paddleslim.analysis package¶

paddleslim.analysis.flops(program, only_conv=True, detail=False)¶

Get FLOPs of target graph.

Parameters:	program (Program) – The program used to calculate FLOPS. only_conv (bool) – Just return number of mul-adds in convolution and FC layer if only_conv is true. default: True. detail (bool) – Whether to return detail of each convolution layer.
Returns:	If detail is true, then return a tuple in format (FLOPs, details), otherwise it will just return FlOPs. The details is a dict whose key is the parameter name of convlution layer and value is the FLOPs of each convolution layer.
Return type:	int\|tuple

paddleslim.analysis.model_size(program)¶

Get total value numbers of all parameters.

Parameters:	program (fluid.Program) – The program used to calculate model size.
Returns:	The total count of all parameters.
Return type:	int

class paddleslim.analysis.LatencyEvaluator¶

Bases: object

Base class of latency evaluator.

latency(graph)¶

Get latency of graph. It is an abstract method.

Parameters:	graph (GrapWrapper \| Program) – The graph to be evaluated.
Returns:	The latency of given graph on current evaluator.
Return type:	latency(float)

class paddleslim.analysis.TableLatencyEvaluator(table_file, delimiter=', ')¶

The evaluator used to get graph’s latency on some devices and infer engines.

Parameters:	table_file (str) – The path of file that records the devices latency of operators. delimiter (str) – The delimiter used in table_file.

latency(graph, only_conv=True)¶

Get latency of target graph.

Parameters:	graph (GrapWrapper \| Program) – The graph to be evaluated. only_conv (bool) – only evaluated convolution layer if only_conv is true. Default: True.
Returns:	The latency of given graph on current evaluator.
Return type:	latency(float)

Submodules¶

paddleslim.analysis.flops.flops(program, only_conv=True, detail=False)¶

Get FLOPs of target graph.

Parameters:	program (Program) – The program used to calculate FLOPS. only_conv (bool) – Just return number of mul-adds in convolution and FC layer if only_conv is true. default: True. detail (bool) – Whether to return detail of each convolution layer.
Returns:	If detail is true, then return a tuple in format (FLOPs, details), otherwise it will just return FlOPs. The details is a dict whose key is the parameter name of convlution layer and value is the FLOPs of each convolution layer.
Return type:	int\|tuple

Define latency evaluators that evaluate the performance of mode on devices.

class paddleslim.analysis.latency.LatencyEvaluator¶