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
162
163
164
165
166
167
168
169
170
171
172
173
174
175
176
177
178
179
180
181
182
183
184
185
186
187
188
189
190
191
192
193
194
195
196
197
198
199
200
201
202
203
204
205
206
207
208
209
210
211
212
213
214
215
216
217
218
219
220
221
222
223
224
225
226
227
228
229
230
231
232
233
234
235
236
237
238
239
240
241
242
243
244
245
246
247
248
249
250
251
252
253
254
255
256
257
258
259
260
261
262
263
264
265
266
267
268
269
270
271
272
273
274
275
276
277
278
279
280
281
282
283
284
285
286
287
288
289
290
291
292
293
294
295
296
297
298
299
300
|
%global _empty_manifest_terminate_build 0
Name: python-pyrevolve
Version: 2.2.2
Release: 1
Summary: Python wrapper for Revolve checkpointing
License: MIT
URL: https://github.com/opesci/pyrevolve/
Source0: https://mirrors.nju.edu.cn/pypi/web/packages/72/4e/d0926921a4fedd0fdf786e4ce5450e5a1eeebb5a2225110fc72bf17dd0a1/pyrevolve-2.2.2.tar.gz
BuildArch: noarch
%description
# Checkpointing
The adjoint computation of an unsteady nonlinear primal function requires the
full primal trajectory in reverse temporal order. Storing this can exceed the
available memory. In that case, Checkpointing can be used to store the state
only at carefully selected points in time. From there, the forward computation
can be restarted to recompute lost sections of the trajectory when they are
needed during the adjoint computation. This is always a tradeoff between memory
and runtime. The classic and provably optimal way to do this for a known number
of time steps is [Revolve](https://doi.org/10.1145/347837.347846), and there are other algorithms for optimal online
checkpointing if the number of steps is unknown a priori, or for multistage
checkpointing if there are multiple layers of storage, e.g. memory and hard
drive.
# PyRevolve
The pyrevolve library contains two parts: crevolve, which is a thin Python
wrapper around a previously published [C++ implementation](https://www.mathematik.hu-berlin.de/de/forschung/forschungsgebiete/mathematische-optimierung/andrea-walther/walther-software), and pyrevolve
itself, which sits on top of crevolve and manages data and computation
management for the user.
The C++ files in this package are slightly modified to play more nicely with
Python, but the original is available from the link below. In addition, there
is a C wrapper around the C++ library, to simplify the interface with Python.
This C wrapper is taken from [libadjoint](https://bitbucket.org/dolfin-adjoint/libadjoint).
# Installation
Since PyRevolve is distributed as source, you must have a functional C/C++
compiler in your environment before trying to install PyRevolve. The
installation procedure respects the standard environment variables
like CC/CXX.
The simplest installation is through pip by simply doing:
pip install pyrevolve
If you would like to have a local development copy that you can edit,
clone the repo and call the following command in the home directory:
pip install -e .
If you face any installation issues, first check that you really do
have a functional C++ compiler, either available as `g++` in the path
or pointed to by the environment variable `CXX`. If this is not the
reason, please raise a issue here or on
[this](https://opesci-slackin.now.sh) Slack channel.
# Usage
There are two wrappers: a _classic_ wrapper that follows the behaviour of Revolve
as described in the papers, and leaves the data mangement, the actual copying
of data, and the calling of operators to the user. An example of how to use it
can be executed by calling
python examples/use_classic.py
The other, _modernised_ wrapper, takes care of all this. The user creates a
Revolver object, and passes a forward operator, reverse operator, and
checkpoint operator to it. The Revolver provides two important methods:
`apply_forward`, and `apply_reverse`. A call to `apply_forward` executes the
forward computation, while creating the necessary checkpoints for the reverse
computation. After this, a user may also call the `apply_reverse` method to
compute the adjoints.
For this to work, the user is responsible that the operators have an `apply()`
method that takes arguments `t_start` and `t_end`, and that the checkpoint
object has a property `size` to report the size of one checkpoint, and methods
`load(ptr)` and `save(ptr)` that deep-copy all time-dependent live data into
a location given in `ptr`.
An example of this can be found here:
python examples/use_modernised.py
%package -n python3-pyrevolve
Summary: Python wrapper for Revolve checkpointing
Provides: python-pyrevolve
BuildRequires: python3-devel
BuildRequires: python3-setuptools
BuildRequires: python3-pip
%description -n python3-pyrevolve
# Checkpointing
The adjoint computation of an unsteady nonlinear primal function requires the
full primal trajectory in reverse temporal order. Storing this can exceed the
available memory. In that case, Checkpointing can be used to store the state
only at carefully selected points in time. From there, the forward computation
can be restarted to recompute lost sections of the trajectory when they are
needed during the adjoint computation. This is always a tradeoff between memory
and runtime. The classic and provably optimal way to do this for a known number
of time steps is [Revolve](https://doi.org/10.1145/347837.347846), and there are other algorithms for optimal online
checkpointing if the number of steps is unknown a priori, or for multistage
checkpointing if there are multiple layers of storage, e.g. memory and hard
drive.
# PyRevolve
The pyrevolve library contains two parts: crevolve, which is a thin Python
wrapper around a previously published [C++ implementation](https://www.mathematik.hu-berlin.de/de/forschung/forschungsgebiete/mathematische-optimierung/andrea-walther/walther-software), and pyrevolve
itself, which sits on top of crevolve and manages data and computation
management for the user.
The C++ files in this package are slightly modified to play more nicely with
Python, but the original is available from the link below. In addition, there
is a C wrapper around the C++ library, to simplify the interface with Python.
This C wrapper is taken from [libadjoint](https://bitbucket.org/dolfin-adjoint/libadjoint).
# Installation
Since PyRevolve is distributed as source, you must have a functional C/C++
compiler in your environment before trying to install PyRevolve. The
installation procedure respects the standard environment variables
like CC/CXX.
The simplest installation is through pip by simply doing:
pip install pyrevolve
If you would like to have a local development copy that you can edit,
clone the repo and call the following command in the home directory:
pip install -e .
If you face any installation issues, first check that you really do
have a functional C++ compiler, either available as `g++` in the path
or pointed to by the environment variable `CXX`. If this is not the
reason, please raise a issue here or on
[this](https://opesci-slackin.now.sh) Slack channel.
# Usage
There are two wrappers: a _classic_ wrapper that follows the behaviour of Revolve
as described in the papers, and leaves the data mangement, the actual copying
of data, and the calling of operators to the user. An example of how to use it
can be executed by calling
python examples/use_classic.py
The other, _modernised_ wrapper, takes care of all this. The user creates a
Revolver object, and passes a forward operator, reverse operator, and
checkpoint operator to it. The Revolver provides two important methods:
`apply_forward`, and `apply_reverse`. A call to `apply_forward` executes the
forward computation, while creating the necessary checkpoints for the reverse
computation. After this, a user may also call the `apply_reverse` method to
compute the adjoints.
For this to work, the user is responsible that the operators have an `apply()`
method that takes arguments `t_start` and `t_end`, and that the checkpoint
object has a property `size` to report the size of one checkpoint, and methods
`load(ptr)` and `save(ptr)` that deep-copy all time-dependent live data into
a location given in `ptr`.
An example of this can be found here:
python examples/use_modernised.py
%package help
Summary: Development documents and examples for pyrevolve
Provides: python3-pyrevolve-doc
%description help
# Checkpointing
The adjoint computation of an unsteady nonlinear primal function requires the
full primal trajectory in reverse temporal order. Storing this can exceed the
available memory. In that case, Checkpointing can be used to store the state
only at carefully selected points in time. From there, the forward computation
can be restarted to recompute lost sections of the trajectory when they are
needed during the adjoint computation. This is always a tradeoff between memory
and runtime. The classic and provably optimal way to do this for a known number
of time steps is [Revolve](https://doi.org/10.1145/347837.347846), and there are other algorithms for optimal online
checkpointing if the number of steps is unknown a priori, or for multistage
checkpointing if there are multiple layers of storage, e.g. memory and hard
drive.
# PyRevolve
The pyrevolve library contains two parts: crevolve, which is a thin Python
wrapper around a previously published [C++ implementation](https://www.mathematik.hu-berlin.de/de/forschung/forschungsgebiete/mathematische-optimierung/andrea-walther/walther-software), and pyrevolve
itself, which sits on top of crevolve and manages data and computation
management for the user.
The C++ files in this package are slightly modified to play more nicely with
Python, but the original is available from the link below. In addition, there
is a C wrapper around the C++ library, to simplify the interface with Python.
This C wrapper is taken from [libadjoint](https://bitbucket.org/dolfin-adjoint/libadjoint).
# Installation
Since PyRevolve is distributed as source, you must have a functional C/C++
compiler in your environment before trying to install PyRevolve. The
installation procedure respects the standard environment variables
like CC/CXX.
The simplest installation is through pip by simply doing:
pip install pyrevolve
If you would like to have a local development copy that you can edit,
clone the repo and call the following command in the home directory:
pip install -e .
If you face any installation issues, first check that you really do
have a functional C++ compiler, either available as `g++` in the path
or pointed to by the environment variable `CXX`. If this is not the
reason, please raise a issue here or on
[this](https://opesci-slackin.now.sh) Slack channel.
# Usage
There are two wrappers: a _classic_ wrapper that follows the behaviour of Revolve
as described in the papers, and leaves the data mangement, the actual copying
of data, and the calling of operators to the user. An example of how to use it
can be executed by calling
python examples/use_classic.py
The other, _modernised_ wrapper, takes care of all this. The user creates a
Revolver object, and passes a forward operator, reverse operator, and
checkpoint operator to it. The Revolver provides two important methods:
`apply_forward`, and `apply_reverse`. A call to `apply_forward` executes the
forward computation, while creating the necessary checkpoints for the reverse
computation. After this, a user may also call the `apply_reverse` method to
compute the adjoints.
For this to work, the user is responsible that the operators have an `apply()`
method that takes arguments `t_start` and `t_end`, and that the checkpoint
object has a property `size` to report the size of one checkpoint, and methods
`load(ptr)` and `save(ptr)` that deep-copy all time-dependent live data into
a location given in `ptr`.
An example of this can be found here:
python examples/use_modernised.py
%prep
%autosetup -n pyrevolve-2.2.2
%build
%py3_build
%install
%py3_install
install -d -m755 %{buildroot}/%{_pkgdocdir}
if [ -d doc ]; then cp -arf doc %{buildroot}/%{_pkgdocdir}; fi
if [ -d docs ]; then cp -arf docs %{buildroot}/%{_pkgdocdir}; fi
if [ -d example ]; then cp -arf example %{buildroot}/%{_pkgdocdir}; fi
if [ -d examples ]; then cp -arf examples %{buildroot}/%{_pkgdocdir}; fi
pushd %{buildroot}
if [ -d usr/lib ]; then
find usr/lib -type f -printf "/%h/%f\n" >> filelist.lst
fi
if [ -d usr/lib64 ]; then
find usr/lib64 -type f -printf "/%h/%f\n" >> filelist.lst
fi
if [ -d usr/bin ]; then
find usr/bin -type f -printf "/%h/%f\n" >> filelist.lst
fi
if [ -d usr/sbin ]; then
find usr/sbin -type f -printf "/%h/%f\n" >> filelist.lst
fi
touch doclist.lst
if [ -d usr/share/man ]; then
find usr/share/man -type f -printf "/%h/%f.gz\n" >> doclist.lst
fi
popd
mv %{buildroot}/filelist.lst .
mv %{buildroot}/doclist.lst .
%files -n python3-pyrevolve -f filelist.lst
%dir %{python3_sitelib}/*
%files help -f doclist.lst
%{_docdir}/*
%changelog
* Wed May 10 2023 Python_Bot <Python_Bot@openeuler.org> - 2.2.2-1
- Package Spec generated
|
