Skip to main content

Module pinn_extension

Module pinn_extension 

Source
Expand description

PINN Extension for QualiaDB Advanced

Physics-Informed Neural Networks with SMX formatting and 1.58-bit ternary quantization for solving differential equations and continuous physical systems while maintaining core engine constraints.

This extension uses the native Qualia LLM pipeline (wgpu + WGSL shaders) for neural network inference, ensuring zero-allocation hot paths and GPU acceleration without external ML frameworks.

Structs§

BoundaryCondition
Boundary condition for PINN
ConvergenceMetrics
Convergence metrics for PINN
NativePinnBackend
Native Qualia LLM backend for PINN inference
PhysicsConstraint
Physics constraint for PINN
PhysicsViolation
Physics violation detected in PINN
PinnExecutionResult
PINN execution result with SMX formatting
PinnExtension
PINN Extension implementation with SMX formatting and ternary quantization
PinnJobParams
PINN execution parameters
QuantizationMetrics
Quantization metrics for PINN execution
SmxFormatter
SMX Formatter for structured model exchange
SmxMetadataSchema
SMX Metadata Schema for PINN models
SmxOutput
SMX output format for PINN results
TensorMetadata
Metadata for ternary tensors
TernaryPinnModel
Ternary Physics-Informed Neural Network model
TernaryPinnModelManager
Ternary PINN Model Manager with 1.58-bit quantization support
TernaryQuantizationConfig
Ternary Quantization Configuration for 1.58-bit models
TernaryTensor
Ternary tensor representation for 1.58-bit quantization
TrainingMetadata
Training metadata for SMX format

Enums§

BoundaryType
Types of boundary conditions
EquationType
Types of physics equations
PhysicsDomain
Physics domains supported by PINN models