Struct emulator::ModelIOConfig

ClassList > emulator > ModelIOConfig

Model I/O variable configuration. More...

• #include <emulator_config.hpp>

Public Attributes

Type	Name
std::vector< std::string >	input_variables Fields to pack as model input.
std::vector< std::string >	output_variables Fields to extract from output.
bool	spatial_mode = true Enable spatial mode for CNN-based models (e.g., ACE2).

Detailed Description

Specifies which fields are used as inputs and outputs for the AI model, and controls the data layout during inference.

Public Attributes Documentation

variable input_variables

Fields to pack as model input.

std::vector<std::string> emulator::ModelIOConfig::input_variables;

---

variable output_variables

Fields to extract from output.

std::vector<std::string> emulator::ModelIOConfig::output_variables;

---

variable spatial_mode

Enable spatial mode for CNN-based models (e.g., ACE2).

bool emulator::ModelIOConfig::spatial_mode;

When true (CNN mode):

• Input data is reshaped from [H*W, C] to [1, C, H, W] before inference
• Output data is reshaped from [1, C, H, W] to [H*W, C] after inference
• The entire grid is processed as a single sample (batch_size=1)

When false (pointwise/MLP mode):

• Data remains in [batch_size, channels] format
• Each grid point is processed independently (batch_size=H*W)

Note:

Default: true (assumes CNN model like ACE2)

---

---

The documentation for this class was generated from the following file components/emulator_comps/common/src/emulator_config.hpp