Issues

A setOption line ending in a stray backslash runs in BNG2.pl, but PyBNF refuses it (edition 2) or breaks its simulations (edition 1), and neither message mentions the backslash

lanl/PyBNF#983 · opened

on Oct 5, 2026

An invisible character beside a setOption line (a non-breaking space, for example) makes BNG2.pl skip the line while PyBNF applies it

lanl/PyBNF#982 · opened

on Oct 5, 2026

A setOption written after generate_network is applied by a PyBNF fit but ignored by BNG2.pl, so PyBNF fits a different model from the one in the file

lanl/PyBNF#981 · opened

on Oct 5, 2026

Add a cross-backend differential harness: evaluate one job at a fixed parameter set on every backend that can run it, and compare the simulated outputs and the objective value

lanl/PyBNF#980 · opened

on Oct 5, 2026

Feature request: semiparametric (monotone-spline) link for qualitative/ordinal data — learn the measurement-model shape instead of fixing probit/logit

lanl/PyBNF#979 · opened

on Oct 4, 2026

job_type = check crashes with KeyError when the noise scale is a free parameter (sigma = fit s)

lanl/PyBNF#978 · opened

on Oct 3, 2026

Gradient fits reject an experiment whose only data file is a .prop: 'no routing was supplied'

lanl/PyBNF#977 · opened

on Oct 3, 2026

Gradient fits of function-heavy models spend almost all their time re-running bngsim's SymPy analyses on every evaluation

lanl/PyBNF#976 · opened

on Oct 3, 2026

1

PEtab v2 cannot carry a negative-binomial noise model or the cumulative-to-incident transform: encode the transform with per-period conditions, and add negative-binomial noise as a versioned PyBNF extension

lanl/PyBNF#974 · opened

on Sep 29, 2026

Add an IBIS sampler (job_type = ibis) that updates the exact posterior as observations arrive, absorbs revised observations, and finishes within wall_time_fit

lanl/PyBNF#973 · opened

on Sep 29, 2026

An edition-2 job cannot set ODE solver tolerances for a BNGL model: the model: line refuses rtol:/atol: for .bngl, and the simulations built from experiment: lines never carry them

lanl/PyBNF#971 · opened

on Sep 26, 2026

Check whether the bngsim SBML backend's sensitivity warm-up (#543) serves a stale chain rule when a condition or fitted value writes a derived parameter (lanl/bngsim#708)

lanl/PyBNF#967 · opened

on Sep 25, 2026

A condition that sets the same target twice (perturbations: L = 2, L = 7) is accepted, and the last value silently wins

lanl/PyBNF#964 · opened

on Sep 25, 2026

Follow-up to the PEtab fixes: three valid jobs are refused, by checks added in PRs #940 and #943 and by numbered BNGL action lines that PyBNF does not recognise

lanl/PyBNF#963 · opened

on Sep 25, 2026

Nine loud failures that block a valid job or give the wrong cause: RuleMonkey experiments, one-time-point data, PEtab import and export cases, error messages that hide their reason, and a crash on deeply nested formulas

lanl/PyBNF#962 · opened

on Sep 25, 2026

A noise formula that uses a fixed model parameter (PEtab noiseFormula `0.5 * v3`) passes every load check, then fails each parameter set as it is scored, so the fit either stops with "Your simulations are failing to run" or finishes with an objective of inf

lanl/PyBNF#961 · opened

on Sep 25, 2026

On the bngsim network-free path, a bifurcate() action is skipped without any message, so the simulations after it start from the seed species instead of the state the bifurcate left

lanl/PyBNF#960 · opened

on Sep 25, 2026

With job_type = ms, an ave_norm_sos (column-mean sigma) objective divides each segment by that segment's own data mean, so the segmented stages fit a differently weighted objective from the one the user asked for

lanl/PyBNF#959 · opened

on Sep 25, 2026

Per-row noise values in a measurement_params file are silently ignored when the column's sigma is column_mean, ave_norm_sos or fix_at, because no formula reads them

lanl/PyBNF#958 · opened

on Sep 25, 2026

The PEtab importer ignores a problem.yaml extension marked `required: true`, where the PEtab v2 spec says a tool that does not support it must reject the problem

lanl/PyBNF#957 · opened

on Sep 25, 2026

The PEtab importer accepts a fixed (estimate = false) parameter row whose name matches nothing in the problem, such as the typo V3 for v3, and drops it without a message

lanl/PyBNF#956 · opened

on Sep 25, 2026

A measurement formula that uses log10, log2 or log(x, b) cannot be evaluated, and the fit reports failed simulations instead of naming the formula; a PEtab import can also silently rewrite exp(n*log10(A)) as A^n

lanl/PyBNF#955 · opened

on Sep 25, 2026

The PEtab importer reads a blank time in the experiments table as 0 instead of refusing the problem, so a pre-equilibration whose -inf was left blank is dropped from a dose-response

lanl/PyBNF#954 · opened

on Sep 25, 2026

An experiment with equil_t_end: but no preequilibrate: loads and fits, but the equilibration it asks for is never run

lanl/PyBNF#953 · opened

on Sep 25, 2026

A data file with a negative time is accepted, and the default backend then starts the simulation at that time, so every reading is compared with the model's value too late

lanl/PyBNF#952 · opened

on Sep 25, 2026

Navigation Menu

Repository navigation

Issues

A setOption line ending in a stray backslash runs in BNG2.pl, but PyBNF refuses it (edition 2) or breaks its simulations (edition 1), and neither message mentions the backslash

An invisible character beside a setOption line (a non-breaking space, for example) makes BNG2.pl skip the line while PyBNF applies it

A setOption written after generate_network is applied by a PyBNF fit but ignored by BNG2.pl, so PyBNF fits a different model from the one in the file

Add a cross-backend differential harness: evaluate one job at a fixed parameter set on every backend that can run it, and compare the simulated outputs and the objective value

Feature request: semiparametric (monotone-spline) link for qualitative/ordinal data — learn the measurement-model shape instead of fixing probit/logit

job_type = check crashes with KeyError when the noise scale is a free parameter (sigma = fit s)

Gradient fits reject an experiment whose only data file is a .prop: 'no routing was supplied'

Gradient fits of function-heavy models spend almost all their time re-running bngsim's SymPy analyses on every evaluation

PEtab v2 cannot carry a negative-binomial noise model or the cumulative-to-incident transform: encode the transform with per-period conditions, and add negative-binomial noise as a versioned PyBNF extension

Add an IBIS sampler (job_type = ibis) that updates the exact posterior as observations arrive, absorbs revised observations, and finishes within wall_time_fit

An edition-2 job cannot set ODE solver tolerances for a BNGL model: the model: line refuses rtol:/atol: for .bngl, and the simulations built from experiment: lines never carry them

Check whether the bngsim SBML backend's sensitivity warm-up (#543) serves a stale chain rule when a condition or fitted value writes a derived parameter (lanl/bngsim#708)

A condition that sets the same target twice (perturbations: L = 2, L = 7) is accepted, and the last value silently wins

Follow-up to the PEtab fixes: three valid jobs are refused, by checks added in PRs #940 and #943 and by numbered BNGL action lines that PyBNF does not recognise

Nine loud failures that block a valid job or give the wrong cause: RuleMonkey experiments, one-time-point data, PEtab import and export cases, error messages that hide their reason, and a crash on deeply nested formulas

A noise formula that uses a fixed model parameter (PEtab noiseFormula `0.5 * v3`) passes every load check, then fails each parameter set as it is scored, so the fit either stops with "Your simulations are failing to run" or finishes with an objective of inf

On the bngsim network-free path, a bifurcate() action is skipped without any message, so the simulations after it start from the seed species instead of the state the bifurcate left

With job_type = ms, an ave_norm_sos (column-mean sigma) objective divides each segment by that segment's own data mean, so the segmented stages fit a differently weighted objective from the one the user asked for

Per-row noise values in a measurement_params file are silently ignored when the column's sigma is column_mean, ave_norm_sos or fix_at, because no formula reads them

The PEtab importer ignores a problem.yaml extension marked `required: true`, where the PEtab v2 spec says a tool that does not support it must reject the problem

The PEtab importer accepts a fixed (estimate = false) parameter row whose name matches nothing in the problem, such as the typo V3 for v3, and drops it without a message

A measurement formula that uses log10, log2 or log(x, b) cannot be evaluated, and the fit reports failed simulations instead of naming the formula; a PEtab import can also silently rewrite exp(n*log10(A)) as A^n

The PEtab importer reads a blank time in the experiments table as 0 instead of refusing the problem, so a pre-equilibration whose -inf was left blank is dropped from a dose-response

An experiment with equil_t_end: but no preequilibrate: loads and fits, but the equilibration it asks for is never run

A data file with a negative time is accepted, and the default backend then starts the simulation at that time, so every reading is compared with the model's value too late