pymdkit 1.7.3__tar.gz → 1.7.4__tar.gz

{pymdkit-1.7.3 → pymdkit-1.7.4}/CHANGELOG.md

@@ -1,5 +1,14 @@
 # Changelog
 
+## 1.7.4
+
+- `pmk msd -ne` retains finite, positive Einstein diffusion coefficients and
+  Nernst–Einstein conductivity when the fit-quality diagnostics would reject the
+  window. The result is reported and aggregated as WARNING, with the original
+  diagnostic details preserved. This applies to automatic and manual windows.
+- Missing valid regression, failed requested estimators, and invalid NE inputs
+  still fail the calculation and preserve existing aggregate summaries.
+
 ## 1.7.3
 
 - `pmk msd` analyzes fixed-cell GPUMD NVT trajectories, selects or accepts a

{pymdkit-1.7.3/src/pymdkit.egg-info → pymdkit-1.7.4}/PKG-INFO

@@ -1,6 +1,6 @@
 Metadata-Version: 2.4
 Name: pymdkit
-Version: 1.7.3
+Version: 1.7.4
 Summary: A Python command-line toolkit for high-throughput atomistic workflows in AI-driven materials design.
 Author-email: Yueda Wang <ydwang0608@ustc.edu.cn>
 License-Expression: GPL-3.0-or-later
@@ -36,6 +36,15 @@ materials design. Install `pymdkit` once and call its tools from anywhere as
 Commands expose explicit, documented CLI interfaces designed for reproducible shell,
 HPC, and AI-agent workflows, and each underlying script is still runnable on its own.
 
+## 1.7.4 release notes
+
+- `pmk msd -ne` retains valid Einstein diffusion coefficients and conductivity
+  as WARNING when fit quality would otherwise be REJECT. This applies to
+  automatic and manual windows and preserves the individual diagnostic results.
+- Missing valid regression, failed requested estimators, and invalid NE inputs
+  still cause calculation errors. Without `-ne`, fit-quality rejection still
+  withholds diffusion coefficients.
+
 ## 1.7.3 release notes
 
 - `pmk msd` analyzes GPUMD trajectories from fixed-cell NVT production. OLS is
@@ -97,7 +106,7 @@ installed.
 On a machine with Internet access, run:
 
 ```bash
-python -m pip download pymdkit==1.7.3 -d pymdkit-offline
+python -m pip download pymdkit==1.7.4 -d pymdkit-offline
 ```
 
 This downloads `pymdkit` and the dependencies resolved by pip into the
@@ -105,14 +114,14 @@ This downloads `pymdkit` and the dependencies resolved by pip into the
 system. Then, inside the target pymdkit Python or conda environment, run:
 
 ```bash
-python -m pip install --no-index --find-links=pymdkit-offline pymdkit==1.7.3
+python -m pip install --no-index --find-links=pymdkit-offline pymdkit==1.7.4
 ```
 
 For optional GPU use, prepare the bundle on a networked machine compatible with
 the target HPC platform and Python environment:
 
 ```bash
-python -m pip download "pymdkit[gpu-cuda12]==1.7.3" -d pymdkit-offline
+python -m pip download "pymdkit[gpu-cuda12]==1.7.4" -d pymdkit-offline
 ```
 
 Transfer the whole `pymdkit-offline` directory, then install from it on the
@@ -120,7 +129,7 @@ target system:
 
 ```bash
 python -m pip install --no-index --find-links=pymdkit-offline \
-    "pymdkit[gpu-cuda12]==1.7.3"
+    "pymdkit[gpu-cuda12]==1.7.4"
 ```
 
 This GPU bundle includes the optional `cupy-cuda12x` package. Its wheels must also
@@ -877,22 +886,33 @@ diffusion in cm²/s, `%.2e`) and, only with `-ne`, `msd/sigma.txt` (x/y/z/total
 Nernst–Einstein conductivity in mS/cm, `%.2f`). Headers are comments. Total
 diffusion and conductivity are directional arithmetic means before rounding;
 display rounding may cause last-digit differences. PASS and WARNING write these
-coefficients with status in their comment headers. Fit rejection returns nonzero
-but preserves `msd/msd.txt`, `fit-windows.txt` (all evaluated candidates, rejection
-reasons, beta, directional D, R², intercept, RSS, points and upper-lag origin
-count, intercept diagnostics and selected-window segment D values), and
-`fit-diagnostic.txt` (settings, selection and sensitivity). Beta-qualified window
-sensitivity, all-valid-candidate sensitivity, and three-segment slope spread are
-reported separately. Segments and MSD time origins are not independent samples. Rejected
-fits write no diffusion or conductivity files. Window-to-window spread is a
-sensitivity diagnostic, not a statistical confidence interval. Formal statistical uncertainty is not estimated by default. Existing `msd` directories cause an error: move them aside
-before repeating a calculation. Temporary coordinate storage is always removed.
+coefficients with the quality status in their comment headers. `provenance.json`
+records the final quality and full-precision total coefficients.
+
+With `-ne`, an automatic or manual window with finite, positive x/y/z/total
+Einstein diffusion coefficients retains D and NE conductivity even when the fit
+quality would otherwise be REJECT. The overall status becomes WARNING. The
+original reason and the beta, intercept-consistency, and three-segment slope
+diagnostics remain in the fit output. A WARNING can indicate that the physical
+diffusion regime failed quality criteria. Without `-ne`, REJECT returns nonzero
+and retains only raw MSD and fit diagnostics. PASS and WARNING retain D.
+
+`fit-windows.txt` records evaluated candidates, rejection reasons, beta,
+directional D, R², intercept, RSS, points, upper-lag origin count, intercept
+diagnostics, and selected-window segment D values. `fit-diagnostic.txt` records
+settings, selection, and sensitivity. Beta-qualified window sensitivity,
+all-valid-candidate sensitivity, and three-segment slope spread are reported
+separately. Segments and MSD time origins are correlated. Window-to-window spread
+measures sensitivity. Formal statistical uncertainty requires an explicit
+request. Move an existing `msd` directory aside before repeating a calculation.
+Temporary coordinate storage is removed when the calculation ends.
 
 The five-column raw MSD is published immediately after FFT succeeds, before
 window search, optional uncertainty, or thermodynamic/NE processing. Later errors
-retain this curve and any completed fit diagnostics; parsing/FFT failures are
-reported separately and do not publish a valid curve. REJECT prints the saved
-path and withholds formal diffusion/conductivity files and new aggregates.
+retain this curve and any completed fit diagnostics. A missing valid regression,
+a failed requested WLS/GLS estimator, or invalid required thermodynamic inputs
+returns nonzero and prevents publication of new coefficients and aggregates.
+Parsing and FFT failures occur before the raw curve is published.
 FFT uses float64 atom chunks sized from an internal 128 MiB workspace budget,
 with bounded remapping of time slabs to avoid retaining the entire coordinate
 memmap in RAM. For 100000 frames and 784 selected atoms the temporary coordinate
@@ -962,9 +982,10 @@ missing NVT targets produce an explicit error). Genuine calculation failures ret
 nonzero and preserve previous summaries; skipped templates are not failures.
 A successful run without `-ne` removes any stale root `sigma-total.txt`.
 The companion `msd-quality-total.txt` retains each directory's PASS/WARNING
-status without changing the existing three-column numeric summaries. A rejected
-directory prevents all new aggregate publication; prior summaries remain prior
-results and the terminal/log explicitly reports that they were preserved.
+status. In NE mode, a directory whose valid coefficients were retained despite
+fit-quality rejection appears as WARNING and contributes to both numeric
+summaries. Without `-ne`, a directory with overall REJECT prevents publication
+of new aggregates.
 See [the automatic-fit report](devtools/MSD_AUTO_FIT_REPORT.md) for validation,
 and [the intercept/stability update](devtools/MSD_AUTO_FIT_PHASE4A1_REPORT.md)
 for the current policy, 500 K MSD comparison and limitations.

{pymdkit-1.7.3 → pymdkit-1.7.4}/README.md

@@ -7,6 +7,15 @@ materials design. Install `pymdkit` once and call its tools from anywhere as
 Commands expose explicit, documented CLI interfaces designed for reproducible shell,
 HPC, and AI-agent workflows, and each underlying script is still runnable on its own.
 
+## 1.7.4 release notes
+
+- `pmk msd -ne` retains valid Einstein diffusion coefficients and conductivity
+  as WARNING when fit quality would otherwise be REJECT. This applies to
+  automatic and manual windows and preserves the individual diagnostic results.
+- Missing valid regression, failed requested estimators, and invalid NE inputs
+  still cause calculation errors. Without `-ne`, fit-quality rejection still
+  withholds diffusion coefficients.
+
 ## 1.7.3 release notes
 
 - `pmk msd` analyzes GPUMD trajectories from fixed-cell NVT production. OLS is
@@ -68,7 +77,7 @@ installed.
 On a machine with Internet access, run:
 
 ```bash
-python -m pip download pymdkit==1.7.3 -d pymdkit-offline
+python -m pip download pymdkit==1.7.4 -d pymdkit-offline
 ```
 
 This downloads `pymdkit` and the dependencies resolved by pip into the
@@ -76,14 +85,14 @@ This downloads `pymdkit` and the dependencies resolved by pip into the
 system. Then, inside the target pymdkit Python or conda environment, run:
 
 ```bash
-python -m pip install --no-index --find-links=pymdkit-offline pymdkit==1.7.3
+python -m pip install --no-index --find-links=pymdkit-offline pymdkit==1.7.4
 ```
 
 For optional GPU use, prepare the bundle on a networked machine compatible with
 the target HPC platform and Python environment:
 
 ```bash
-python -m pip download "pymdkit[gpu-cuda12]==1.7.3" -d pymdkit-offline
+python -m pip download "pymdkit[gpu-cuda12]==1.7.4" -d pymdkit-offline
 ```
 
 Transfer the whole `pymdkit-offline` directory, then install from it on the
@@ -91,7 +100,7 @@ target system:
 
 ```bash
 python -m pip install --no-index --find-links=pymdkit-offline \
-    "pymdkit[gpu-cuda12]==1.7.3"
+    "pymdkit[gpu-cuda12]==1.7.4"
 ```
 
 This GPU bundle includes the optional `cupy-cuda12x` package. Its wheels must also
@@ -848,22 +857,33 @@ diffusion in cm²/s, `%.2e`) and, only with `-ne`, `msd/sigma.txt` (x/y/z/total
 Nernst–Einstein conductivity in mS/cm, `%.2f`). Headers are comments. Total
 diffusion and conductivity are directional arithmetic means before rounding;
 display rounding may cause last-digit differences. PASS and WARNING write these
-coefficients with status in their comment headers. Fit rejection returns nonzero
-but preserves `msd/msd.txt`, `fit-windows.txt` (all evaluated candidates, rejection
-reasons, beta, directional D, R², intercept, RSS, points and upper-lag origin
-count, intercept diagnostics and selected-window segment D values), and
-`fit-diagnostic.txt` (settings, selection and sensitivity). Beta-qualified window
-sensitivity, all-valid-candidate sensitivity, and three-segment slope spread are
-reported separately. Segments and MSD time origins are not independent samples. Rejected
-fits write no diffusion or conductivity files. Window-to-window spread is a
-sensitivity diagnostic, not a statistical confidence interval. Formal statistical uncertainty is not estimated by default. Existing `msd` directories cause an error: move them aside
-before repeating a calculation. Temporary coordinate storage is always removed.
+coefficients with the quality status in their comment headers. `provenance.json`
+records the final quality and full-precision total coefficients.
+
+With `-ne`, an automatic or manual window with finite, positive x/y/z/total
+Einstein diffusion coefficients retains D and NE conductivity even when the fit
+quality would otherwise be REJECT. The overall status becomes WARNING. The
+original reason and the beta, intercept-consistency, and three-segment slope
+diagnostics remain in the fit output. A WARNING can indicate that the physical
+diffusion regime failed quality criteria. Without `-ne`, REJECT returns nonzero
+and retains only raw MSD and fit diagnostics. PASS and WARNING retain D.
+
+`fit-windows.txt` records evaluated candidates, rejection reasons, beta,
+directional D, R², intercept, RSS, points, upper-lag origin count, intercept
+diagnostics, and selected-window segment D values. `fit-diagnostic.txt` records
+settings, selection, and sensitivity. Beta-qualified window sensitivity,
+all-valid-candidate sensitivity, and three-segment slope spread are reported
+separately. Segments and MSD time origins are correlated. Window-to-window spread
+measures sensitivity. Formal statistical uncertainty requires an explicit
+request. Move an existing `msd` directory aside before repeating a calculation.
+Temporary coordinate storage is removed when the calculation ends.
 
 The five-column raw MSD is published immediately after FFT succeeds, before
 window search, optional uncertainty, or thermodynamic/NE processing. Later errors
-retain this curve and any completed fit diagnostics; parsing/FFT failures are
-reported separately and do not publish a valid curve. REJECT prints the saved
-path and withholds formal diffusion/conductivity files and new aggregates.
+retain this curve and any completed fit diagnostics. A missing valid regression,
+a failed requested WLS/GLS estimator, or invalid required thermodynamic inputs
+returns nonzero and prevents publication of new coefficients and aggregates.
+Parsing and FFT failures occur before the raw curve is published.
 FFT uses float64 atom chunks sized from an internal 128 MiB workspace budget,
 with bounded remapping of time slabs to avoid retaining the entire coordinate
 memmap in RAM. For 100000 frames and 784 selected atoms the temporary coordinate
@@ -933,9 +953,10 @@ missing NVT targets produce an explicit error). Genuine calculation failures ret
 nonzero and preserve previous summaries; skipped templates are not failures.
 A successful run without `-ne` removes any stale root `sigma-total.txt`.
 The companion `msd-quality-total.txt` retains each directory's PASS/WARNING
-status without changing the existing three-column numeric summaries. A rejected
-directory prevents all new aggregate publication; prior summaries remain prior
-results and the terminal/log explicitly reports that they were preserved.
+status. In NE mode, a directory whose valid coefficients were retained despite
+fit-quality rejection appears as WARNING and contributes to both numeric
+summaries. Without `-ne`, a directory with overall REJECT prevents publication
+of new aggregates.
 See [the automatic-fit report](devtools/MSD_AUTO_FIT_REPORT.md) for validation,
 and [the intercept/stability update](devtools/MSD_AUTO_FIT_PHASE4A1_REPORT.md)
 for the current policy, 500 K MSD comparison and limitations.

{pymdkit-1.7.3 → pymdkit-1.7.4}/pyproject.toml

@@ -4,7 +4,7 @@ build-backend = "setuptools.build_meta"
 
 [project]
 name = "pymdkit"
-version = "1.7.3"
+version = "1.7.4"
 description = "A Python command-line toolkit for high-throughput atomistic workflows in AI-driven materials design."
 readme = "README.md"
 requires-python = ">=3.9"

{pymdkit-1.7.3 → pymdkit-1.7.4}/src/pymdkit/__init__.py

@@ -1,3 +1,3 @@
 """A Python command-line toolkit for high-throughput atomistic workflows in AI-driven materials design."""
 
-__version__ = "1.7.3"
+__version__ = "1.7.4"

{pymdkit-1.7.3 → pymdkit-1.7.4}/src/pymdkit/commands/msd.py

@@ -34,12 +34,12 @@ def production_temperature(production):
 
 
 def add_arguments(parser):
-    parser.epilog = ('Automatic fits use positive MSD lags below 0.7T; beta, intercept consistency '
-                     'and three-segment Einstein slopes diagnose the regime. '
-                     'Einstein regression computes D. PASS/WARNING retain coefficients; REJECT '
-                     'retains msd.txt, fit-windows.txt and fit-diagnostic.txt only. '
-                     'Raw beta rejection can become WARNING only with positive-intercept '
-                     'agreement and acceptable segment slopes. '
+    parser.epilog = ('Automatic fits use positive MSD lags below 0.7T. Beta, intercept consistency '
+                     'and three-segment Einstein slopes diagnose the diffusion regime. '
+                     'Einstein regression calculates D. With -ne, valid Einstein coefficients and '
+                     'conductivity are retained as WARNING if fit quality would otherwise be REJECT. '
+                     'Without -ne, PASS/WARNING retain D and REJECT retains only raw MSD and fit diagnostics. '
+                     'Missing valid regression or invalid NE thermodynamic inputs cause calculation errors. '
                      'Window spread is sensitivity, not a confidence interval. '
                      'Optional --uncertainty covariance reports conditional Bayesian uncertainty separately.')
     parser.add_argument('--uncertainty', choices=['covariance'])
@@ -196,11 +196,15 @@ def analyze(folder, args):
                     (result/'diagnostics.json').write_text(json.dumps(dict(selection=segment_metadata, status='ERROR', reason=str(exc)), indent=2), encoding='utf-8')
                 raise ValueError(f'fitting-window search failed after MSD calculation; raw MSD retained: {exc}') from exc
             selected = diagnostic['selected']
-            if selected is not None and diagnostic['status'] != 'REJECT':
+            diffusion = None
+            fit_error = None
+            if selected is not None and (args.ne or diagnostic['status'] != 'REJECT'):
                 interval = (selected['start'], selected['end'])
                 beta = selected['beta']
-                diffusion = np.array([selected['D_'+k] for k in ('x', 'y', 'z', 'total')])
-                if args.fit_method != 'ols':
+                diffusion = np.array([selected.get('D_'+k, np.nan) for k in ('x', 'y', 'z', 'total')], dtype=float)
+                if args.ne and (not np.isfinite(diffusion).all() or np.any(diffusion <= 0)):
+                    diffusion = None
+                if diffusion is not None and args.fit_method != 'ols':
                     try:
                         fit = fit_directions(mapped, dt, interval, args.fit_method)
                         diffusion = np.array([r['diffusion'] for r in fit['axes'].values()] + [fit['diffusion']])*1e-4
@@ -210,9 +214,19 @@ def analyze(folder, args):
                         diagnostic['method_intercepts_A2'] = {
                             axis: estimate['intercept'] for axis, estimate in fit['axes'].items()
                         }
-                    except (ValueError, RuntimeError, ImportError) as exc:
-                        diagnostic.update(status='REJECT', overall_status='REJECT', reason=str(exc))
-                        selected['overall_status'] = 'REJECT'
+                    except (ValueError, RuntimeError, ImportError, np.linalg.LinAlgError) as exc:
+                        fit_error = f'{args.fit_method.upper()} Einstein regression failed: {exc}' if args.ne else str(exc)
+            if args.ne and diffusion is None:
+                fit_error = f"No valid Einstein regression available. {diagnostic['reason']}"
+            if fit_error is not None:
+                diagnostic.update(status='REJECT', overall_status='REJECT', reason=fit_error)
+                if selected is not None:
+                    selected['overall_status'] = 'REJECT'
+            elif args.ne and diagnostic['status'] == 'REJECT':
+                diagnostic.update(status='WARNING', overall_status='WARNING', reason=(
+                    'NE mode retained valid Einstein coefficients despite the original fit-quality REJECT. '
+                    + diagnostic['reason']))
+                selected['overall_status'] = 'WARNING'
             write_diagnostics(result, diagnostic, duration, dt, len(times), len(indices), args.fit_method, mode,
                               getattr(args, 'mobile_site_distance', None))
             if selection is not None:
@@ -223,7 +237,8 @@ def analyze(folder, args):
             if diagnostic['status'] == 'REJECT':
                 print('  No self-diffusion or NE conductivity output generated.')
                 saved = 'msd/' if selection is None else str(result)
-                raise ValueError('REJECT: '+diagnostic['reason']+f'; MSD and diagnostics saved in {saved}')
+                failure = 'MSD calculation failed' if args.ne else 'REJECT'
+                raise ValueError(f"{failure}: {diagnostic['reason']}; MSD and diagnostics saved in {saved}")
             uncertainty = None
             if getattr(args, 'uncertainty', None):
                 from ._msd_uncertainty import estimate, write_outputs

{pymdkit-1.7.3 → pymdkit-1.7.4/src/pymdkit.egg-info}/PKG-INFO

@@ -1,6 +1,6 @@
 Metadata-Version: 2.4
 Name: pymdkit
-Version: 1.7.3
+Version: 1.7.4
 Summary: A Python command-line toolkit for high-throughput atomistic workflows in AI-driven materials design.
 Author-email: Yueda Wang <ydwang0608@ustc.edu.cn>
 License-Expression: GPL-3.0-or-later
@@ -36,6 +36,15 @@ materials design. Install `pymdkit` once and call its tools from anywhere as
 Commands expose explicit, documented CLI interfaces designed for reproducible shell,
 HPC, and AI-agent workflows, and each underlying script is still runnable on its own.
 
+## 1.7.4 release notes
+
+- `pmk msd -ne` retains valid Einstein diffusion coefficients and conductivity
+  as WARNING when fit quality would otherwise be REJECT. This applies to
+  automatic and manual windows and preserves the individual diagnostic results.
+- Missing valid regression, failed requested estimators, and invalid NE inputs
+  still cause calculation errors. Without `-ne`, fit-quality rejection still
+  withholds diffusion coefficients.
+
 ## 1.7.3 release notes
 
 - `pmk msd` analyzes GPUMD trajectories from fixed-cell NVT production. OLS is
@@ -97,7 +106,7 @@ installed.
 On a machine with Internet access, run:
 
 ```bash
-python -m pip download pymdkit==1.7.3 -d pymdkit-offline
+python -m pip download pymdkit==1.7.4 -d pymdkit-offline
 ```
 
 This downloads `pymdkit` and the dependencies resolved by pip into the
@@ -105,14 +114,14 @@ This downloads `pymdkit` and the dependencies resolved by pip into the
 system. Then, inside the target pymdkit Python or conda environment, run:
 
 ```bash
-python -m pip install --no-index --find-links=pymdkit-offline pymdkit==1.7.3
+python -m pip install --no-index --find-links=pymdkit-offline pymdkit==1.7.4
 ```
 
 For optional GPU use, prepare the bundle on a networked machine compatible with
 the target HPC platform and Python environment:
 
 ```bash
-python -m pip download "pymdkit[gpu-cuda12]==1.7.3" -d pymdkit-offline
+python -m pip download "pymdkit[gpu-cuda12]==1.7.4" -d pymdkit-offline
 ```
 
 Transfer the whole `pymdkit-offline` directory, then install from it on the
@@ -120,7 +129,7 @@ target system:
 
 ```bash
 python -m pip install --no-index --find-links=pymdkit-offline \
-    "pymdkit[gpu-cuda12]==1.7.3"
+    "pymdkit[gpu-cuda12]==1.7.4"
 ```
 
 This GPU bundle includes the optional `cupy-cuda12x` package. Its wheels must also
@@ -877,22 +886,33 @@ diffusion in cm²/s, `%.2e`) and, only with `-ne`, `msd/sigma.txt` (x/y/z/total
 Nernst–Einstein conductivity in mS/cm, `%.2f`). Headers are comments. Total
 diffusion and conductivity are directional arithmetic means before rounding;
 display rounding may cause last-digit differences. PASS and WARNING write these
-coefficients with status in their comment headers. Fit rejection returns nonzero
-but preserves `msd/msd.txt`, `fit-windows.txt` (all evaluated candidates, rejection
-reasons, beta, directional D, R², intercept, RSS, points and upper-lag origin
-count, intercept diagnostics and selected-window segment D values), and
-`fit-diagnostic.txt` (settings, selection and sensitivity). Beta-qualified window
-sensitivity, all-valid-candidate sensitivity, and three-segment slope spread are
-reported separately. Segments and MSD time origins are not independent samples. Rejected
-fits write no diffusion or conductivity files. Window-to-window spread is a
-sensitivity diagnostic, not a statistical confidence interval. Formal statistical uncertainty is not estimated by default. Existing `msd` directories cause an error: move them aside
-before repeating a calculation. Temporary coordinate storage is always removed.
+coefficients with the quality status in their comment headers. `provenance.json`
+records the final quality and full-precision total coefficients.
+
+With `-ne`, an automatic or manual window with finite, positive x/y/z/total
+Einstein diffusion coefficients retains D and NE conductivity even when the fit
+quality would otherwise be REJECT. The overall status becomes WARNING. The
+original reason and the beta, intercept-consistency, and three-segment slope
+diagnostics remain in the fit output. A WARNING can indicate that the physical
+diffusion regime failed quality criteria. Without `-ne`, REJECT returns nonzero
+and retains only raw MSD and fit diagnostics. PASS and WARNING retain D.
+
+`fit-windows.txt` records evaluated candidates, rejection reasons, beta,
+directional D, R², intercept, RSS, points, upper-lag origin count, intercept
+diagnostics, and selected-window segment D values. `fit-diagnostic.txt` records
+settings, selection, and sensitivity. Beta-qualified window sensitivity,
+all-valid-candidate sensitivity, and three-segment slope spread are reported
+separately. Segments and MSD time origins are correlated. Window-to-window spread
+measures sensitivity. Formal statistical uncertainty requires an explicit
+request. Move an existing `msd` directory aside before repeating a calculation.
+Temporary coordinate storage is removed when the calculation ends.
 
 The five-column raw MSD is published immediately after FFT succeeds, before
 window search, optional uncertainty, or thermodynamic/NE processing. Later errors
-retain this curve and any completed fit diagnostics; parsing/FFT failures are
-reported separately and do not publish a valid curve. REJECT prints the saved
-path and withholds formal diffusion/conductivity files and new aggregates.
+retain this curve and any completed fit diagnostics. A missing valid regression,
+a failed requested WLS/GLS estimator, or invalid required thermodynamic inputs
+returns nonzero and prevents publication of new coefficients and aggregates.
+Parsing and FFT failures occur before the raw curve is published.
 FFT uses float64 atom chunks sized from an internal 128 MiB workspace budget,
 with bounded remapping of time slabs to avoid retaining the entire coordinate
 memmap in RAM. For 100000 frames and 784 selected atoms the temporary coordinate
@@ -962,9 +982,10 @@ missing NVT targets produce an explicit error). Genuine calculation failures ret
 nonzero and preserve previous summaries; skipped templates are not failures.
 A successful run without `-ne` removes any stale root `sigma-total.txt`.
 The companion `msd-quality-total.txt` retains each directory's PASS/WARNING
-status without changing the existing three-column numeric summaries. A rejected
-directory prevents all new aggregate publication; prior summaries remain prior
-results and the terminal/log explicitly reports that they were preserved.
+status. In NE mode, a directory whose valid coefficients were retained despite
+fit-quality rejection appears as WARNING and contributes to both numeric
+summaries. Without `-ne`, a directory with overall REJECT prevents publication
+of new aggregates.
 See [the automatic-fit report](devtools/MSD_AUTO_FIT_REPORT.md) for validation,
 and [the intercept/stability update](devtools/MSD_AUTO_FIT_PHASE4A1_REPORT.md)
 for the current policy, 500 K MSD comparison and limitations.

{pymdkit-1.7.3 → pymdkit-1.7.4}/tests/test_msd_large.py

@@ -64,7 +64,7 @@ class LargeMSDTests(unittest.TestCase):
         args.uncertainty = 'covariance'
         args.min_fit_window = 1.
         with patch('pymdkit.commands._msd_uncertainty.estimate', side_effect=RuntimeError('unavailable')) as estimate:
-            with self.assertRaisesRegex(ValueError, 'REJECT'):
+            with self.assertRaisesRegex(ValueError, 'No valid Einstein regression'):
                 msd.analyze(self.root, args)
             estimate.assert_not_called()
         for name in ('msd.txt', 'fit-windows.txt', 'fit-diagnostic.txt'):

{pymdkit-1.7.3 → pymdkit-1.7.4}/tests/test_msd_phase3.py

@@ -1,3 +1,5 @@
+import argparse
+import json
 import os
 from pathlib import Path
 import subprocess
@@ -13,7 +15,8 @@ from pymdkit.commands._gpumd_run import parse_run, trajectory_schedule
 from pymdkit.commands._msd_thermo import load_thermo
 from pymdkit.commands._msd_conductivity import conductivity, ELEMENTARY_CHARGE, BOLTZMANN
 from pymdkit.commands._msd_coordinates import Coordinates
-from pymdkit.commands import msd
+from pymdkit.commands import _msd_auto, _msd_fit, msd
+from pymdkit.commands._pred_msd_plan import read_plan
 
 
 HPC_RUN = '''time_step 1.0
@@ -258,24 +261,24 @@ class Phase3Tests(unittest.TestCase):
         with self.assertRaisesRegex(ValueError, 'ordering'):
             p.update(f)
 
-    def fixture(self, folder):
+    def fixture(self, folder, *, retention=1., step_scale=.05, temperature=300):
         folder.mkdir(exist_ok=True)
         n, count = 129, 64
         model = Atoms('Li'*count+'O'*count, positions=np.zeros((count*2, 3)), cell=[100]*3, pbc=True)
         model.arrays['group'] = np.r_[np.zeros(count, dtype=int), np.ones(count, dtype=int)]
         write(folder/'model.xyz', model)
-        (folder/'run.in').write_text(f'time_step 1\nensemble nvt_nhc 300 300 100\ndump_thermo 1\ndump_exyz 1 1 1\nrun {n}\n')
+        (folder/'run.in').write_text(f'time_step 1\nensemble nvt_nhc {temperature} {temperature} 100\ndump_thermo 1\ndump_exyz 1 1 1\nrun {n}\n')
         rng = np.random.default_rng(17)
         raw = np.zeros((count*2, 3))
         for i in range(n):
-            raw[:count] += rng.normal(0, .05, (count, 3))
+            raw[:count] = retention*raw[:count] + rng.normal(0, step_scale, (count, 3))
             f = model.copy()
             f.positions = raw % 100
             f.arrays['unwrapped_position'] = raw.copy()
             f.info['Time'] = i+1
             write(folder/'dump.xyz', f, append=i>0)
         data = np.zeros((n, 18))
-        data[:, 0] = 300
+        data[:, 0] = temperature
         data[:, 9:] = np.tile(np.diag([100]*3).ravel(), (n, 1))
         np.savetxt(folder/'thermo.out', data)
 
@@ -310,7 +313,7 @@ class Phase3Tests(unittest.TestCase):
         args = type('Args', (), dict(group=0, ne=True, charge=1., fit_range=None,
                                     fit_method='ols', min_fit_window=1.))()
         with patch.object(_msd_fit, 'fft_msd', wraps=_msd_fit.fft_msd) as fft:
-            with self.assertRaisesRegex(ValueError, 'REJECT'):
+            with self.assertRaisesRegex(ValueError, 'No valid Einstein regression'):
                 msd.analyze(self.root, args)
             self.assertEqual(fft.call_count, 1)
         for name in ('msd.txt', 'fit-windows.txt', 'fit-diagnostic.txt'):
@@ -347,6 +350,129 @@ class Phase3Tests(unittest.TestCase):
         for name in ('self-diff.txt', 'sigma.txt'):
             self.assertIn('quality=WARNING', (self.root/'msd'/name).read_text())
 
+    def test_quality_rejection_with_valid_regression(self):
+        t = np.arange(129)*.001
+        values = t[:, None]**.6*np.array([.5, .6, .7])
+        parser = argparse.ArgumentParser()
+        msd.add_arguments(parser)
+        for name, flags in (
+            ('non-ne', []),
+            ('auto-ne', ['-ne']),
+            ('manual-ne', ['-ne', '-fit-range', '.01', '.075', '-trajectory-range', '0', '.000129']),
+        ):
+            with self.subTest(mode=name):
+                folder = self.root/name
+                self.fixture(folder)
+                args = parser.parse_args(['-group', '0', *flags])
+                original = _msd_auto.search(t, values, .129, interval=args.fit_range)
+                self.assertEqual(original['status'], 'REJECT')
+                self.assertIsNotNone(original['selected'])
+                directions = ('x', 'y', 'z', 'total')
+                expected = np.array([original['selected']['D_'+axis] for axis in directions])
+                self.assertTrue(np.all(np.isfinite(expected) & (expected > 0)))
+                with patch.object(Path, 'cwd', return_value=self.root), patch.object(
+                        _msd_fit, 'fft_msd', return_value=values):
+                    if args.ne:
+                        result = msd.analyze(folder, args)
+                    else:
+                        with self.assertRaisesRegex(ValueError, '^REJECT: Raw beta'):
+                            msd.analyze(folder, args)
+
+                out = folder/'msd'
+                if args.trajectory_range:
+                    out = out/'segments/0-0.000129ns'
+                    fit = json.loads((out/'diagnostics.json').read_text())['fit']
+                    self.assertEqual(fit['status'], 'WARNING')
+                    self.assertEqual(fit['overall_status'], 'WARNING')
+                    self.assertEqual(fit['selected']['overall_status'], 'WARNING')
+                    self.assertEqual(fit['segments'], original['segments'])
+                self.assertTrue((out/'msd.txt').is_file())
+                windows = (out/'fit-windows.txt').read_text()
+                diagnostic = (out/'fit-diagnostic.txt').read_text()
+                for key in ('beta_status', 'intercept_status', 'einstein_slope_status'):
+                    self.assertIn(f'\n{key}: {original[key]}\n', diagnostic)
+                self.assertIn(original['reason'], diagnostic)
+                if not args.ne:
+                    self.assertIn('\nstatus: REJECT\n', diagnostic)
+                    for output in ('self-diff.txt', 'sigma.txt', 'provenance.json'):
+                        self.assertFalse((out/output).exists())
+                    continue
+                self.assertEqual(result.quality, 'WARNING')
+                self.assertIn('\nstatus: WARNING\n', diagnostic)
+                self.assertIn('\noverall_status: WARNING\n', diagnostic)
+                self.assertIn('NE mode retained valid Einstein coefficients', diagnostic)
+                self.assertTrue(any(row.endswith('\tWARNING') for row in windows.splitlines()))
+                expected_sigma = conductivity(expected, 64, 1, 1e6, 300, ['Li'])
+                self.assertAlmostEqual(result.diffusion, expected[-1])
+                self.assertAlmostEqual(result.conductivity, expected_sigma[-1])
+                for output, prefix, coefficients, fmt in (
+                    ('self-diff.txt', 'D', expected, '.2e'),
+                    ('sigma.txt', 'sigma', expected_sigma, '.2f'),
+                ):
+                    header, *rows = (out/output).read_text().splitlines()
+                    self.assertIn('quality=WARNING', header)
+                    self.assertEqual([row.split() for row in rows], [
+                        [f'{prefix}_{axis}', format(value, fmt)] for axis, value in zip(directions, coefficients)])
+                provenance = json.loads((out/'provenance.json').read_text())
+                self.assertEqual(provenance['quality'], 'WARNING')
+                self.assertEqual(provenance['diffusion_cm2_s'], result.diffusion)
+                self.assertEqual(provenance['conductivity_mS_cm'], result.conductivity)
+
+    def test_ne_rejected_quality_does_not_hide_weighted_estimator_failure(self):
+        t = np.arange(129)*.001
+        values = t[:, None]**.6*np.array([.5, .6, .7])
+        for method in ('wls', 'gls'):
+            with self.subTest(method=method):
+                folder = self.root/method
+                self.fixture(folder)
+                args = argparse.Namespace(group=0, ne=True, charge=1., fit_range=[.013, .064], fit_method=method)
+                self.assertEqual(_msd_auto.search(t, values, .129, interval=args.fit_range)['status'], 'REJECT')
+                with patch.object(_msd_fit, 'fft_msd', return_value=values):
+                    with self.assertRaisesRegex(ValueError, f'{method.upper()} Einstein regression failed: .*binary points'):
+                        msd.analyze(folder, args)
+                for output in ('msd.txt', 'fit-windows.txt', 'fit-diagnostic.txt'):
+                    self.assertTrue((folder/'msd'/output).is_file())
+                for output in ('self-diff.txt', 'sigma.txt', 'provenance.json'):
+                    self.assertFalse((folder/'msd'/output).exists())
+
+    def test_cli_ne_aggregates_rejected_quality_and_predicts_from_warning(self):
+        confined, brownian = self.root/'confined', self.root/'brownian'
+        self.fixture(confined, retention=.96)
+        self.fixture(brownian, temperature=500)
+        result = self.cli('msd', '-group', '0', '-ne')
+        self.assertEqual(result.returncode, 0, result.stdout+result.stderr)
+        for folder in (confined, brownian):
+            for output in ('self-diff.txt', 'sigma.txt', 'provenance.json'):
+                self.assertTrue((folder/'msd'/output).is_file())
+        for output in ('self-diff-total.txt', 'sigma-total.txt'):
+            _, *rows = (self.root/output).read_text().splitlines()
+            self.assertEqual([row.split()[0] for row in rows], ['confined', 'brownian'])
+        self.assertIn('confined WARNING', (self.root/'msd-quality-total.txt').read_text())
+        diagnostic = (confined/'msd/fit-diagnostic.txt').read_text()
+        self.assertIn('NE mode retained valid Einstein coefficients', diagnostic)
+        result = self.cli('pred-msd-time', '-ea', '.2', '-temp-interval', '50', '-max-temp', '350')
+        self.assertEqual(result.returncode, 0, result.stdout+result.stderr)
+        plan = read_plan(self.root/'pred-msd-time.log')
+        self.assertEqual(plan['reference']['temperature_K'], 300)
+        self.assertEqual(plan['reference']['quality'], 'WARNING')
+        self.assertTrue(any('BASELINE WARNING' in warning for warning in plan['warnings']))
+
+    def test_cli_ne_zero_msd_preserves_diagnostics_and_aggregate_summaries(self):
+        self.fixture(self.root, step_scale=0.)
+        msd.write_summaries(self.root, [msd.AnalysisResult('prior', 300., 1.e-5, 1.)], ne=True)
+        summaries = {p: p.read_bytes() for p in self.root.glob('*-total.txt')}
+        result = self.cli('msd', '-group', '0', '-ne')
+        self.assertNotEqual(result.returncode, 0, result.stdout)
+        self.assertIn('No valid Einstein regression', result.stdout)
+        self.assertIn('existing summaries preserved', result.stdout)
+        out = self.root/'msd'
+        np.testing.assert_array_equal(np.loadtxt(out/'msd.txt')[:, 1:], 0.)
+        for output in ('fit-windows.txt', 'fit-diagnostic.txt'):
+            self.assertTrue((out/output).is_file())
+        for output in ('self-diff.txt', 'sigma.txt', 'provenance.json'):
+            self.assertFalse((out/output).exists())
+        self.assertEqual({p: p.read_bytes() for p in summaries}, summaries)
+
     def test_cli_multi_temperature_templates_and_failure_preserves_summary(self):
         for name, target, measured in [('hot', 1000, 1001), ('cold-b', 300, 301), ('cold-a', 300, 301)]:
             folder = self.root/name