Skip to content
体验新版
项目
组织
正在加载...
登录
切换导航
打开侧边栏
MegEngine 天元
MegEngine
提交
b22ceba7
MegEngine
项目概览
MegEngine 天元
/
MegEngine
1 年多 前同步成功
通知
404
Star
4705
Fork
582
代码
文件
提交
分支
Tags
贡献者
分支图
Diff
Issue
0
列表
看板
标记
里程碑
合并请求
0
DevOps
流水线
流水线任务
计划
Wiki
0
Wiki
分析
仓库
DevOps
项目成员
Pages
MegEngine
项目概览
项目概览
详情
发布
仓库
仓库
文件
提交
分支
标签
贡献者
分支图
比较
Issue
0
Issue
0
列表
看板
标记
里程碑
合并请求
0
合并请求
0
Pages
DevOps
DevOps
流水线
流水线任务
计划
分析
分析
仓库分析
DevOps
Wiki
0
Wiki
成员
成员
收起侧边栏
关闭侧边栏
动态
分支图
创建新Issue
流水线任务
提交
Issue看板
提交
b22ceba7
编写于
11月 25, 2021
作者:
M
Megvii Engine Team
浏览文件
操作
浏览文件
下载
电子邮件补丁
差异文件
docs(imperative): fix code exmaple indent warnings
GitOrigin-RevId: aa533bf2964b4b40ea366329da66bd47a2ca2e52
上级
7ac693a8
变更
3
隐藏空白更改
内联
并排
Showing
3 changed file
with
69 addition
and
53 deletion
+69
-53
imperative/python/megengine/tools/graph_info_analyze.py
imperative/python/megengine/tools/graph_info_analyze.py
+9
-5
imperative/python/megengine/traced_module/traced_module.py
imperative/python/megengine/traced_module/traced_module.py
+28
-28
imperative/python/megengine/utils/tensorboard.py
imperative/python/megengine/utils/tensorboard.py
+32
-20
未找到文件。
imperative/python/megengine/tools/graph_info_analyze.py
浏览文件 @
b22ceba7
...
...
@@ -205,13 +205,17 @@ def convert(args):
def
main
():
"""`graph_info_analyze.py` is uesed to convert json dumped by `VisableDataSet`
"""`graph_info_analyze.py` is uesed to convert json dumped by `VisableDataSet`
class to logs which can be read by python `tensorboard`.
Now `get_static_memory_alloc_info()` support this feature,it will dump a dir
Now `get_static_memory_alloc_info()` support this feature,it will dump a dir
which can be convert by `graph_info_analyze.py`.
Examples::
graph_info_analyze.py -i <input_dir_name> -o <output_dir_name>
tensorboard --logdir <output_dir_name>
Examples:
.. code-block:: shell
graph_info_analyze.py -i <input_dir_name> -o <output_dir_name>
tensorboard --logdir <output_dir_name>
"""
parser
=
argparse
.
ArgumentParser
(
"convert json dumped by c to logs which can be read by python tensorboard"
,
...
...
imperative/python/megengine/traced_module/traced_module.py
浏览文件 @
b22ceba7
...
...
@@ -495,7 +495,7 @@ class InternalGraph:
inp = F.zeros(shape = (3, 4))
traced_module = tm.trace_module(net, inp)
Will produce the following ``InternalGraph``::
print(traced_module.graph)
...
...
@@ -728,7 +728,7 @@ class InternalGraph:
# graph : InternalGraph
print("{:p}".format(graph))
print(graph.__format__("p"))
Args:
name: a string in glob syntax that can contain ``?`` and
``*`` to match a single or arbitrary characters.
...
...
@@ -749,7 +749,7 @@ class InternalGraph:
def
get_dep_exprs
(
self
,
nodes
:
Sequence
[
Node
])
->
List
[
Expr
]:
r
"""Get the dependent Exprs of the ``nodes``.
Args:
nodes: a list of :class:`Node`.
Returns:
...
...
@@ -817,7 +817,7 @@ class InternalGraph:
shape: the shape of the new input Node.
dtype: the dtype of the new input Node.
Default: float32
name: the name of the new input Node. When the name is used in the graph,
name: the name of the new input Node. When the name is used in the graph,
a suffix will be added to it.
"""
forma_mnode
=
self
.
inputs
[
0
]
...
...
@@ -850,16 +850,16 @@ class InternalGraph:
def
reset_outputs
(
self
,
outputs
):
r
"""Reset the output Nodes of the graph.
.. note::
This method only supports resetting the output of graphs
that do not have a parent graph.
Args:
outputs: an object which inner element is Node. Support tuple, list
dict, etc.
For example, the following code
.. code-block::
...
...
@@ -882,7 +882,7 @@ class InternalGraph:
out_node = graph.outputs[0]
graph.reset_outputs((out_node, {"input": inp_node}))
out = traced_module(inp)
Will produce the following ``InternalGraph`` and ``out``::
print(graph)
...
...
@@ -910,9 +910,9 @@ class InternalGraph:
def
add_output_node
(
self
,
node
:
TensorNode
):
r
"""Add an output node to the Graph.
The Graph output will become a ``tuple`` after calling ``add_output_node``.
The first element of the ``tuple`` is the original output, and the second
The first element of the ``tuple`` is the original output, and the second
is the ``node``.
For example, the following code
...
...
@@ -938,7 +938,7 @@ class InternalGraph:
graph.add_output_node(inp_node)
graph.add_output_node(out_node)
out = traced_module(inp)
Will produce the following ``InternalGraph`` and ``out``::
print(graph)
...
...
@@ -977,11 +977,11 @@ class InternalGraph:
... # inert exprs into graph and resotre normal mode
Args:
expr: the ``expr`` after which to insert. If None, the insertion position will be
expr: the ``expr`` after which to insert. If None, the insertion position will be
automatically set based on the input node.
Returns:
A resource manager that will initialize trace mode on ``__enter__`` and
A resource manager that will initialize trace mode on ``__enter__`` and
restore normal mode on ``__exit__``.
"""
if
expr
is
not
None
:
...
...
@@ -990,7 +990,7 @@ class InternalGraph:
def
replace_node
(
self
,
repl_dict
:
Dict
[
Node
,
Node
]):
r
"""Replace the Nodes in the graph.
Args:
repl_dict: the map {old_Node: new_Node} that specifies how to replace the Nodes.
"""
...
...
@@ -1752,7 +1752,7 @@ class BaseFilter:
class
ExprFilter
(
BaseFilter
):
"""Filter on Expr iterator.
This class is an iterator of :class:`.Expr` objects and multiple
This class is an iterator of :class:`.Expr` objects and multiple
filtering conditions and mappers can be chained.
"""
...
...
@@ -1777,7 +1777,7 @@ class ExprFilter(BaseFilter):
class
NodeFilter
(
BaseFilter
):
"""Filter on Node iterator.
This class is an iterator of :class:`
.Node` objects and multiple
This class is an iterator of :class:`
~.traced_module.Node` objects and multiple
filtering conditions and mappers can be chained.
"""
...
...
@@ -1905,14 +1905,14 @@ class ExprFilterExprId(ExprFilter):
class
TracedModule
(
Module
):
r
"""``TracedModule`` is the Module created by tracing normal module.
It owns an argdef to graph(InternalGraph) map. The forward method of ``TracedModule``
will get a graph from ``argdef_graph_map`` according to the argdef of input ``args/kwargs``
r
"""``TracedModule`` is the Module created by tracing normal module.
It owns an argdef to graph(InternalGraph) map. The forward method of ``TracedModule``
will get a graph from ``argdef_graph_map`` according to the argdef of input ``args/kwargs``
and interpret it.
.. note::
``TracedModule`` can only be created by :func:`~.trace_module`. See :func:`~.trace_module`
``TracedModule`` can only be created by :func:`~.trace_module`. See :func:`~.trace_module`
for more details.
"""
# m_node = None # type: ModuleNode
...
...
@@ -1956,10 +1956,10 @@ class TracedModule(Module):
r
"""Initialize the :attr:`~.TracedModule.watch_points`.
You can call this function to get the ``Tensor/Module`` corresponding to a ``Node`` at runtime.
Args:
nodes: a list of ``Node``.
For example, the following code
.. code-block::
...
...
@@ -1981,7 +1981,7 @@ class TracedModule(Module):
traced_module.set_watch_points(add_1_node)
out = traced_module(inp)
Will get the following ``watch_node_value``::
print(traced_module.watch_node_value)
...
...
@@ -2010,7 +2010,7 @@ class TracedModule(Module):
r
"""Initialize the :attr:`~.TracedModule.end_points`.
When all the ``nodes`` are generated, the Module will stop execution and return directly.
Args:
nodes: a list of ``Node``.
...
...
@@ -2035,7 +2035,7 @@ class TracedModule(Module):
traced_module.set_end_points(add_1_node)
out = traced_module(inp)
Will get the following ``out``::
print(out)
...
...
@@ -2354,7 +2354,7 @@ def wrap(func: Callable):
@tm.wrap
def my_func(x, y):
return x + y
Args:
func: the function of the global function to insert into the graph when it's called.
"""
...
...
imperative/python/megengine/utils/tensorboard.py
浏览文件 @
b22ceba7
...
...
@@ -141,21 +141,26 @@ class SummaryWriterExtend(SummaryWriter):
def
add_text
(
self
,
tag
,
text_string_list
,
global_step
=
None
,
walltime
=
None
):
"""Add text data to summary.
Args:
tag (string): Data identifier
text_string_list (string list): String to save
global_step (int): Global step value to record
walltime (float): Optional override default walltime (time.time())
seconds after epoch of event
Examples::
# text can be divided into three levels by tag and global_step
from writer import SummaryWriterExtend
writer = SummaryWriterExtend()
writer.add_text('level1.0/level2.0', ['text0'], 0)
writer.add_text('level1.0/level2.0', ['text1'], 1)
writer.add_text('level1.0/level2.1', ['text2'])
writer.add_text('level1.1', ['text3'])
Examples:
.. code-block:: python
# text can be divided into three levels by tag and global_step
from writer import SummaryWriterExtend
writer = SummaryWriterExtend()
writer.add_text('level1.0/level2.0', ['text0'], 0)
writer.add_text('level1.0/level2.0', ['text1'], 1)
writer.add_text('level1.0/level2.1', ['text2'])
writer.add_text('level1.1', ['text3'])
"""
self
.
_get_file_writer
().
add_summary
(
...
...
@@ -171,8 +176,11 @@ class SummaryWriterExtend(SummaryWriter):
dtype
=
None
,
attributes
=
{},
):
"""Add node raw datas that can help build graph.After add all nodes, call
add_graph_by_node_raw_list() to build graph and add graph data to summary.
"""Add node raw datas that can help build graph.
After add all nodes, call ``add_graph_by_node_raw_list()`` to build
graph and add graph data to summary.
Args:
name (string): opr name.
op (string): opr class name.
...
...
@@ -180,15 +188,19 @@ class SummaryWriterExtend(SummaryWriter):
outputshape (list): output shape.
dtype (string): output data dtype.
attributes (dict): attributes info.
Examples::
from writer import SummaryWriterExtend
writer = SummaryWriterExtend()
writer.add_node_raw('node1', 'opr1', outputshape=[6, 2, 3], dtype="float32", attributes={
"peak_size": "12MB", "mmory_alloc": "2MB, percent: 16.7%"})
writer.add_node_raw('node2', 'opr2', outputshape=[6, 2, 3], dtype="float32", input="node1", attributes={
"peak_size": "12MB", "mmory_alloc": "2MB, percent: 16.7%"})
writer.add_graph_by_node_raw_list()
Examples:
.. code-block:: python
from writer import SummaryWriterExtend
writer = SummaryWriterExtend()
writer.add_node_raw('node1', 'opr1', outputshape=[6, 2, 3], dtype="float32", attributes={
"peak_size": "12MB", "mmory_alloc": "2MB, percent: 16.7%"})
writer.add_node_raw('node2', 'opr2', outputshape=[6, 2, 3], dtype="float32", input="node1", attributes={
"peak_size": "12MB", "mmory_alloc": "2MB, percent: 16.7%"})
writer.add_graph_by_node_raw_list()
"""
# self.node_raw_list.append(
...
...
编辑
预览
Markdown
is supported
0%
请重试
或
添加新附件
.
添加附件
取消
You are about to add
0
people
to the discussion. Proceed with caution.
先完成此消息的编辑!
取消
想要评论请
注册
或
登录