quark.torch.quantization

Submodules

• quark.torch.quantization.api

Package Contents

Classes

Config	A class that encapsulates comprehensive quantization configurations for a machine learning model, allowing for detailed and hierarchical control over quantization parameters across different model components.
QuantizationConfig	A data class that specifies quantization configurations for different components of a module, allowing hierarchical control over how each tensor type is quantized.
Uint4PerTensorSpec	Helper class that provides a standard way to create an ABC using
Uint4PerChannelSpec	Helper class that provides a standard way to create an ABC using
Uint4PerGroupSpec	Helper class that provides a standard way to create an ABC using
Int4PerTensorSpec	Helper class that provides a standard way to create an ABC using
Int4PerChannelSpec	Helper class that provides a standard way to create an ABC using
Int4PerGroupSpec	Helper class that provides a standard way to create an ABC using
Uint8PerTensorSpec	Helper class that provides a standard way to create an ABC using
Uint8PerChannelSpec	Helper class that provides a standard way to create an ABC using
Uint8PerGroupSpec	Helper class that provides a standard way to create an ABC using
Int8PerTensorSpec	Helper class that provides a standard way to create an ABC using
Int8PerChannelSpec	Helper class that provides a standard way to create an ABC using
Int8PerGroupSpec	Helper class that provides a standard way to create an ABC using
FP8E4M3PerTensorSpec	Helper class that provides a standard way to create an ABC using
FP8E4M3PerChannelSpec	Helper class that provides a standard way to create an ABC using
FP8E4M3PerGroupSpec	Helper class that provides a standard way to create an ABC using
FP8E5M2PerTensorSpec	Helper class that provides a standard way to create an ABC using
FP8E5M2PerChannelSpec	Helper class that provides a standard way to create an ABC using
FP8E5M2PerGroupSpec	Helper class that provides a standard way to create an ABC using
Float16Spec	Helper class that provides a standard way to create an ABC using
Bfloat16Spec	Helper class that provides a standard way to create an ABC using
MXSpec	Helper class that provides a standard way to create an ABC using
MX6Spec	Helper class that provides a standard way to create an ABC using
MX9Spec	Helper class that provides a standard way to create an ABC using
BFP16Spec	Helper class that provides a standard way to create an ABC using
QuantizationSpec	A data class that defines the specifications for quantizing tensors within a model.
AWQConfig	Configuration for Activation-aware Weight Quantization (AWQ).
GPTQConfig	A data class that defines the specifications for Accurate Post-Training Quantization for Generative Pre-trained Transformers (GPTQ).
RotationConfig	A data class that defines the specifications for rotation settings in processing algorithms.
SmoothQuantConfig	A data class that defines the specifications for Smooth Quantization.
AutoSmoothQuantConfig	A data class that defines the specifications for AutoSmoothQuant.
QuaRotConfig	A data class that defines the specifications for the QuaRot algorithm.

Functions

load_pre_optimization_config_from_file(→ PreQuantOptConfig)	Load pre-optimization configuration from a JSON file.
load_quant_algo_config_from_file(→ AlgoConfig)	Load quantization algorithm configuration from a JSON file.