1
2
3
4
5
6
7
8
9
10
11
12
13
14
15
16
17
18
19
20
21
22
23
24
25
26
27
28
29
30
31
32
33
34
35
36
37
38
39
40
41
42
43
44
45
46
47
48
49
50
51
52
53
54
55
56
57
58
59
60
61
62
63
64
65
66
67
68
69
70
71
72
73
74
75
76
77
78
79
80
81
82
83
84
85
86
87
88
89
90
91
92
93
94
95
96
97
98
99
100
101
102
103
104
105
106
107
108
109
110
111
112
113
114
115
116
117
118
119
120
121
122
123
124
125
126
127
128
129
130
131
132
133
134
135
136
137
138
139
140
141
142
143
144
145
146
147
148
149
150
151
152
153
154
155
156
157
158
159
160
161
|
from enum import Enum
from typing import List, Optional, Tuple, Union
from torch._C import device, dtype, layout
# defined in torch/csrc/profiler/python/init.cpp
class RecordScope(Enum):
FUNCTION = ...
BACKWARD_FUNCTION = ...
TORCHSCRIPT_FUNCTION = ...
KERNEL_FUNCTION_DTYPE = ...
CUSTOM_CLASS = ...
BUILD_FEATURE = ...
LITE_INTERPRETER = ...
USER_SCOPE = ...
STATIC_RUNTIME_OP = ...
STATIC_RUNTIME_MODEL = ...
class ProfilerState(Enum):
Disable = ...
CPU = ...
CUDA = ...
NVTX = ...
ITT = ...
KINETO = ...
KINETO_GPU_FALLBACK = ...
class ActiveProfilerType(Enum):
NONE = ...
LEGACY = ...
KINETO = ...
NVTX = ...
ITT = ...
class ProfilerActivity(Enum):
CPU = ...
CUDA = ...
class _EventType(Enum):
Allocation = ...
Backend = ...
PyCall = ...
PyCCall = ...
TorchOp = ...
Kineto = ...
class _ExperimentalConfig:
def __init__(
self,
profiler_metrics: List[str] = ...,
profiler_measure_per_kernel: bool = ...,
verbose: bool = ...,
) -> None: ...
...
class ProfilerConfig:
def __init__(
self,
state: ProfilerState,
report_input_shapes: bool,
profile_memory: bool,
with_stack: bool,
with_flops: bool,
with_modules: bool,
experimental_config: _ExperimentalConfig,
) -> None: ...
...
class _ProfilerEvent:
start_tid: int
start_time_ns: int
children: List[_ProfilerEvent]
extra_fields: Union[
_ExtraFields_TorchOp,
_ExtraFields_Backend,
_ExtraFields_Allocation,
_ExtraFields_OutOfMemory,
_ExtraFields_PyCall,
_ExtraFields_PyCCall,
_ExtraFields_Kineto,
]
@property
def name(self) -> str: ...
@property
def tag(self) -> _EventType: ...
@property
def id(self) -> int: ...
@property
def parent(self) -> Optional[_ProfilerEvent]: ...
@property
def correlation_id(self) -> int: ...
@property
def end_time_ns(self) -> int: ...
@property
def duration_time_ns(self) -> int: ...
class _Inputs:
shapes: List[List[int]]
dtypes: List[str]
strides: List[List[int]]
ivalues: List[Union[int, float, bool, complex]]
tensor_metadata: List[Optional[_TensorMetadata]]
class _TensorMetadata:
impl_ptr: Optional[int]
storage_data_ptr: Optional[int]
id: Optional[int]
@property
def layout(self) -> layout: ...
@property
def device(self) -> device: ...
@property
def dtype(self) -> dtype: ...
class _ExtraFields_TorchOp:
inputs: _Inputs
sequence_number: int
allow_tf32_cublas: bool
@property
def scope(self) -> RecordScope: ...
class _ExtraFields_Backend: ...
class _ExtraFields_Allocation:
ptr: int
id: Optional[int]
alloc_size: int
total_allocated: int
total_reserved: int
@property
def device(self) -> device: ...
class _ExtraFields_OutOfMemory: ...
class _PyFrameState:
line_number: int
function_name: str
@property
def file_name(self) -> str: ...
class _NNModuleInfo:
@property
def params(self) -> List[Tuple[str, int]]: ...
@property
def cls_name(self) -> str: ...
class _ExtraFields_PyCCall:
callsite: _PyFrameState
caller: _PyFrameState
module: Optional[_NNModuleInfo]
class _ExtraFields_PyCall:
caller: _PyFrameState
class _ExtraFields_Kineto: ...
|
