mlir::quantizer::AttributeTensorStatistics Class Reference

#include <Statistics.h>

Inheritance diagram for mlir::quantizer::AttributeTensorStatistics:

Collaboration diagram for mlir::quantizer::AttributeTensorStatistics:

Public Member Functions
	AttributeTensorStatistics (Attribute attr)
bool	get (TensorAxisStatistics &stats) const override
Public Member Functions inherited from mlir::quantizer::AbstractTensorStatistics
virtual	~AbstractTensorStatistics ()=default
virtual bool	supportsPerAxis () const
virtual unsigned	getAxisCount () const
	Count of axes supported in a per-axis query. More...
virtual bool	getForAxis (unsigned axis, TensorAxisStatistics &stats) const

Detailed Description

Wraps an MLIR Attribute and returns statistics about it. It is expected that the attribute be one of: FloatAttr (scalar) DenseFPElementsAttr OpaqueElementsAttr (with Float based type) SparseElementAttr (with Float based type)

Constructor & Destructor Documentation

AttributeTensorStatistics()

mlir::quantizer::AttributeTensorStatistics::AttributeTensorStatistics ( Attribute  attr )

inline

Member Function Documentation

get()

bool AttributeTensorStatistics::get ( TensorAxisStatistics &  stats ) const

overridevirtual

Gets statistics across the whole tensor. Returns true if statistics are valid and were populated.

Reimplemented from mlir::quantizer::AbstractTensorStatistics.

---

The documentation for this class was generated from the following files:

• include/mlir/Quantizer/Support/Statistics.h
• lib/Quantizer/Support/Statistics.cpp