未验证 提交 d969c309 编写于 作者: R ronnywang 提交者: GitHub

add custom profier doc (#48629)

* add custom profier doc

* update

* format; test=document_fix
Co-authored-by: NLigoml <39876205+Ligoml@users.noreply.github.com>
上级 493825a5
...@@ -31,17 +31,14 @@ from paddle.fluid.core import ( ...@@ -31,17 +31,14 @@ from paddle.fluid.core import (
enable_memory_recorder, enable_memory_recorder,
enable_op_info_recorder, enable_op_info_recorder,
) )
from paddle.profiler import utils
from .utils import RecordEvent, wrap_optimizers
from .profiler_statistic import ( from .profiler_statistic import (
SortedKeys,
StatisticData, StatisticData,
_build_table, _build_table,
SortedKeys,
gen_layer_flops, gen_layer_flops,
) )
from paddle.profiler import utils
from .profiler_statistic import SortedKeys, StatisticData, _build_table
from .timer import benchmark from .timer import benchmark
from .utils import RecordEvent, wrap_optimizers from .utils import RecordEvent, wrap_optimizers
...@@ -359,6 +356,7 @@ class Profiler: ...@@ -359,6 +356,7 @@ class Profiler:
be timed and profiled. Default: False. be timed and profiled. Default: False.
record_shapes (bool, optional): If it is True, collect op's input shape information. Default: False. record_shapes (bool, optional): If it is True, collect op's input shape information. Default: False.
profile_memory (bool, optional): If it is True, collect tensor memory allocation and release information. Default: False. profile_memory (bool, optional): If it is True, collect tensor memory allocation and release information. Default: False.
custom_device_types (list, optional): If targets contain profiler.ProfilerTarget.CUSTOM_DEVICE, custom_device_types select the custom device type for profiling. The default value represents all custom devices will be selected.
with_flops (bool, optional): If it is True, the flops of the op will be calculated. Default: False. with_flops (bool, optional): If it is True, the flops of the op will be calculated. Default: False.
Examples: Examples:
......
Markdown is supported
0% .
You are about to add 0 people to the discussion. Proceed with caution.
先完成此消息的编辑!
想要评论请 注册