infini::ops::AwqMarlinRepack Class Reference

#include <awq_marlin_repack.h>

Inheritance diagram for infini::ops::AwqMarlinRepack:

Public Member Functions
	AwqMarlinRepack (const Tensor b_q_weight, const int64_t size_k, const int64_t size_n, const int64_t num_bits, const bool is_a_8bit, Tensor out)
virtual void	operator() (const Tensor b_q_weight, const int64_t size_k, const int64_t size_n, const int64_t num_bits, const bool is_a_8bit, Tensor out) const =0
Public Member Functions inherited from infini::ops::Operator< AwqMarlinRepack >
void	operator() (const Handle &handle, const Args &... args)
void	operator() (const Args &... args) const
void	operator() (const Handle &handle, const Args &... args)
void	operator() (const Args &... args) const
Public Member Functions inherited from infini::ops::OperatorBase
virtual	~OperatorBase ()=default
virtual std::size_t	workspace_size_in_bytes () const
void	set_handle (const Handle &handle)
void	set_config (const Config &config)
void	set_stream (void *stream)
void	set_workspace (void *workspace)
void	set_workspace_size_in_bytes (std::size_t workspace_size_in_bytes)
virtual	~OperatorBase ()=default
virtual std::size_t	workspace_size_in_bytes () const
void	set_handle (const Handle &handle)
void	set_config (const Config &config)
void	set_stream (void *stream)
void	set_workspace (void *workspace)
void	set_workspace_size_in_bytes (std::size_t workspace_size_in_bytes)

Protected Member Functions
void	ValidateCallMetadata (const Tensor b_q_weight, const int64_t size_k, const int64_t size_n, const int64_t num_bits, const bool is_a_8bit, const Tensor out) const

Protected Attributes
Tensor	b_q_weight_metadata_
Tensor	out_metadata_
int64_t	size_k_ {0}
int64_t	size_n_ {0}
int64_t	num_bits_ {0}
int64_t	pack_factor_ {0}
bool	is_a_8bit_ {false}
int	device_index_ {0}
Protected Attributes inherited from infini::ops::OperatorBase
std::unique_ptr< Handle >	handle_ptr_
std::unique_ptr< Config >	config_ptr_
void *	stream_ {nullptr}
void *	workspace_ {nullptr}
std::size_t	workspace_size_in_bytes_ {0}

Additional Inherited Members
Static Public Member Functions inherited from infini::ops::Operator< AwqMarlinRepack >
static void	clear_cache ()
static void	clear_cache ()
static std::unique_ptr< Operator >	Make (const Config &config, const Tensor tensor, Args &&... args)
static std::unique_ptr< Operator >	Make (const Tensor tensor, Args &&... args)
static std::unique_ptr< Operator >	Make (const Config &config, const std::vector< Tensor > tensors, Args &&... args)
static std::unique_ptr< Operator >	Make (const std::vector< Tensor > tensors, Args &&... args)
static std::unique_ptr< Operator >	Make (const Config &config, const Tensor tensor, Args &&... args)
static std::unique_ptr< Operator >	Make (const Tensor tensor, Args &&... args)
static std::unique_ptr< Operator >	Make (const Config &config, const std::vector< Tensor > tensors, Args &&... args)
static std::unique_ptr< Operator >	Make (const std::vector< Tensor > tensors, Args &&... args)
static void	Call (const Handle &handle, const Config &config, const Args &... args)
static void	Call (const Tensor tensor, const Args &... args)
static auto	Call (const TensorLike &tensor, const Args &... args)
static void	Call (const Handle &handle, const Config &config, const Args &... args)
static void	Call (const Tensor tensor, const Args &... args)
static auto	Call (const TensorLike &tensor, const Args &... args)
static std::vector< std::size_t >	active_implementation_indices (Device::Type dev_type)
static std::vector< std::size_t >	active_implementation_indices (Device::Type dev_type)
Static Protected Attributes inherited from infini::ops::Operator< AwqMarlinRepack >
static constexpr Device::Type	device_type_
static constexpr std::size_t	implementation_index_

Constructor & Destructor Documentation

AwqMarlinRepack()

infini::ops::AwqMarlinRepack::AwqMarlinRepack ( const Tensor  b_q_weight, const int64_t  size_k, const int64_t  size_n, const int64_t  num_bits, const bool  is_a_8bit, Tensor  out  )

inline

Member Function Documentation

operator()()

virtual void infini::ops::AwqMarlinRepack::operator() ( const Tensor  b_q_weight, const int64_t  size_k, const int64_t  size_n, const int64_t  num_bits, const bool  is_a_8bit, Tensor  out  ) const

pure virtual

ValidateCallMetadata()

void infini::ops::AwqMarlinRepack::ValidateCallMetadata ( const Tensor  b_q_weight, const int64_t  size_k, const int64_t  size_n, const int64_t  num_bits, const bool  is_a_8bit, const Tensor  out  ) const

inlineprotected

Member Data Documentation

b_q_weight_metadata_

Tensor infini::ops::AwqMarlinRepack::b_q_weight_metadata_

protected

device_index_

int infini::ops::AwqMarlinRepack::device_index_ {0}

protected

is_a_8bit_

bool infini::ops::AwqMarlinRepack::is_a_8bit_ {false}

protected

num_bits_

int64_t infini::ops::AwqMarlinRepack::num_bits_ {0}

protected

out_metadata_

Tensor infini::ops::AwqMarlinRepack::out_metadata_

protected

pack_factor_

int64_t infini::ops::AwqMarlinRepack::pack_factor_ {0}

protected

size_k_

int64_t infini::ops::AwqMarlinRepack::size_k_ {0}

protected

size_n_

int64_t infini::ops::AwqMarlinRepack::size_n_ {0}

protected

---

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

• awq_marlin_repack.h