ludus::graphics::renderer::LightTransformerModelDescription¶

Struct

Borrowed trained-model setup for the executable Light Transformer graph template. Initialize consumes/copies assets. Shader artifacts survive through graph Poll. No model files, Python runtime or download is required at rendering time.

Declared in modules/graphics/renderer/include/ludus/graphics/renderer/light_transformer_encoder.hpp

#include <ludus/graphics/renderer/light_transformer_encoder.hpp>

struct ludus::graphics::renderer::LightTransformerModelDescription final

Fields¶

EncoderAsset¶

foundation::BorrowedView<const foundation::uint8> ludus::graphics::renderer::LightTransformerModelDescription::EncoderAsset

Offline --encoder-output asset.

Related types: foundation::BorrowedView · foundation::uint8

Declared in modules/graphics/renderer/include/ludus/graphics/renderer/light_transformer_encoder.hpp

DecoderAsset¶

foundation::BorrowedView<const foundation::uint8> ludus::graphics::renderer::LightTransformerModelDescription::DecoderAsset

Offline --decoder-output asset.

Related types: foundation::BorrowedView · foundation::uint8

Declared in modules/graphics/renderer/include/ludus/graphics/renderer/light_transformer_encoder.hpp

EncoderShader¶

rhi::RasterShaderDescription ludus::graphics::renderer::LightTransformerModelDescription::EncoderShader {}

Synchronous renderer_light_transformer_encoder recipe.

Related types: rhi::RasterShaderDescription

Declared in modules/graphics/renderer/include/ludus/graphics/renderer/light_transformer_encoder.hpp

WorldUnitsPerModelUnit¶

foundation::float32 ludus::graphics::renderer::LightTransformerModelDescription::WorldUnitsPerModelUnit =
0

Required positive fixed training scale; never per-scene fitting.

Related types: foundation::float32

Declared in modules/graphics/renderer/include/ludus/graphics/renderer/light_transformer_encoder.hpp

ScenePoints¶

foundation::uint32 ludus::graphics::renderer::LightTransformerModelDescription::ScenePoints = LUDUS_LIGHT_ENCODER_MAX_POINTS

Surface samples, 128..20000.

Related types: foundation::uint32

Declared in modules/graphics/renderer/include/ludus/graphics/renderer/light_transformer_encoder.hpp

Seed¶

foundation::uint64 ludus::graphics::renderer::LightTransformerModelDescription::Seed = 1

Stable surface sampling stream; unchanged by camera movement.

Related types: foundation::uint64

Declared in modules/graphics/renderer/include/ludus/graphics/renderer/light_transformer_encoder.hpp

MaximumBufferBytes¶

foundation::usize ludus::graphics::renderer::LightTransformerModelDescription::MaximumBufferBytes =
0

Required logical model GPU buffer budget; separate RHI device limits still apply.

Related types: foundation::usize

Declared in modules/graphics/renderer/include/ludus/graphics/renderer/light_transformer_encoder.hpp

MaximumSurfacePreparationBytes¶

foundation::usize ludus::graphics::renderer::LightTransformerModelDescription::MaximumSurfacePreparationBytes = 0

Required independent CPU surface scene/sampling budget.

Related types: foundation::usize

Declared in modules/graphics/renderer/include/ludus/graphics/renderer/light_transformer_encoder.hpp

MaximumSamplingAttempts¶

foundation::uint64 ludus::graphics::renderer::LightTransformerModelDescription::MaximumSamplingAttempts =
foundation::uint64{LUDUS_LIGHT_ENCODER_MAX_POINTS} *
LUDUS_LIGHT_ENCODER_DEFAULT_ATTEMPTS_PER_POINT

Surface proposal work limit, at least ScenePoints.

Related types: foundation::uint64

Declared in modules/graphics/renderer/include/ludus/graphics/renderer/light_transformer_encoder.hpp