forked from brinckmann/montepython_public
-
Notifications
You must be signed in to change notification settings - Fork 0
/
changelog.dat
292 lines (287 loc) · 19.2 KB
/
changelog.dat
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
v3.5.0 + Added framework for spectral distortions likelihoods, plus mock FIRAS and PIXIE (D. C. Hooper, M. Lucca, N. Schoeneberg)
+ Added new Pantheon likelihood with prior on the absolute magnitude (D. Camarena)
+ Updated Pantheon likelihood to latest value
+ Various bugfixes related to python2 / python3 compatibility (D. C. Hooper, N. Schoeneberg)
+ Various additional bugfixes
---------------------------------------------------------------------
v3.4.0 + Added BBN likelihoods from 1907.11594 (N. Schoeneberg, D. C. Hooper)
+ Added H0LICOW likelihoods from 1907.04869 (S. Taubenberger, S. Suyu)
+ New feature: experiment specifications passed in .param files now overwrite ones in .data files, which allows for testing different likelihood specifications simultaneously (N. Schoeneberg)
+ New plotting features: to_derive allows to add new parameters in the plot file without overwriting the old parameter, and to_reorder allows to choose the order of parameters for the plots (N. Schoeneberg)
+ Significant speed-up of Pantheon likelihood (J. Renk)
+ Improved treatment of tick limits, sigma bounds, and Omega_m / omega_m (N. Schoeneberg)
+ Various bugfixes related to python2 / python3 compatibility (D. C. Hooper, N. Schoeneberg, F. Agocs, J. Renk)
+Various additional bugfixes
---------------------------------------------------------------------
v3.3.0 + Added python 3 compatibility, python 2.7 still supported (N. Schöneberg, F. Köhlinger)
+ Added Lyman-alpha alpha-beta-gamma likelihood from 1907.01496 (M. Archidiacono, D.C. Hooper, R. Murgia)
+ Added new type of parameters, derived_lkl, that behave like derived parameters of the likelihood and are not passed to class (D.C. Hooper)
+ Added joint prior on sz nuisance parameters for all Planck 18 likelihoods. Bestfits and covmats updated to reflect this (D.C. Hooper)
+ Various bugfixes and minor improvements, e.g. updated example plot file with more options and clarifications
---------------------------------------------------------------------
v3.2.0 + Added Planck 2018 likelihoods and example param files (Deanna Hooper)
+ Added BOSS DR12 and eBOSS DR14 Lya and QSO BAO likelihoods from 1906.11628 (James Farr)
+ Added KiDS+VIKING-450 likelihood from 1812.06076 (KiDS collaboration)
+ Added KiDS-450 correlation function likelihood from 1809.01406 (Fabian Köhlinger)
+ Various bugfixes
---------------------------------------------------------------------
v3.1.0 + Added PolyChord sampler (Handley, Hobson, Lasenby 1502.01856)
thanks to Will Handley and Deanna C. Hooper (for usage, see
the description of option "-m PC" in "run --help", the
PolyChord help files, and montepython/PolyChord.py)
+ Updated likelihoods euclid_pk and euclid_lensing to version of 1801.08331
+ Added 21cm intensity mapping likelihoods ska1_IM_band1 and ska1_IM_band2 from 1801.08331
+ Added cosmic shear likelihoods ska1_lensing and ska2_lensing from 1801.08331
+ Added galaxy clustering likelihoods ska1_pk and ska2_pk from 1801.08331
+ Added CMB likelihoods cmb_s4 and pico from 1808.05955
+ Updated CMB likelihoods core_m5, litebird and fake_planck_realistic for 1808.05955
+ Added CMB-S4 + CMB satellite combination likelihood cmb_s4_highl from 1808.05955
+ Added litebird_lowl, litebird_highl_fsky0p3 from 1808.05955
+ Added core_m5_lowl, core_m5_highl_fsky0p3 from 1808.05955
+ Added fake_planck_realistic_lowl, fake_planck_realistic_highl_fsky0p17 from 1808.05955
+ Pantheon likelihood added courtesy of Rodrigo von Martens
+ wrapper_wmap updated with workaround for WMAP v5.0 likelihood by S. More
+ Bugfixes affecting convergence rate of superupdate
+ Reverted default behavior of analyze to pre v3.0 where analyze would
remove the part of the chain before the latest covmat update
+ added option "--keep-only-markovian" corresponding to v3.1 behavior,
which also removes the part of the chain where the jumping factor was
changed (see "info --help" for details)
---------------------------------------------------------------------
v3.0.1 + (detailed changelog missing) bugfixes
+ Some new likelihoods (see folder montepython/likelihoods/)
---------------------------------------------------------------------
v3.0.0 + (detailed changelog missing) main features summarized
+ superupdate metropolis-hastings sampling (see description
of option "--superupdate SUPERUPDATE" in "run --help").
Automatically adjust jumping factor to improve convergence rate
+ Fisher matrix computation (see description of option
"--fisher" in "run --help") useful for producing starting covmat
+ Many new likelihoods (see folder montepython/likelihoods/)
+ Improved plotting (see options with "info --help")
---------------------------------------------------------------------
v2.2.1 + to avoid I/O bug, only max(R-1) written in chains during update
---------------------------------------------------------------------
v2.2.0 + Automatic update of the proposal density's covariance matrix, with
or without MPI (for usage, see the description of option
"--update UPDATE" in "run --help").
+ When analyse run, keep only the markovian part of the chains,
unless you use the flag "--keep-non-markovian" in order to get
better (but slightly biased) statistics.
+ When analyse run, new covmat is not computed anymore by default
(to avoid interfering with possible on-going run in update mode).
New covmat calculation can be forced with flag --want-covmat
----------------------------------------------------------------------
v2.1.4 + Points to Planck 2015 likelihoods (default TT, TTTEEE, lite)
+ New example files base2015*.param for Planck 2015
----------------------------------------------------------------------
v2.1.3 + Added SDSS LRG4 data (sdss_lrgDR4)
+ Additional functionality to compute and use fisher matrix
+ Added a --minimal flag for info to only compute covmat/bestfit
+ Corrected the sigma8 likelihoods: minimum requirement k_max>=1h/Mpc
----------------------------------------------------------------------
v2.1.2 + Fixed the ReadTheDocs website.
----------------------------------------------------------------------
v2.1.1: + Implemented a new observable, cosmic clocks, that measure the Hubble
rate at different redshifts. Full information can be found in
data/cosmic_clocks/README.txt (credits Licia Verde)
+ Implemented the BAO measurements from WiggleZ, from 1401.0358v2. Use
the new likelihood, called WiggleZ_bao, in your parameter file.
+ Added a new BAO data point from Ross et. al. 2014, available in
data/bao_2014.txt along with all the previous ones. Notice that all
reference to the historical and unfortunate rs_rescale is now gone -
all the data points are stored as if computed with the true value of
the sound horizon, and not the Eisenstein and Hu 98 formula. Thanks
to Antonio Cuesta for his help in annihilating this issue.
+ Fixed the ReadTheDocs website.
----------------------------------------------------------------------
v2.1.0: + Implemented the JLA new SuperNovae analysis (arXiv:1401.4064). There
are two versions available, 'JLA' and 'JLA_simple' (the latter being a
faster version of the former, so they are mutually exclusive). Thanks
to Martin Kunz, Stefano Foffa, Yves Dirian and Valeria Pettorino for
their testing and help in debugging. The data being massive, you are
asked to download it manually.
+ Euclid fiducial likelihoods have been cleaned, and adapted to the
latest version of the wrapper
+ Implemented the POLARBEAR likelihood (measurement of the BB spectrum
in four bins at large ell, see arXiv:1403.2369).
+ Major rewriting of the 'analyze' module. Now handles arbitrary number
of folders/set of files. The '--comp' flag has been removed - you
simply specify more elements to the 'info' call.
+ You can now combine parameters in 'analyze' (at your own risk, beware
of the priors), using a new dictionary in the 'extra plot file'. See
the example in 'plot_files/example.plot'
+ Unified scheme for handling complex parameters from CLASS. You can
now specify for instance 'm_ncdm__1', 'm_ncdm__2' as running
parameters, and they will be treated correctly.
+ It is now possible to add derived parameters that were not chosen
before the run began. Simply list the parameters to follow on an
existing folder with the flag '-m Der', and the options
'--Der-target-folder' and '--Der-param-list'.
+ 'mpi_run' method reinforced - should now work on more possible
configurations.
+ Rewriting of the '--help' method, with long information printed if
'--help' is called, and short if '-h' is used instead. Credits to
Francesco Montesano.
+ added a '--silent' option to avoid printing on-screen the steps in
parameter space.
+ Fixed bao_2013.txt data, it now uses the right sound horizon, thanks
to Martin Kunz, Stefano Foffa and Yves Dirian for finding it.
+ Monte Python is now under the MIT License (permissive BSD-type
license)
-=-=-=-=-=-=-=-=-=-=-=-=-=-=-=-=-=-=-=-=-=-=-=-=-=-=-=-=-=-=-=-=-=-=-=
v2.0.5: + Nested Sampling can now be used with MPI - see the documentation.
+ The histograms are now also stored in files if the '--all' flag is
specified.
+ Improved the communication with the classy wrapper, notably on the
derived parameters - now the function `get_current_derived_parameters`
does not need to know how Monte Python works, which is better.
----------------------------------------------------------------------
v2.0.4: + Importance Sampling implemented. You can ask for this by specifying
the method flag to IS, `-m IS`. It also requires a starting folder,
`--IS-starting-folder something/` in order to work. The parameter file
must be identical to the one in the starting folder, excepted at the
level of the likelihoods. The method then reads the existing chains,
and compute the new likelihood and multiplicity, and stores this in a
new folder.
+ Bug-fix for the previously cleaned parser, notably in the analyze
module.
----------------------------------------------------------------------
v2.0.3: + New BOSS BAO likelihoods - courtesy of the BOSS collaboration via Jan
Grieb. There are now two possibilities, isotropic and anisotropic
measurements. See
https://github.com/baudren/montepython_public/pull/9 for more details.
+ There are now options to also plot the comparison between two
experiments (folders), not only on the 1d posterior distributions, but
also the 2d posterior distributions (credits to Jan Grieb)
+ Separated the different way of calling MontePython (for running and
analyzing), to improve clarity of the help messages when asking
`MontePython.py -h` (credits Francesco Montesano). Note that
compatibility is maintained with the old way of calling things! Note
also that to obtain the help on the running, or information
subcommands, you should call: `MontePython.py run -h` or
`MontePython.py info -h`.
+ Tentative implementation of basic MPI support. With mpi, and mpi4py
(the Python wrapper of MPI), you can now an arbitrary amount of chains
on a cluster and prevent (normally) any name conflict.
----------------------------------------------------------------------
v2.0.2: + BICEP2 likelihood implemented. You can specify which field you want to
use ('B', 'E', or 'T' so far according to their documentation), with
the line: bicep2.fields = ['B', 'E']
See the bicep2.param for reference.
+ You can now specify likelihood parameters directly in the parameter
file. This overwrites any setting present in the .data file of the
likelihood. Simply write: `mylkl.something = 'that'`.
+ Help messages have been improved with the help of Francesco Montesano.
More improvements will come in the next version.
+ Improved error messages, reading of the conf file now supports the
tilde character.
----------------------------------------------------------------------
v2.0.1: + Bug-fix from 2.0.0: the memory was not properly freed from CLASS
calls. There was a problem with imports of the errors imported from
the cosmological module.
+ Lots of optimization, improvements and fixing of the analyze module.
It now runs much faster for heavy runs with hundreds of chains to
analyze.
+ Fixed the routine in nested_sampling: `from_NS_output_to_chains()`,
which was bugged. Now one can analyze properly the output from
MultiNest, translate it to Monte Python format, and compute a
covariance matrix.
----------------------------------------------------------------------
v2.0.0: + Over sampling implemented: it helps running with likelihoods with a
high number of varying nuisance parameters.
+ **Repackaging**: the whole folder now should behave more like a proper
python package. This included renaming the `code` directory into
`montepython`, and moving the `likelihoods` folder inside this new
`montepython`. Also, considering a `lkl` likelihood, instead of
following the previous structure:
`likelihoods/lkl/lkl.py and .data,
you should now use:
`montepython/likelihoods/lkl/__init__.py and lkl.data`.
+ Wrapping of the Cosmo Hammer within Monte Python. This should allow
runs being able to scale on the biggest available clusters (big thanks
to Joel Akeret and Sebastian Seehars for the help in implementation).
+ Wrapping of MultiNest within Monte Python, thanks to the already
existing Python wrapper written by Johannes Buchner (big thanks to
Jesus Torrado for coding most of the communication between MP and
PyNS).
+ Experimental feature: automated separating in subfolders a run that
was found by MultiNest to be multimodal (Jesus Torrado)
+ New WiggleZ analysis (Signe Riemer, David Parkinson)
+ First version of the test module, with basic functionalities being
checked.
+ Convention of path changed to comply with Python standards
=-=-=-=-=-=-=-=-=-=-=-=-=-=-=-=-=oOo=-=-=-=-=-=-=-=-=-=-=-=-=-=-=-=-=-
-=-=-=-=-=-=-=-=-=-=-=-=-=-=-=-=-oOo-=-=-=-=-=-=-=-=-=-=-=-=-=-=-=-=-=
v1.2.5: + Compatible with CLASS v2.0.x
----------------------------------------------------------------------
v1.2.4: + Fixed behaviour for the -all option when analyzing folders.
----------------------------------------------------------------------
v1.2.3: + Tentative error message to catch a typical error appearing when
running with a badly initialized log.param file. It now states that
you should probably remove the folder entirely. It is not sure that
this catches all types of situation, but it does catch some of them.
----------------------------------------------------------------------
v1.2.2: + Several bug fixes. One with analyze module, that failed when a file
was not complete (wrong number of columns) - fixed. The two other
were more serious, and detailed below:
+ When using as an input a covariance matrix containing less parameters
than the current run, wrong numbers were used for the remaining
proposal distribution - fixed
+ When running with *no* varying parameters, instead of using the values
specified in the parameter file, the code would instead run with
default Class values - fixed
----------------------------------------------------------------------
v1.2.1: + Planck lensing is now working
----------------------------------------------------------------------
v1.2.0: + Added the Planck likelihoods (called Planck_highl, Planck_lowl,
Planck_lensing and lowlike)
+ Added a Cholesky decomposition method to speed up sampling for
likelihoods with a large number of nuisance parameters based on
A.Lewis 1304.4473, which groups parameters by blocks of same speed.
+ Transfered the documentation to a sphinx-documentation. Most comments
in the code have been replaced by docstrings, automatically extracted
to create a web-site. You can now navigate the different functions,
see the dependencies, see the source code in one convenient place.
+ Renamed both io.py and parser.py into io_mp.py and parser_mp.py (mp
for Monte Python). It was conflicting with some installed packages of
python, both being some very often used names.
+ Removed the comparison between input param and log.param: always
choose the latest.
+ Changed the architecture of analyze.py file. The whole thing is no
longer a class, but instead a collection of functions. To exchange
information between these functions, a new class called `information`
has been created.
+ Removed the common.py file, merging its content (lockfile system) with
io_mp.py. Note that the lockfile is still being tested (works on our
machine)
+ Transfered the log_flag from being an indendent quantity being
exchanged between data and likelihoods to a proper attribute of the
data class. Since no comparison is done any-more, there was no reason
to keep it this way anymore.
-=-=-=-=-=-=-=-=-=-=-=-=-=-=-=-=-=-=-=-=-=-=-=-=-=-=-=-=-=-=-=-=-=-=-=
v1.1.0: + Added the following likelihoods: SPT full data (called spt_2500 in the
likelihood directory), compilation of BAO scale data "bao",
compilation of quasar time-delay data "timedelay" and updated the
value for hst (1103.2976)
+ To speed up the production of plots, you can play with a new extension
defining the print-out format: use the command-line option "-ext png",
or eps, or pdf (from the fastest to the slowest format)
+ Added a lockfile system, that should prevent a file from being written
simultaneously by two programs, when launching several different runs
at the same time.
+ Changed a detail in the input file syntax, concerning parameter prior
edges. To leave the prior edge undefined, you can still use "-1" as
before, or "None". But if your prior edge is really minus one (for
instance for w), you can write "-1."
+ Function cosmo_update_arguments() modified for better readability and
simpler use (this function allows to rename or remap input parameters
at each step before passing them to CLASS). The syntax pattern "try,
except" is not needed anymore
-=-=-=-=-=-=-=-=-=-=-=-=-=-=-=-=-=-=-=-=-=-=-=-=-=-=-=-=-=-=-=-=-=-=-=
v1.0.1: + Removed the first and unwanted line of code/mcmc.py, that was
pointlessly trying to import again Class. This conflicted in case of a
non-installation of the Class wrapper.
+ Modified '0.9' in '1.0' in a print statement in MontePython.py
----------------------------------------------------------------------
v1.0.0: + Initial release