- development tree is rift_O4c
- pseudo_pipe backwards compatibility fix for coinc file reading w/o eccentricity. PUFF settings for BNS use both LambdaTilde and DeltaLambdaTilde to avoid loss of points. gwsignal phase shift hlm modes (rotation) phi_star=pi/2 to match convention. CIP/ILE plugin acess without portfolios. pp_RIFT/pp_RIFT_with_ini cleanup. dag_utils periodic_release for ILE file transfer issues. PUFF lambda1,lambda2 in log coords. helper matter uses lambda1,lambda2. --internal-precompute-ignore-threshold (default None) to future-protect analysis of high-amplitude signals. asimov-> pesummary handoff of calmarg output. (rc0)
- pp_RIFT_with_ini improve OSG use, sanity plots, various bugfixes ( mass range, dmax/dmin); GWSignalWriteFrame fixes for inj; PUFF reflection fix; various demos (populations) and pp plotting scripts and waveform/CI plot script; pseudo_pipe cip-expl-de-jobs-auto last explode sanity for n_eff (rc1)
- bootstrap asimov using --manual-initial-grid-supplements (new option); subdag memory/disk limits correctly passed; wf interface update INRPhenomXAS; mcsamplerNFlow updates; ILE fix precompute-ignore-threshold so non-default value possible. calmarg/rift_source interfaces to eccentric waveforms (may need external edits), calmarg argument quoting. cosmo_sourceframe works with distance-marginalization. hyperpipe minor. puff file transfer fix. CIP adds rf_pca, rbf fit options. asimov defaults AV/AV no transverse coord as preferred, asimov scitokens; updates for pp test infrastructure (AV allows pinned sky, test/pp localizer adjust to be AV-relevant); factored_likelihood/ILE minimize gwsignal import to reduce overhead (pyseobnr expensive!); pp_RIFT_with_ini updates/bugfixes from Jake. ILE_batchmode dmarg NINF-> -inf protect numpy 2.0. rift.ini 'n input samples' option to control grid sizes. (rc2)
- pseudo_pipe quoted waveform arg passing to calibration_reweighting fixes; util_RandomizeOverlapOrder fix in case of many workers/few points per worker; remove X509 access; fix e^2 prior normalization (rc3)
- minor review-requested edits (mostly unused code), but note factored_likelihood precompute was doing redundant calculation (since 0.0.17.0). asimov_test_convergence.py. helper corner case if gid input coinc crazy in mc; undo RandomizeOverlapOrder --n-min test since needs to be passed consistently, defer until future (rc4)
- cepp_basic change pattern for CIP worker join_cip to only matches to wroker jobs + util_RandomizeOverlapOder.py drop CIP worker output with catastrophically too few points, so the use of min(n_samples) from each worker doesn't cause dramatic loss of precision - fix to change made in rc3 (rc5)
- asimov/rift.ini access to ILE 'n copies', as backstop for highest SNR (rc6)
- update requirements.txt so ligolw_print installed, to allow CI to work - with release, commit c723ae29c
Release is after last commit
- development tree is rift_O4c
- lalsimutils.periodic_params for plot_posterior_corner and puff; pipeline transverse puff better z coordinates noteworthy, ligo.lw->igwn_ligoow update, backend changes to CIP column order parsing to be more flexible, CIP lambda_min; ILE --internal-waveform-extra-kwargs for gwsignal, changing t_window; pipeline --internal-use-oauthfiles; hyperpipe convergence test nodes. Mean anomaly merge. Disable old ilwdchar-compatible reading to reduce warning messages. misc/dag_utils/write_ILE... now quotes, enabling passing arguments. asimov cleanup manual-extra-ile-args and allow waveform/arguments. PUFF bugfix (only applied last). rift_container.def with pyseobnr. asimov with ecc, it0 eccentric arch, AV does cupy-based random numbers, fixing conflict with cosmo_sourceframe. dag_utils and related handle osdf container identification better, also cip-request-disk. mcsamplerGPU np.isfinite fix. ILE less process_params output (rc0)
- FYI coinc.xml files from MBTA/spiir malformed and produce parse errors. calmarg: fix misnamed arguments in calibration_reweighting causing block, fix handoff from asimov to find bilby ini files; also more RETRY. pp_RIFT_with_ini use provided fmin/fmax/seglen correctly. helper initial grids respect force-mc-range. asimov boostrap from previous samples, including coinc; note samples_utils gets pesummary/h5 snarfing tool. eccentricity (lalsimutils/pipeline/ILE/CEPP): multiple updates to enable TEOBResumSDali use (pull #104), also --internal-cip-use-periodic-ecc-vars and related support, and minor bugfixes associated with meanPerAno merges (rc1)
- PUFF issue resolved (lalsimutils convert_waveform_coordinates corner case for transverse puff+downselect), plus provide for eta=0,1/4 reflections to preserve points; asimov/rift.py logger, bootstraping; dag_utils argument quoting to pass waveform dictionaries to ILE; pseudo_pipe/calibration_reweighting take fref, and also make files unique. bugfix blocking change with row.alpha for extrinsic export from recent eccentricity merge above. calibration_reweighting minor typo fix arg name (blocking error). (rc2)
- asimov/rift.py edit to prevent getting stuck identifying job needing rescue, also more useful error message when upstream dependency fail (calmarg needs bilby ini file). (rc3)
release is rc3
- development tree is rift_O4c
- true cosmological distance priors for ILE. Fix --cip-explode-jobs-subdag in cepp_basic to only operate when requested. Update requirements. Note lalsuite==7.23,7.24 releases don't work (and have never worked) due to broken executable installs (lal_path2cache): never use those two versions. (rc0)
- natsort in requirements, EOSManager tabular can extract one EOS realization, puff adds --fail-if-empty option, GWSignal/gwsignal interface forcing hlm tapering, plot_posterior_corner access external code to plot other (e.g., exact) marginals, hyperpipe randomize grid*.dat when joining to fix sort issue, hyperpipe SINGULARITY_BASE_EXE_DIR_HYPERPIPE, ILE/dmarg tables new cosmo/cosmo_sourceframe priors, pp_RIFT_with_ini boundary range fix ported, hyperpipe fix file transfer issues for PUFF outputs (rc1)
release is rc1
- development tree is rift_O4c
start roughly 0.0.16.0rc2. asimov integation, documentation, CIP --n-events-to-analyze; mcsamplerNFlow; mcsamplerPortfolio; unreliable_oracles including reference samples; implement mcsamplerPortfolio and some oracles in ILE and CIP; hyperpipe dag writer integer events; pp_RIFT_with_ini merge; hyperpipe EOS_POST other integrators; helper parses d-min from ini file; hyperpipe change EOS_POST default to include all data; more asimov (n-output-samples, request_disk, additional-files-to-transfer, distance prior, fmax, dmin; gwdata interface), AV integrator pinned params, inv-spec-trunc-time access at pipeline, source_redshift in convert_waveform_coordinates vectorized; statutils xpy.max fix for new cupy/python; helper_transfer_files access for eg surrogate files; hlmoft tapering for some cases (ChooseTDModes, SimInspiralTDModesFromPolarizations) not previously start-tapered;
Public OSG operation setup, focusing on hyperpipe (condor-local-nonworker-igwn-prefix, etc). Fix barf where pseudo_pipe required coinc with row.alpha4 (eccentricity).
bugfix sign error reflection for some modes (TEOBResumS and SEOBv4HM - not precessing); yet more asimov (calmarg, improved rift.py/rift.ini handoff of args, resurrect/completion detection). ILE hard fail if --use-gwsignal but no gwsignal. Tabular EOS fixes. Calmarg fix so weight_files present, and use alt_reweight to avoid prior problems. Expose tukey window length to allow top-level user to change it. important factored_likelihood change to ComputeModeCrossTermIP to speed up by x2. (rc0)
- TEOBResumS external interface bugfixes (m>=0 modes only by accident if aligned; pass phiref too; bugfix data.data missing); asimov interface updates in rift.ini/rift.py (approx; cache file collection; q-max; roll off time; rift/environment backend; RIFT_BOOLEAN_LIST; 'extra eccentric arguments'); factored_likelihood/internal_hlm_generator now requests tapering; backend issues (e.g., try/except around GetApproximantFromString; lalsimutils/puffball support vectorized; mcsamplerPortfolio uses entrypoints; minor scoping issues; code updates to match scipy changes to mvnun/integrate.simps; HyperCombine handles empty data; add --internal-XXX-request-memory; r strings to avoid warnings; don't call estimateWaveformDuration if not needed to avoid barf if fmin==0; float128/float64; disable default printing of long comment/logging messages; str conversion chIeff range fix; upstream change to event_id in coinc.xml). Transverse puffball. cip-explode-jobs-auto-scale. Updates to using NR simulations from KW. Oracle hill climber improvements. Bugfix hoft : psi not passed correctly (only for injections), now have waveform matching; add test/check_waveform_random . hlmoft ChooseFDModes: add conditioning to return from ChooseFDModes, work on phase factors/geometry so now excellent matching, including fd_centering_factor. calmarg code clean up (rc1)
- public OSG settings (--use-osg-public); basic CIP/ILE subdag system with while loop as option (cip-explode-jobs-subdag) ; minor bugfixes (ile-runtime-max-minutes to convergence subdag; remove simple_unique ILE_puff/ILE_fetch argument); minor refactoring of ILE nodes in dag writing (internal) in prep for improved subdag system (rc2)
release is rc2
- Since last release
- Start roughly 0.0.15.9 in rift_O4b, merge with 0.0.15.10 into it. ROMWaveformManager hlmoft backwards compatbile snarf extra options; pp_RIFT_with_ini prototype; pp_RIFT minor fixes (lalapps_path2cache); AV eos-tabular-infereence updates; hyperpipe; scitokens; fix online PE use case for O4 (PSD; gracedb-id logic if ini provided); request_memory units; EOSPlotUtilities; mcsamplerAdaptiveVolume (rc1)
- merge in 0.0.15.11, as well as below ( ILE_batchmode hard fail on JIT compilation error, from 15.12 below); AV in rosenbrock test; AV n_chunk size for portfolio; hyperpipe
- integer events; bugfix (-1)^m->(-1)^l reflection for aligned TEOBResumS external call (rc2)
- asimov integration (as in rift_O4a/0.0.15.12 rc0); documentation update for hyperpipe, etc; CIP --n-events-to-analyze for hyperpipe; expanded asimov integration (from rift_O4c branch) - calmarg, improved rift.py/rift.ini handoff of args
No release intended (modulo LVK requirements), folded into 0.0.17
- Since last release
- gracedb get file psd.xml.gz fix for online; ILE hard fail if CUDA/JIT compilation error; bugfix (-1)^m->(-1)^l reflection for aligned TEOBResumS external call; asimov integration
No release intended (modulo LVK requirements), folded into above.
- Since last release
- bugfix util_JoinExtrXML to catch last batch; bugfix --calibration-reweighting-initial-extra-args argument passing (rc1)
- bugfix Lmag high-order PN coefficient; ourparams glue.ligolw -> ligo.lw (rc2)
- CI fix so integration test sane (rc3)
Release is rc3
- Since last release
- ChooseFDWaveform J frame and fourier-transform-conditioning fixes as described in T2300304; add RIFT_BOOLEAN_LIST environment variable (rc1)
- getenv=True workaround; CIP spin prior normalization fixes (just needed for evidence); gwsignal implement Lmax in hlmoft (rc2)
- RIFT_GETENV_OSG; enable gwsignal aways even if RIFT_LOWLATENCY active; mcsamplerGPU typo fix (self.n_total) for corner use case (rc3)
- add --allow-subsolar to prevent hardcoding 1Msun limit; dockerfile cleanup; has_GWS scoping fix to avoid crash in factored_likelihood; remove glue.ligo_lw reference in util_SimInspiralToCoinc (rc4)
- bugfixes calmarg, mainly for use-gwsignal which wasn't implemented (rc5)
- minor fix to extrinsic export scripts to enable arbitrary output sample size (rc6)
Release is rc6
- Since last release
- compatibility minor updates for numpy>=1.24 (see #27); calmarg import for conda build fix; plot_posterior_corner psi mod pi plot option; query_singularity_path executable; CI adds test-build.sh (rc1)
- minor corner-case bugfixes re gpu/cpu typing; ILE_batchmode correctly resets when using GMM each iteration for dL/incl, and GMM+force-adapt-all implemented (not silently ignored); misc updates for contemporary online operation (ecp-cert-info; psd inside coinc.xml) dump reproducibility info and ini by default; prior_utils better checking cupy active; still more compatibility updates for numpy >=1.24, including follow-on changes to mcsamplerGPU n_eff test; placate XML backend change; --assume-matter-eos correctly set CIP tide option; ILE-specific GMM setup cleaned up (rc2)
- --assume-matter-conservatively (allow crazy tides), --rom-group (gwsurrogate) implies initial tapering,
- CIP --assume-eos-but-primary-bh, convergence_test_samples JS base 2 not e, vectorized_general_tools histogram try to avoid memory errors (rc3)
- EOSManager+reprimand minor review updates; mcsamplerGPU works on GPUs when fixing parameters; CI updates; fix --assume-matter-eos / --assume-eos-but-primary-bh again; CIP lambda export with --use-eos fix; helper fixes for --assume-matter-eos; lalsimutils CreateCompatibleComplexOverlap update for contemporary python; lmax_nyquist for gwsignal (ILE: --use-gwsignal-lmax-nyquist); helper sets both eta limits on initial grid with --force-eta-range; extrinsic export with eccentricity bugfix (rc4)
- misc hyperpipe/hyperpuff/CEP fixes (filenames/interface issues); plot_posterior_corner allow for composite with labelled fields; lalsimutils convert_vector_coordinates prevent fallthrough to non-vectorized; CIP 9-parameter fit variable typo; pipeline transfer gp pickle if on OSG; various int casts for modern / in python (rc5)
- plot_posterior_corner can use composite files with labelled fields; hypercombine product outcome, length consistency; EOSManager protect lambda_from_m for BHs; hyperpipe handoffs; CEPP set n_eff ofor last iteration tied to cip-explode-jobs-last (rc6)
release is rc6
- Since last release
- bugfix pseudo_pipe so --internal-*-use-lnL passed correctly to helper. CIP_gauss and CQL working correctly (rc1)
- bugfixes mcsamplerGPU (wrong var name mcsamplerGPU in type check; self.ntotal init at start of loop). mcsamplerGPU/statutils protect against cupyx.scipy.special not being present (rc2)
- documentation; mcsampler GPU/ILE_batchmode exports for use-lnL; dockerfile builds; pipeline --cip-explode-jobs-auto to auto-select appropriate CIP worker count; CIP --lnL-downscale-factor to help sample loud signals; pipeline --use-downscale-early to auto-select that factor; pipeline can use CIP_gauss in iterations, and can request via --use-gauss-early,
merge last TEOBResumS; collections.abc.Iterable for py3.10 support; helper minor misc (rc3)
- user control of n-iterations-subdag-max, and puff in all subdag iterations; CIP/fail-unless/n-eff all floating point; plot_posterior_corner.py can use matplotlibrc; pipeline correctly reduces goal of labor per worker in many-worker limit; helper edit (tanmay) to help using coinc as input; pipeline internal-cip-tripwire and --internal-n-evaluations-per-iteration options; fix extrinsic output for binaries with tides; minor misc bugfixes to obscure code paths; bugfix sky rotation and phase rotation; add GWSignal interface; tweak zero-spin run settings; ILE add 'supplementary-likelihood-factor' interface to enable call to external runtime-specified code; increase worker count for high SNR jobs with cip-explode-jobs-auto; ditto more for matter jobs; bugfix tidal export to XML; EOSManager updates. Note incompatible with lalsuite 7.13 still (rc4)
- fix CI; dmarg+phasemarg patch from soichiro (nonprecessing only); add --manual-extra-puff-args; add forward-looking approx names; add --force-adapt-all to ILE; add non-time-marginalized likelihood output if user is resampling in time (i.e., an snr-like output). EOSManager QueryLS, repirmand and causal-spectral; xml patch for lalsuite >=7.13; cosmo prior on gpu debug; dmarg allows pseudo_cosmo prior; misc osg minor updates; ILE --zero-likelihood for testing; various --manual-extra-X-args; cal marg from Jake (rc5)
- fix CI again; hyperpipe/hyperpost, framed for EOS; cal marg debug; Atul EOSManager updates (reprimand, etc); add missing fairdraw code to GMM and AC+lnL mcsamplerAC minor normalization cleaning for low-precision GPU arithmetic (right-edge CDF effect); row.time_geocent method in lalsimutils (rc6)
- catch various ILE errors; correctly set n_eff goal for CIP workers for last iteration consistent with cip-explode-jobs-last; add (inaccessible) option to manage XPHM version change; bugfix mcsamplerGMM error estimate; bugfix mcsamplerGMM in case of use-lnL/use-lnI; NSBH puffball fix lambda_BH=0; calmarg can use --calibration-reweighting-batchsize; remove print in hlmoft; calmarg more fixes paths; calibration_reweighting fix missing data for BBHs; add RIFT_AVOID_HOSTS variable so user can identify hosts for ILE to avoid (rc7)
- XPHM J frame workaround draft; cal marg workflow minor fixes so runs; some cupy/cuda workflow error handling, including --force-reset-all from top level (rc8)
- bugfix argparsing/typo in pipeline from last commits in rc8; add hlmoft conditioning for ChooseFDModes (rc9)
- cal marg pass --calibration-reweighting-count, alternate recombination methods, access (low-level-only) to alternate h_method, and minor bugfixes from typos (rc10)
- pipeline-level access to XPHM L-frame and condition control; XPHM workaround for L frame; ILE --force-reset-all typo fix (rc11)
- calmarg J--> L frame access (rc12)
release is rc12
- Since last release
- bugfix lalsimutils vectorized coordinate transform (sph coordinate cos_theta_2 use), add test in 'tests' for conversion; fix CIP issue where mc prior range could be changed by input grid despite --mc-range; implement geocenter-time posterior option at last stage, along with edits needed to implement (e.g., fairdraw output option for extrinsic stage); bugfix xml event time export; chi_pavg implement updates; EOSManager minor edits; GMM in ILE finer control over adaptation variables; overflow protection GMM+ILE now user-accessible; address typing issues in mcsampler (returning object type) for selected arguments; bugfix syntax errors introduced in some mergers; CIP/EOSManager methods for quick inference with tabulated sequence of EOSs (EOSSequenceLandry; etc); mcsampler avoid infinite loop for 'no contribution to integral,skipping'; CIP_gauss defined (gaussian fit + resampling based on gaussian); scitokens-ready ILE (rc1)
- dockerfile prototype in this repo; CQL vectorized; OSG updates (local.cache duplication); fix cupy memory warning; add CUDA memory limit to avoid landing on overstrained GPU hosts; convergence_test_samples has JS test used elsewhere; bugfix mcsamplerGPU adaptive (intermittent array size error); mcsampler default/gpu standardize n_history; integrator test update; lalapps_path2cache->lal_path2cache change; TEOBResumS integration as external package; new pseudo-cylindrical coordinates; new CIP option to put change-of-coordinates prior reweighting inside adaptive integrand, so it is done live instread of at-end reweighting; fix some fallthroughs in lalsimutils to 'slow' non-vectorized code; update vectorized tranform test to cover standard use cases and put into CI; pp_RIFT updates; start sphinx documentation (rc2)
- CIP running on OSG as option (--use-cip-osg); lalsimutils.convert_waveform_coordinates fix non-production transform and update tests; minor bugfixes (formatting ligolw_add arguments; dmarg+sky rotation wasn't coded). Change default fit to rf. cosmo prior gpu-ized and exposed for use. Rosenbrock test cleanup for paper. More sphinx documentation. --auto-logarithm-offset in ILE (and access via --internal-ile-auto-logarithm-offset). Tweaks to better automate interpretation of asymmetric binaries like BHNS (pipeline can set lambda1,lambda2 prior upper bounds, --force-chi-small-max, and allow tides only on one object). CIP_gaussian updates. (rc3)
- helper missing argument for --internal-ile-auto-logarithm-offset; fix --auto-logarithm-offset implementation SNR scale (and debugs thereof); ILE request_disk; expose --internal-rotate-phase; add --internal-loud-signal-mitigation-suite; add util_ForOSG_MakeTruncatedLocalFramesDir.sh and assocated .py script to trim frames for a remote-machine run; pp plot minor typing issues; fix accidental mangling of rosenbrock test commit; docs. (rc4)
- mcsamplerGPU use-lnL mode, via statutils; pipeline --cip-sigma-cut, --scale-mc-range, --internal-ile/cip-use-lnL; intermediate posterior*dat files have tides & eccentricity auto-produced (rc5)
- var name bugfix in helper; uniform lambda prior in iteration 0 option added (rc6)
Release is rc6, to facilitate early igwn-testing/igwn use.
- Since last release
- pseudo and helper (--use-legacy-gracedb at top level); lalsimutils overlaps using psi4 input; pseudo (gwsurrogate logic/reference location fixes) (rc1)
- pseudo_pipe (path fixes for osg; add --condor-nogrid-nonworker) (rc2)
- helper (typo in V1 data lookup/hard fail; fix corner case for burst trigger hard fail; don't override --force-eta-range at low mass); CEPP/dag_utils (--force-gpu-only, some OSG edits to clean requirements; expand --condor-nogrid-nonworker to apply to CIP), pseudo_pipe (add --force-hint-snr) (rc3)
- typo fix pseudo_pipe missing colon; convert_ile2inference convention change update py3 print; dag_utils fix PUFF issue no_grid-> PUFF fails; merge Yelikar edits to pp_RIFT for OSG, NRSur; request_disk option for ILE in CEPP/dag_utils for OSG runs; dag_utils add periodic_release etc update for OSG operation (rc4)
- carriage return bugfixes in convert_output_format ile2inference (rc5)
- grids from analytic fisher matrix (util_AnalyticFisherGrid) using gwbench; MOG gets latin hypercube sampling; new pseudo_pipe_lowlatency for low latency tunings (rc6)
- ILE, ILE_batchmode (option export extrinsic per iteration; hope to fix cosmo prior sampling; Dan's suggestion to remove np.vectorize calls), mcsampler (help export extrinsic per iteration), general analytic fisher matrix via gwbench calls (pseudo_pipe option), converter add missing py2->py3 print statements (rc7)
- remove ^M from CIPs; add eccentricity capability from Champion; add Henshaw/Gerosa chip_avg; CIP can import gaussian likelihood fits; ILE extrinsic export bugfix for likelihood export (missing --manual-logarithm-offset undo); CIP can use cos_theta1 and cos_theta2 as coordinates for sampling, and these are made default (rc8)
- fix merge error in CIP in above - travis fail! (rc9)
- fix another merge problem from CIP above which dropped the gaussians; add ILE hard fail on cuda errors (rc10) Probably should have been major release around July 19, 2021
- GMM updates and bugfixes; 'fetch' mode to grab info from related jobs; chip_av; GP fits informed by lnL errors; alternate fits for placement (cov, quadratic), glue->ligo.lw, assorted minor edits
UWM hackathon outcomes (distance marginalization (Wysocki/Morisaki); AMR grids) (rc11)
- soichiro mu1,mu2 coordinates; subdag iterate to convergence ('Z'); lalsimutils convert_waveform_coordinates vectorized (duplicate implementations for transforms); helper fixes for architecture to use new subdags; ILE_batchmode fix if no events to analyze; dag auto-completes if test successful (for subdag system); merge procedure for workers randomizes results, so next iteration isn't dominated by one worker; architecture change to use transverse spins earlier in fit, with suitable prior for sampling, and generally be more efficient for precessing systems; dag checks if composite files are nonempty; partial untested import from Vinaya of using Soichiro mu1,mu2 coordinates for util_AMRGrid; tentative change in 200a505dbad6c3d6911e5043aabfe2880c991545 of xmax in dmarg, pending review [wrong]; pp_RIFT updates including testing d_marg; GMM sampler fix overflow protection, can now analyze high-lnL sources; allow last iteration explode size to be larger than others; fix bug with convert_output_format_ile2inference introduced by change in upstream astropy; more glue.ligo_lw -> ligo.lw and many changes; more on util_AMRGrid.py as refinement engine; improved tests for MC integration tools, validating GMM and mcsamplerGPU; minor fixes (rc12)
- pseudo_pipe/helper updates to use ini files/coinc-embedded PSD appropriate to low-latency; lalsimutils update psd parser; util_InitMargTable undo tentative change noted in rc12; pp_RIFT more flexible ini file parsing (rc13)
- rotated sky coordinates in ILE/ILE_batchmode (not comprehensive, use different adaptation); mcsamplerGPU bugfixes; ILE/ILE_batchmode changes to avoid GPU reallocations; bugfixes for join_grids in dag_utils, cepp_basic subdag system and +flock_local for OSG; reduce imports and superfluous setup for low latency; only generate hlm(t) once in factored_likelihood; better running variance estimate, that GPU-izes; lalsimutils PSD init vectorized; initMargTable save metadata; ongoing increments to amrlib/util_AMRGrid; pseudo_pipe 'last-iteration-extrinsic' and 'batched-extrinsic'; test/provenance to validate information flow; CEPP_basic miscellaneous minor fenceposting (--first-iteration-jumpstart); test/ has integrator tests updated (rosenbrock, Ensemble_extended); tools to let pseudo_pipe take fake data (and generate coincs); pp plot updates from AY; dmarg fix (soichiro) temper-log in all integrators; enhanced initial grids for low mass sources and rf; initial grid wider chieff range for low mass pseudo-pipe recommend GMM correlate mc,delta,s1z at high q; force-away smaller for low-mass events; integrator CI test (rc14)
- TEOBResumS/TEOBResumSE (eccentricity); MultiMetaPipe; PUFF active in subdag; refactor plot_posterior_corner/samples_utils; ILE_batchmode reset sampling if hit certain errors; OSG file transfer mode revitalize (rc15)
Release is rc15
- Since last release
- pseudo_pipe ini parsing (halting bug if fake-cache used) (rc1)
- fix temporary path issue with ini using abs paths, fix typecast to int (rc2)
- fmax ini file parsing (rc3)
release is rc3
- Since last release
- dag_utils missing 'no_grid' when building extrinsic (halting bug); add runmon interface; lalsimutils list() in hlmoft_SEOB_dict; convert_...all2xml updated (rc1)
- ini file srate (rc2)
- CIP/mcsampler cos_theta sampling; pipeline --manual-ifo-list; workers contribute to net goal piecemeal; ini file parser can use fake-cache (rc3)
- waveforms (NRHybSur3dq8Tidal via gwsurrogate; logic for IMRPhenomXP via ModesFromPolarizations; logic for IMRPhenomTP/TPHM); user control over whether pipeline generates precessing analysis (--assume-precessing, --assume-nonprecessing); pseudo_pipe minor (full path to target_params, for ini-file operation) (rc4)
- waveforms (fix typos with IMRPhenomTP), ILE add --force-gpu-only to hard fail if GPU not used (rc5)
- pipeline --force-gpu-only; puffball nan checks; pseudo pipe cache if ini logic fix; FrameZeroNoiseSNR 2to3 (rc6)
- waveforms (ChooseFDModes: PHM,XHM,PXHM, ...), bugfix in --force-gpu-only logic in pseudo_pipe (rc7)
- waveforms (lalsimutils, fix patch) (rc8)
- waveforms (still fixing that damn ChooseFDModes patch) (rc9)
- updating mcsamplerGPU for testing; minor edits to util_CleanILE (skip files of zero length) and ILE (--force-gpu-only logic; change some sys.exit(0) to sys.exit(1)) (rc10)
- tool to save sklearn GPs (not yet used); ILE cupy.show_config; pseudo_pipe not error with --force-gpu-only (rc11)
release is rc11
- Since last release
- range limit on a2 (rc1)
- more xpy==cupy checks in factored_likelihood, protect a lalsimutils coordinate conversion against error, OSG update conventions for using local pool, CEPP add --condor-nogrid-nonworker option to use it, xmlutils fix py3 reduce issue (rc2)
- ILE_batchmode integration window 75ms, xmlutils more missing py2->py3 (rc3)
Release is rc3
- Since last release
- minor py3 errors in PP scripts (rc1)
- import 0.0.14.9rc1-rc5 (rc2)
- minor py2->3 fixes for merged code. config_yank (rc3)
release is rc3
- Since last release
*NR*scripts : fixes for py3/restructure (rc1)
another NR fix (not calling py3 version in NRWriteFrame) (rc2)
- import 0.0.14.8rc1 (rc3)
- import 0.0.14.8rc2 (rc4)
Release is rc4
- Since last release
- py3 port, including most of changes up to 0.0.14.7 (rc1)
- py27 import changes through 0.0.14.7rc4. Add gpytorch. (rc2)
- py27 import changes through 0.0.14.7rc5 (rc3)
- minor fixes for latest py3 (func_code->__code__, 'not subscriptable', / float) (rc4)
release is rc4
- Since last release
bugfix for parsing ini files (indentation error; handling overspecified channel names); pp OSG; NRWriteFrame latest glue; plot_posterior_corner fix tex label issue (rc1)
bugfix ini file parsing (not parsing distance-max) (rc2)
bugfix in ini file use (overriding distance-max if ini used) (rc3)
change ILE time integration window default to 75 ms. --propose-flat-strategy. Better --internal-correlate-parameters arg parsing. Fix enforce_kerr constraint on conversion. RF protect against out of range error. pseudo_pipe GPS->str prevent truncation when moving args around. add --transverse-prior. helper for mc>25 uses mc/delta_mc instead of mc/eta. Add PEsummary output option. Add --general-retries. Pass search --hint-snr in pseudo_pipe. Pass --fref to convert, so reference spins specified correctly. Paths for gwsurrogate.
Other minor non-ILE/CIP modifications (rc4)
infrastructure speed improvements (puffball distance force away function; interpolated cosmology); error protection and handling (workarounds for bugs in error handling in lalsuite); CIP always stream error/out; helper updtes (option for --assume-well-placed to flatten architectures if exploration needs minimal; bugfix highq strategy transverse spin dependence) (rc5)
- Since last release
- bugfix pseudo_pipe (space); pp plot puff enforces mc range; OSG updates (option to copy frames, not cvmfs; local workers; requirements avoid blackhole nodes; minor fixes);
workflow generation test; bugfix NR script restructure; TROUBLESHOOTING (rc1)
- helper fix (cache file name had directory prefix at times); docs (rc2)
Release is rc2
- bugfix in helper introduced by use_ini. PP pipeline. CIP allows arbitrary user-specified priors. Update singularity_base_exe. Fix bitrot to old FactoredLogLikelihood. (rc1)
- automated PP pipeline. Modify BNS tidal grid. CVMFS frames on OSG. Improve NN. (rc2)
- miscellaneous (pp proxies/permissions; pipeline parameter limits; bugfix parsing v4HM in xml) (rc3)
- NR surrogates (gwsurrogate API update; lalsim calls to surrogates; lalsim surrogate is default in pipeline);
puffball more flexible; NN/senni update; other minor (option to cap runtime; plotter; V1 sept 2019 channels) (rc4)
- OSG updates (alt requirements, local universe for non-workers); pp updates (volumetric spins), puffball (force-away), periodic_remove option, bugfix for helper logic for first puffball (rc5)
Release is rc5
Reminder: 0.0.14.x will be the last versions with py27 support; from version 0.0.15 and upward, we should exclusively use py3
- Since last release
- pipline script in main repo (rc1)
- bugfix in GMM integrator interface; in pipeline interface (rc2)
- more bugfixes in GMM (rc3)
- fix access to gp-sparse in CIP
Note GMM, adapt_cart, gp-sparse, rf all validated with this version. Note 0.0.14.x will be the last versions with py27 support; from version 0.0.15 and upward, we should exclusively use py3
- Since last release
- packaging improvements and fixing bugs introduced in restructuring (rc1-rc4)
- fix bug in mcsamplerEnsemble (used with --sampler-method GMM) (rc5)
- helper can parse LI ini files
- Since last release
- Adding CI tests
- minor bugfixes associated with packaging
- minor bugfixes and improvements [helper grid placement at high mass; lnL cutoff for GMM; C-1_nonlinear frame label; other]
- This is the initial release.