neural_compressor.adaptor.query

Module Contents

Classes

QueryBackendCapability

Base class that defines Query Interface.

class neural_compressor.adaptor.query.QueryBackendCapability

Base class that defines Query Interface. Each adaption layer should implement the inherited class for specific backend on their own.

abstract get_version()

Get the current backend’s version string.

abstract get_precisions()

Get the supported low precisions, e.g [‘int8’, ‘bf16’]

abstract get_op_types()

Get the op types for specific backend per low precision. e.g {‘2.3.0’: {‘int8’: [‘Conv2D’, ‘MatMuL’]}}

abstract get_fuse_patterns()

Get the fusion patterns for specified op type for every specific precision

abstract set_quantization_config(q_config)

Set the quantization config to backend. :param q_config: set the organized quantization configuration to backend. :type q_config: yaml content?

abstract get_quantization_capability()

Get the quantization capability of low precision op types. e.g, granularity, scheme and etc.

abstract get_mixed_precision_combination(unsupported_precisions)

Get the valid precision combination base on hardware and user’ config. e.g[‘fp32’, ‘bf16’, ‘int8’]