# API Reference ## coreai_opt coreai_opt - A library for PyTorch model compression and optimizations. | [`coreai_opt.CoreMLExportError`](generated/coreai_opt.CoreMLExportError.md#coreai_opt.CoreMLExportError)(dtype, context) | Raised when a model cannot be exported to the CoreML backend. | |----------------------------------------------------------------------------------------------------------------------------|-----------------------------------------------------------------| | [`coreai_opt.ExportBackend`](generated/coreai_opt.ExportBackend.md#coreai_opt.ExportBackend)(value, \*args, \*\*kwargs) | Enum representing supported model export backends. | ## coreai_opt.casting Casting related utilities including FP32 -> FP16 and INT32 -> INT16 passes. | [`coreai_opt.casting.cast_fp32_to_fp16`](generated/coreai_opt.casting.cast_fp32_to_fp16.md#coreai_opt.casting.cast_fp32_to_fp16)(...) | Convert a torch exported program from FP32 to FP16 where applicable. | |------------------------------------------------------------------------------------------------------------------------------------------------------------|-------------------------------------------------------------------------------------| | [`coreai_opt.casting.cast_int32_to_int16`](generated/coreai_opt.casting.cast_int32_to_int16.md#coreai_opt.casting.cast_int32_to_int16)(...) | Convert INT32/INT64 tensors to INT16 in a torch exported program. | | [`coreai_opt.casting.cast_to_16_bit_precision`](generated/coreai_opt.casting.cast_to_16_bit_precision.md#coreai_opt.casting.cast_to_16_bit_precision)(...) | Convert a torch exported program to 16-bit precision: FP32→FP16 and INT32/64→INT16. | ## coreai_opt.config Configuration and specification modules for coreai_opt. | [`coreai_opt.config.CompressionConfig`](generated/coreai_opt.config.CompressionConfig.md#coreai_opt.config.CompressionConfig) | Top level configuration class for model compression. | |---------------------------------------------------------------------------------------------------------------------------------------------------------------------------|-------------------------------------------------------------------------------| | [`coreai_opt.config.CompressionSpec`](generated/coreai_opt.config.CompressionSpec.md#coreai_opt.config.CompressionSpec) | Base class for compression specifications. | | [`coreai_opt.config.CompressionType`](generated/coreai_opt.config.CompressionType.md#coreai_opt.config.CompressionType)(value[, ...]) | Enum representing compression techniques applied to the model. | | [`coreai_opt.config.ModuleCompressionConfig`](generated/coreai_opt.config.ModuleCompressionConfig.md#coreai_opt.config.ModuleCompressionConfig) | Abstract base configuration class for module-level compression settings. | | [`coreai_opt.config.OpCompressionConfig`](generated/coreai_opt.config.OpCompressionConfig.md#coreai_opt.config.OpCompressionConfig) | Abstract base configuration class for op-level compression settings. | | [`coreai_opt.config.WeightOnlyModuleValidationMixin`](generated/coreai_opt.config.WeightOnlyModuleValidationMixin.md#coreai_opt.config.WeightOnlyModuleValidationMixin)() | Mixin that adds weight-only validation to ModuleCompressionConfig subclasses. | | [`coreai_opt.config.WeightOnlyOpValidationMixin`](generated/coreai_opt.config.WeightOnlyOpValidationMixin.md#coreai_opt.config.WeightOnlyOpValidationMixin)() | Mixin that adds weight-only validation to OpCompressionConfig subclasses. | ### coreai_opt.config.spec Base abstractions for compression specs, simulators, and component factories. | [`coreai_opt.config.spec.CompressionComponentFactoryBase`](generated/coreai_opt.config.spec.CompressionComponentFactoryBase.md#coreai_opt.config.spec.CompressionComponentFactoryBase)() | Abstract base class for compression component factories. | |--------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------|------------------------------------------------------------| | [`coreai_opt.config.spec.CompressionSimulatorBase`](generated/coreai_opt.config.spec.CompressionSimulatorBase.md#coreai_opt.config.spec.CompressionSimulatorBase)(...) | Abstract base class for compression simulators. | | [`coreai_opt.config.spec.CompressionTargetTensor`](generated/coreai_opt.config.spec.CompressionTargetTensor.md#coreai_opt.config.spec.CompressionTargetTensor)(value) | Enum to specify the target tensor for compression. | ## coreai_opt.coreai_utils Core AI MLIR-level compression transforms. | [`coreai_opt.coreai_utils.CompressionGranularity`](generated/coreai_opt.coreai_utils.CompressionGranularity.md#coreai_opt.coreai_utils.CompressionGranularity)(value) | Enum representing the granularity of quantization for Core AI weight compression. | |-------------------------------------------------------------------------------------------------------------------------------------------------------------------------|-------------------------------------------------------------------------------------| | [`coreai_opt.coreai_utils.DType`](generated/coreai_opt.coreai_utils.DType.md#coreai_opt.coreai_utils.DType)(value[, ...]) | Enum representing data types for Core AI weight compression. | | [`coreai_opt.coreai_utils.palettize_weights`](generated/coreai_opt.coreai_utils.palettize_weights.md#coreai_opt.coreai_utils.palettize_weights)(...) | Palettize weights in a Core AI AIProgram (MLIR IR) by using Core AI ops. | | [`coreai_opt.coreai_utils.quantize_weights`](generated/coreai_opt.coreai_utils.quantize_weights.md#coreai_opt.coreai_utils.quantize_weights)(...) | Quantize weights in a Core AI AIProgram (MLIR IR) by using Core AI ops. | | [`coreai_opt.coreai_utils.sparsify_weights`](generated/coreai_opt.coreai_utils.sparsify_weights.md#coreai_opt.coreai_utils.sparsify_weights)(...) | Sparsify weights in a Core AI AIProgram (MLIR IR) by using Core AI ops. | ### coreai_opt.coreai_utils.common Common enums and constants for coreai_opt.coreai_utils. | [`coreai_opt.coreai_utils.common.QScheme`](generated/coreai_opt.coreai_utils.common.QScheme.md#coreai_opt.coreai_utils.common.QScheme)(value) | Enum representing the quantization scheme. | |-------------------------------------------------------------------------------------------------------------------------------------------------|----------------------------------------------| ## coreai_opt.inspection Utilities for inspecting model operations and compression configuration. | [`coreai_opt.inspection.BoundaryEdge`](generated/coreai_opt.inspection.BoundaryEdge.md#coreai_opt.inspection.BoundaryEdge)(op, index) | A single data-flow edge crossing a module boundary. | |----------------------------------------------------------------------------------------------------------------------------------------------|---------------------------------------------------------------------------| | [`coreai_opt.inspection.InputEdge`](generated/coreai_opt.inspection.InputEdge.md#coreai_opt.inspection.InputEdge)(op, output_idx) | One input edge into an op, pairing the producing op with its output slot. | | [`coreai_opt.inspection.ModelInspector`](generated/coreai_opt.inspection.ModelInspector.md#coreai_opt.inspection.ModelInspector)(model, ...) | Inspect operations in a PyTorch model for compression configuration. | | [`coreai_opt.inspection.ModelSummary`](generated/coreai_opt.inspection.ModelSummary.md#coreai_opt.inspection.ModelSummary)(model, mode) | Complete listing of operations discovered in a model. | | [`coreai_opt.inspection.ModuleContext`](generated/coreai_opt.inspection.ModuleContext.md#coreai_opt.inspection.ModuleContext)(...) | One level of the `nn.Module` nesting hierarchy. | | [`coreai_opt.inspection.ModuleInfo`](generated/coreai_opt.inspection.ModuleInfo.md#coreai_opt.inspection.ModuleInfo)(...) | A node in the `nn.Module` hierarchy with its directly-owned ops. | | [`coreai_opt.inspection.OpInfo`](generated/coreai_opt.inspection.OpInfo.md#coreai_opt.inspection.OpInfo)(op_name, ...) | Information about a single operation discovered in a model. | | [`coreai_opt.inspection.SourceFrame`](generated/coreai_opt.inspection.SourceFrame.md#coreai_opt.inspection.SourceFrame)(filename, ...) | A single frame in the source call stack leading to an operation. | ## coreai_opt.palettization Palettization specification and utilities for weight compression via lookup tables. | [`coreai_opt.palettization.KMeansPalettizer`](generated/coreai_opt.palettization.KMeansPalettizer.md#coreai_opt.palettization.KMeansPalettizer)(model) | K-means palettizer with integrated supported operations strategy. | |-------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------|---------------------------------------------------------------------------| | [`coreai_opt.palettization.KMeansPalettizerConfig`](generated/coreai_opt.palettization.KMeansPalettizerConfig.md#coreai_opt.palettization.KMeansPalettizerConfig) | Top-level configuration class for kmeans palettization. | | [`coreai_opt.palettization.ModuleKMeansPalettizerConfig`](generated/coreai_opt.palettization.ModuleKMeansPalettizerConfig.md#coreai_opt.palettization.ModuleKMeansPalettizerConfig) | Configuration for palettizing a specific module using K-means clustering. | | [`coreai_opt.palettization.PalettizationSpec`](generated/coreai_opt.palettization.PalettizationSpec.md#coreai_opt.palettization.PalettizationSpec) | Specification for palettization compression of neural network weights. | ### coreai_opt.palettization.config Palettization configuration classes. | [`coreai_opt.palettization.config.OpKMeansPalettizerConfig`](generated/coreai_opt.palettization.config.OpKMeansPalettizerConfig.md#coreai_opt.palettization.config.OpKMeansPalettizerConfig) | Configuration class for palettization at the operation level. | |------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------|-----------------------------------------------------------------| ### coreai_opt.palettization.spec Palettization specs, granularity classes, and factory functions. | [`coreai_opt.palettization.spec.PalettizationGranularity`](generated/coreai_opt.palettization.spec.PalettizationGranularity.md#coreai_opt.palettization.spec.PalettizationGranularity) | Base class for palettization granularity specifications. | |---------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------|------------------------------------------------------------| | [`coreai_opt.palettization.spec.PerGroupedChannelGranularity`](generated/coreai_opt.palettization.spec.PerGroupedChannelGranularity.md#coreai_opt.palettization.spec.PerGroupedChannelGranularity) | Per-grouped-channel palettization granularity. | | [`coreai_opt.palettization.spec.PerTensorGranularity`](generated/coreai_opt.palettization.spec.PerTensorGranularity.md#coreai_opt.palettization.spec.PerTensorGranularity) | Per-tensor palettization granularity. | | [`coreai_opt.palettization.spec.default_weight_palettization_spec`](generated/coreai_opt.palettization.spec.default_weight_palettization_spec.md#coreai_opt.palettization.spec.default_weight_palettization_spec)() | | ## coreai_opt.pruning Pruning infrastructure for coreai_opt. | [`coreai_opt.pruning.MagnitudePruner`](generated/coreai_opt.pruning.MagnitudePruner.md#coreai_opt.pruning.MagnitudePruner)(model[, ...]) | Apply magnitude-based pruning to a model. | |----------------------------------------------------------------------------------------------------------------------------------------------------------------|------------------------------------------------| | [`coreai_opt.pruning.MagnitudePrunerConfig`](generated/coreai_opt.pruning.MagnitudePrunerConfig.md#coreai_opt.pruning.MagnitudePrunerConfig) | Top-level configuration for magnitude pruning. | | [`coreai_opt.pruning.ModuleMagnitudePrunerConfig`](generated/coreai_opt.pruning.ModuleMagnitudePrunerConfig.md#coreai_opt.pruning.ModuleMagnitudePrunerConfig) | Module-level pruning configuration. | | [`coreai_opt.pruning.PruningSpec`](generated/coreai_opt.pruning.PruningSpec.md#coreai_opt.pruning.PruningSpec) | Specification for pruning tensors. | ### coreai_opt.pruning.config Pruning configuration exports. | [`coreai_opt.pruning.config.ConstantSparsitySchedule`](generated/coreai_opt.pruning.config.ConstantSparsitySchedule.md#coreai_opt.pruning.config.ConstantSparsitySchedule) | Step function: zero before `begin_step`, `target_sparsity` at and after. | |------------------------------------------------------------------------------------------------------------------------------------------------------------------------------|----------------------------------------------------------------------------| | [`coreai_opt.pruning.config.OpMagnitudePrunerConfig`](generated/coreai_opt.pruning.config.OpMagnitudePrunerConfig.md#coreai_opt.pruning.config.OpMagnitudePrunerConfig) | Operation-level pruning configuration. | | [`coreai_opt.pruning.config.PolynomialDecaySchedule`](generated/coreai_opt.pruning.config.PolynomialDecaySchedule.md#coreai_opt.pruning.config.PolynomialDecaySchedule) | Polynomial schedule from `initial_sparsity` to `target_sparsity`. | | [`coreai_opt.pruning.config.SparsityScheduleBase`](generated/coreai_opt.pruning.config.SparsityScheduleBase.md#coreai_opt.pruning.config.SparsityScheduleBase) | Abstract base for sparsity schedules used by `MagnitudePruner`. | ### coreai_opt.pruning.spec Pruning spec components: specs, schemes, and parametrizations. | [`coreai_opt.pruning.spec.ChannelStructured`](generated/coreai_opt.pruning.spec.ChannelStructured.md#coreai_opt.pruning.spec.ChannelStructured) | Channel-structured pruning scheme. | |---------------------------------------------------------------------------------------------------------------------------------------------------------------------------------|------------------------------------------------------------------------| | [`coreai_opt.pruning.spec.PruneImplBase`](generated/coreai_opt.pruning.spec.PruneImplBase.md#coreai_opt.pruning.spec.PruneImplBase)(...) | Abstract base for pruning parametrizations that mask a layer's weight. | | [`coreai_opt.pruning.spec.PruningScheme`](generated/coreai_opt.pruning.spec.PruningScheme.md#coreai_opt.pruning.spec.PruningScheme) | Base class for pruning scheme specifications. | | [`coreai_opt.pruning.spec.Unstructured`](generated/coreai_opt.pruning.spec.Unstructured.md#coreai_opt.pruning.spec.Unstructured) | Unstructured pruning scheme. | | [`coreai_opt.pruning.spec.default_weight_pruning_spec`](generated/coreai_opt.pruning.spec.default_weight_pruning_spec.md#coreai_opt.pruning.spec.default_weight_pruning_spec)() | Return the default pruning spec for weight tensors. | ## coreai_opt.quantization Quantization compressor, configuration, specs, and granularity classes. | [`coreai_opt.quantization.ExecutionMode`](generated/coreai_opt.quantization.ExecutionMode.md#coreai_opt.quantization.ExecutionMode)(value, ...) | Enum representing quantization execution modes. | |-------------------------------------------------------------------------------------------------------------------------------------------------------------|---------------------------------------------------------------------------------------------------------| | [`coreai_opt.quantization.ModuleQuantizerConfig`](generated/coreai_opt.quantization.ModuleQuantizerConfig.md#coreai_opt.quantization.ModuleQuantizerConfig) | Configuration class for quantization at the module level. | | [`coreai_opt.quantization.QuantizationSpec`](generated/coreai_opt.quantization.QuantizationSpec.md#coreai_opt.quantization.QuantizationSpec) | Specification for quantizing tensors in neural networks. | | [`coreai_opt.quantization.Quantizer`](generated/coreai_opt.quantization.Quantizer.md#coreai_opt.quantization.Quantizer)(model[, ...]) | Unified quantizer API that provides a single entry point for various quantization workflows, including: | | [`coreai_opt.quantization.QuantizerConfig`](generated/coreai_opt.quantization.QuantizerConfig.md#coreai_opt.quantization.QuantizerConfig) | Top-level configuration class for quantization. | ### coreai_opt.quantization.config Quantization configuration classes and execution mode. | [`coreai_opt.quantization.config.KVCacheQuantConfig`](generated/coreai_opt.quantization.config.KVCacheQuantConfig.md#coreai_opt.quantization.config.KVCacheQuantConfig) | Enables KV-cache buffer storage in a quantized dtype for one cache-update op type. | |---------------------------------------------------------------------------------------------------------------------------------------------------------------------------|--------------------------------------------------------------------------------------| | [`coreai_opt.quantization.config.OpQuantizerConfig`](generated/coreai_opt.quantization.config.OpQuantizerConfig.md#coreai_opt.quantization.config.OpQuantizerConfig) | Configuration class for quantization at the operation level. | | [`coreai_opt.quantization.config.QATSchedule`](generated/coreai_opt.quantization.config.QATSchedule.md#coreai_opt.quantization.config.QATSchedule) | Schedule for controlling observer and fake quantization state in QAT. | ### coreai_opt.quantization.spec Quantization specs, schemes, granularity classes, and parameter calculators. | [`coreai_opt.quantization.spec.DynamicQParamsCalculator`](generated/coreai_opt.quantization.spec.DynamicQParamsCalculator.md#coreai_opt.quantization.spec.DynamicQParamsCalculator)(...) | Dynamically computes scale/zero-point/minval from the current tensor every forward. | |---------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------|-----------------------------------------------------------------------------------------------------| | [`coreai_opt.quantization.spec.GlobalMinMaxQParamsCalculator`](generated/coreai_opt.quantization.spec.GlobalMinMaxQParamsCalculator.md#coreai_opt.quantization.spec.GlobalMinMaxQParamsCalculator)(...) | Computes scale and zero point by tracking the running min/max. | | [`coreai_opt.quantization.spec.MinMaxRangeCalculator`](generated/coreai_opt.quantization.spec.MinMaxRangeCalculator.md#coreai_opt.quantization.spec.MinMaxRangeCalculator)(...) | Range calculator that computes the range of a given tensor as the min and max values of the tensor. | | [`coreai_opt.quantization.spec.MovingAverageQParamsCalculator`](generated/coreai_opt.quantization.spec.MovingAverageQParamsCalculator.md#coreai_opt.quantization.spec.MovingAverageQParamsCalculator)(...) | Computes the scale and zero point using a moving average of the range. | | [`coreai_opt.quantization.spec.PerBlockGranularity`](generated/coreai_opt.quantization.spec.PerBlockGranularity.md#coreai_opt.quantization.spec.PerBlockGranularity) | Per-block quantization granularity. | | [`coreai_opt.quantization.spec.PerChannelGranularity`](generated/coreai_opt.quantization.spec.PerChannelGranularity.md#coreai_opt.quantization.spec.PerChannelGranularity) | Per-channel quantization granularity. | | [`coreai_opt.quantization.spec.PerTensorGranularity`](generated/coreai_opt.quantization.spec.PerTensorGranularity.md#coreai_opt.quantization.spec.PerTensorGranularity) | Per-tensor quantization granularity. | | [`coreai_opt.quantization.spec.QParamsCalculatorBase`](generated/coreai_opt.quantization.spec.QParamsCalculatorBase.md#coreai_opt.quantization.spec.QParamsCalculatorBase)(...) | Abstract base for qparams calculators — common configuration and helpers. | | [`coreai_opt.quantization.spec.QuantizationComponentFactory`](generated/coreai_opt.quantization.spec.QuantizationComponentFactory.md#coreai_opt.quantization.spec.QuantizationComponentFactory)() | Factory class for creating quantization components from QuantizationSpec. | | [`coreai_opt.quantization.spec.QuantizationFormulation`](generated/coreai_opt.quantization.spec.QuantizationFormulation.md#coreai_opt.quantization.spec.QuantizationFormulation)(value) | Formula used to map between quantized integers and dequantized values. | | [`coreai_opt.quantization.spec.QuantizationGranularity`](generated/coreai_opt.quantization.spec.QuantizationGranularity.md#coreai_opt.quantization.spec.QuantizationGranularity) | Base class for quantization granularity specifications. | | [`coreai_opt.quantization.spec.QuantizationScheme`](generated/coreai_opt.quantization.spec.QuantizationScheme.md#coreai_opt.quantization.spec.QuantizationScheme)(value) | | | [`coreai_opt.quantization.spec.RangeCalculatorBase`](generated/coreai_opt.quantization.spec.RangeCalculatorBase.md#coreai_opt.quantization.spec.RangeCalculatorBase)(...) | Base class and registry for classes used to compute the range of a given tensor. | | [`coreai_opt.quantization.spec.RunningRangeMixin`](generated/coreai_opt.quantization.spec.RunningRangeMixin.md#coreai_opt.quantization.spec.RunningRangeMixin)(...) | Mixin for calculators that maintain running min/max range buffers. | | [`coreai_opt.quantization.spec.StatefulQParamsCalculatorBase`](generated/coreai_opt.quantization.spec.StatefulQParamsCalculatorBase.md#coreai_opt.quantization.spec.StatefulQParamsCalculatorBase)(...) | Stateful base: maintains scale/zero_point/minval as nn.Module buffers across forwards. | | [`coreai_opt.quantization.spec.StatelessQParamsCalculatorBase`](generated/coreai_opt.quantization.spec.StatelessQParamsCalculatorBase.md#coreai_opt.quantization.spec.StatelessQParamsCalculatorBase)(...) | Stateless base: no cached qparams; recomputed every forward. | | [`coreai_opt.quantization.spec.StaticQParamsCalculator`](generated/coreai_opt.quantization.spec.StaticQParamsCalculator.md#coreai_opt.quantization.spec.StaticQParamsCalculator)(...) | Computes scale/zero-point/minval using min/max values from the current tensor. | | [`coreai_opt.quantization.spec.default_activation_quantization_spec`](generated/coreai_opt.quantization.spec.default_activation_quantization_spec.md#coreai_opt.quantization.spec.default_activation_quantization_spec)() | | | [`coreai_opt.quantization.spec.default_weight_quantization_spec`](generated/coreai_opt.quantization.spec.default_weight_quantization_spec.md#coreai_opt.quantization.spec.default_weight_quantization_spec)() | | ### coreai_opt.quantization.spec.fake_quantize Fake quantization implementation base class and default implementation. | [`coreai_opt.quantization.spec.fake_quantize.FakeQuantizeImplBase`](generated/coreai_opt.quantization.spec.fake_quantize.FakeQuantizeImplBase.md#coreai_opt.quantization.spec.fake_quantize.FakeQuantizeImplBase)(...) | Base class for implementing fake quantization | |--------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------|-------------------------------------------------|