------------------------------

### LSTM Config Creator

Source: https://neuralampmodelercore.readthedocs.io/en/latest/api/get_dsp.html

Config parser for ConfigParserRegistry.

```cpp
std::unique_ptr<ModelConfig> create_config(const nlohmann::json &config, double sampleRate)
```

--------------------------------

### set_weights_

Source: https://neuralampmodelercore.readthedocs.io/en/latest/api/get_dsp.html

Set the parameters (weights) of this module.

```C++
void set_weights_(std::vector<float>::iterator &weights)
{
    // Set the parameters (weights) of this module.
    // Parameters:
    // **weights** – Iterator to the weights vector. Will be advanced as weights are consumed.
}
```

--------------------------------

### ConvNet Process Method

Source: https://neuralampmodelercore.readthedocs.io/en/latest/api/convnet.html

Processes audio frames using the ConvNet model.

```cpp
virtual void process(NAM_SAMPLE **input, NAM_SAMPLE **output, const int num_frames) override
```

--------------------------------

### Buffer Class Constructor

Source: https://neuralampmodelercore.readthedocs.io/en/latest/api/get_dsp.html

Constructor for the Buffer class, which is a base class for DSP models requiring input buffering. This class is deprecated.

```C++
Buffer(const int in_channels, const int out_channels, const int receptive_field, const double expected_sample_rate = -1.0)
```

--------------------------------

### has

Source: https://neuralampmodelercore.readthedocs.io/en/latest/api/get_dsp.html

Check whether an architecture name is registered.

```C++
bool has(const std::string &name) const
{
    // Check whether an architecture name is registered.
}
```

--------------------------------

### Activation Class

Source: https://neuralampmodelercore.readthedocs.io/en/latest/api/get_dsp.html

Class for activation functions.

```C++
#include <activations.h>

class Activation;
```

--------------------------------

### SetMaxBufferSize

Source: https://neuralampmodelercore.readthedocs.io/en/latest/api/get_dsp.html

Reset the ring buffer and pre-allocate output buffer.

```C++
void SetMaxBufferSize(const int maxBufferSize)
{
    // Reset the ring buffer and pre-allocate output buffer.
    // Parameters:
    // **maxBufferSize** – Maximum buffer size for output buffer and to size ring buffer
}
```

--------------------------------

### Process_ Function (In-place)

Source: https://neuralampmodelercore.readthedocs.io/en/latest/api/get_dsp.html

Processes input data using conditioning in-place, modifying the input matrix directly.

```cpp
inline void Process_(Eigen::Ref<Eigen::MatrixXf> input, const Eigen::Ref<const Eigen::MatrixXf> &condition, const int num_frames)

    
Process input with conditioning (in-place)
Uses Eigen::Ref to accept matrices and block expressions without creating temporaries (real-time safe). Modifies the input matrix directly.

Parameters:

  * **input** – Input matrix (input_dim x num_frames), will be modified in-place
  * **condition** – Conditioning matrix (condition_dim x num_frames)
  * **num_frames** – Number of frames to process
```

--------------------------------

### DSP Object Construction

Source: https://neuralampmodelercore.readthedocs.io/en/latest/api/get_dsp.html

The primary construction path for DSP objects, used by both JSON and binary loaders. It handles construction, metadata application, and prewarming.

```C++
std::unique_ptr<DSP> create_dsp(std::unique_ptr<ModelConfig> config, std::vector<float> weights, const ModelMetadata &metadata)
```

--------------------------------

### Conv1x1 Constructor

Source: https://neuralampmodelercore.readthedocs.io/en/latest/api/dsp.html

Constructor for the 1x1 convolution, which performs a pointwise convolution.

```C++
Conv1x1(const int in_channels, const int out_channels, const bool _bias, const int groups = 1)
```

--------------------------------

### process_

Source: https://neuralampmodelercore.readthedocs.io/en/latest/api/get_dsp.html

Internal processing function for Conv1D.

```C++
void process_(const Eigen::MatrixXf &input, Eigen::MatrixXf &output, const long i_start, const long ncols, const long j_start) const
{
    // Internal processing function for Conv1D.
}
```

--------------------------------

### Model Configuration Parsing

Source: https://neuralampmodelercore.readthedocs.io/en/latest/api/get_dsp.html

Parses a ModelConfig from a JSON architecture name and config block, given an expected sample rate.

```C++
std::unique_ptr<ModelConfig> parse_model_config_json(const std::string &architecture, const nlohmann::json &config, double sample_rate)
```

--------------------------------

### ConvNetBlock GetOutput Method

Source: https://neuralampmodelercore.readthedocs.io/en/latest/api/convnet.html

Retrieves a block reference to the output from the last Process() call.

```cpp
Eigen::Block<Eigen::MatrixXf> GetOutput(const int num_frames)
```

--------------------------------

### Conv1D Default Constructor

Source: https://neuralampmodelercore.readthedocs.io/en/latest/api/conv1d.html

Initializes a Conv1D layer with default dilation=1 and groups=1. The size parameters can be configured later using set_size_().

```cpp
inline Conv1D() {
    // Initializes with dilation=1 and groups=1. Use set_size_() to configure.
}
```

--------------------------------

### Linear Model Constructor

Source: https://neuralampmodelercore.readthedocs.io/en/latest/api/dsp.html

Constructor for the basic linear model, which implements a simple linear convolution.

```C++
Linear(const int in_channels, const int out_channels, const int receptive_field, const bool _bias, const std::vector<float> &weights, const double expected_sample_rate = -1.0)
```

--------------------------------

### Conv1x1 Process Function (Matrix Input)

Source: https://neuralampmodelercore.readthedocs.io/en/latest/api/dsp.html

Processes an input matrix and returns the output matrix.

```C++
inline Eigen::MatrixXf process(const Eigen::MatrixXf &input) const
```

--------------------------------

### LSTM Config Parser

Source: https://neuralampmodelercore.readthedocs.io/en/latest/api/get_dsp.html

Parse LSTM configuration from JSON.

```cpp
LSTMConfig parse_config_json(const nlohmann::json &config)
// Parameters:
// **config** – JSON configuration object
// Returns:
// LSTMConfig
```

--------------------------------

### parse

Source: https://neuralampmodelercore.readthedocs.io/en/latest/api/get_dsp.html

Parse a ModelConfig from an architecture name, JSON config, and sample rate.

```C++
std::unique_ptr<ModelConfig> parse(const std::string &name, const nlohmann::json &config, double sampleRate) const
{
    // Parse a ModelConfig from an architecture name, JSON config, and sample rate.
    // Throws:
    //  std::runtime_error – If no parser is registered for the given name
}
```

--------------------------------

### ModelMetadata Structure

Source: https://neuralampmodelercore.readthedocs.io/en/latest/api/get_dsp.html

Metadata common to all model formats.

```C++
#include <model_config.h>

struct ModelMetadata {
    std::string version;
    double sample_rate = -1.0;
    std::optional<double> loudness;
    std::optional<double> input_level;
    std::optional<double> output_level;
};
```

--------------------------------

### Conv1x1 Process Function (Matrix and Frames Input)

Source: https://neuralampmodelercore.readthedocs.io/en/latest/api/dsp.html

Processes an input matrix for a specified number of frames and returns the output matrix.

```C++
Eigen::MatrixXf process(const Eigen::MatrixXf &input, const int num_frames) const
```

--------------------------------

### GetOutput Function

Source: https://neuralampmodelercore.readthedocs.io/en/latest/api/get_dsp.html

Retrieves the internal output buffer of the FiLM layer.

```cpp
inline Eigen::MatrixXf &GetOutput()

    
Get the entire internal output buffer.
This is intended for internal wiring between layers; callers should treat the buffer as pre-allocated storage and only consider the first num_frames columns valid for a given processing call. Slice with .leftCols(num_frames) as needed.

Returns:
Reference to the output buffer
```

```cpp
inline const Eigen::MatrixXf &GetOutput() const

    
Get the entire internal output buffer (const version)

Returns:
Const reference to the output buffer
```

--------------------------------

### FiLM Constructor

Source: https://neuralampmodelercore.readthedocs.io/en/latest/api/film.html

Constructor for the FiLM class. Initializes the module with specified dimensions and options for scaling and shifting.

```cpp
inline FiLM(const int condition_dim, const int input_dim, const bool shift, const int groups = 1)
{
    // ... (implementation details)
}

```

--------------------------------

### Conv1x1 Process Function (In-place)

Source: https://neuralampmodelercore.readthedocs.io/en/latest/api/dsp.html

Processes input and stores output to a pre-allocated buffer. Uses Eigen::Ref for real-time safety.

```C++
void process_(const Eigen::Ref<const Eigen::MatrixXf> &input, const int num_frames)
```

--------------------------------

### Process Input

Source: https://neuralampmodelercore.readthedocs.io/en/latest/api/conv1d.html

Processes the input matrix and writes the result to the internal output buffer. The number of frames to process is specified.

```cpp
void Process(const Eigen::MatrixXf &input, const int num_frames)
```

--------------------------------

### lowercase() function signature

Source: https://neuralampmodelercore.readthedocs.io/en/latest/api/util.html

The signature for the `lowercase()` function, which converts a string to lowercase.

```cpp
std::string lowercase(const std::string &s)
```

--------------------------------

### registerParser

Source: https://neuralampmodelercore.readthedocs.io/en/latest/api/get_dsp.html

Register a config parser for an architecture.

```C++
void registerParser(const std::string &name, ConfigParserFunction func)
{
    // Register a config parser for an architecture.
    // Parameters:
    //  * **name** – Architecture name (e.g., “WaveNet”, “LSTM”)
    //  * **func** – Parser function that returns a unique_ptr<ModelConfig>
    // Throws:
    //  std::runtime_error – If the name is already registered
}
```

--------------------------------

### Conv1D Constructor

Source: https://neuralampmodelercore.readthedocs.io/en/latest/api/conv1d.html

Constructs a Conv1D layer with specified input channels, output channels, kernel size, bias, and dilation. Grouped convolution is supported with an optional groups parameter.

```cpp
inline Conv1D(const int in_channels, const int out_channels, const int kernel_size, const int bias, const int dilation, const int groups = 1)
```

--------------------------------

### Process_ Method (In-place)

Source: https://neuralampmodelercore.readthedocs.io/en/latest/api/film.html

Processes the input matrix in-place using the conditioning matrix. Modifies the input matrix directly.

```cpp
inline void Process_(Eigen::Ref<Eigen::MatrixXf> input, const Eigen::Ref<const Eigen::MatrixXf> &condition, const int num_frames)
{
    // ... (implementation details)
}

```

--------------------------------

### ConvNetBlock set_weights_ Method

Source: https://neuralampmodelercore.readthedocs.io/en/latest/api/convnet.html

Sets the parameters (weights) for a ConvNetBlock.

```cpp
void set_weights_(const int in_channels, const int out_channels, const int _dilation, const bool batchnorm, const activations::ActivationConfig &activation_config, const int groups, std::vector<float>::iterator &weights)
```

--------------------------------

### SetMaxBufferSize Function

Source: https://neuralampmodelercore.readthedocs.io/en/latest/api/get_dsp.html

Resizes internal buffers to accommodate a specified maximum number of frames.

```cpp
inline void SetMaxBufferSize(const int maxBufferSize)

    
Resize buffers to handle maxBufferSize frames.

Parameters:

**maxBufferSize** – Maximum number of frames to process in a single call
```

--------------------------------

### SlimmableModel Class

Source: https://neuralampmodelercore.readthedocs.io/en/latest/api/get_dsp.html

Interface for models that support dynamic size reduction. Models implementing this interface can reduce their computational cost at the expense of quality. The interpretation of the size parameter is model-specific (e.g., selecting a sub-model, pruning channels, etc.).

```C++
#include <slimmable.h>

class SlimmableModel {
public:
    virtual ~SlimmableModel() = default;
    virtual void SetSlimmableSize(const double val) = 0;
};
```

--------------------------------

### Set Size Parameters

Source: https://neuralampmodelercore.readthedocs.io/en/latest/api/conv1d.html

Configures the size parameters of the convolution layer, including input/output channels, kernel size, bias usage, dilation factor, and number of groups for grouped convolution.

```cpp
void set_size_(const int in_channels, const int out_channels, const int kernel_size, const bool do_bias, const int _dilation, const int groups = 1)
```
Tokens:
Raw