pymdkit 1.4.2__tar.gz → 1.4.3__tar.gz

{pymdkit-1.4.2/src/pymdkit.egg-info → pymdkit-1.4.3}/PKG-INFO

@@ -1,6 +1,6 @@
 Metadata-Version: 2.4
 Name: pymdkit
-Version: 1.4.2
+Version: 1.4.3
 Summary: A unified command-line toolkit for atomistic / MD structure workflows.
 Author-email: Yueda Wang <ydwang0608@ustc.edu.cn>
 License-Expression: GPL-3.0-or-later
@@ -175,7 +175,7 @@ When `pca` or `umap` receives `-it train.xyz`, it reads `Config_type` from every
 
 For folder conversion, `convert -if INPUT -of OUTPUT` writes `.xyz` by default. Set another target extension with `-oe`, for example `-oe cif` or `-oe vasp`. Single-file and trajectory modes infer the target format from `-o` and `-ot`.
 
-`gpumd-thermo` writes `temperature.txt`, `potential-energy.txt`, `pressure.txt`, `lattice-parameters.txt`, `volume.txt`, `lattice-angles.txt`, and a headless `thermo.png` in every detected GPUMD job folder. Potential energy is the primary equilibrium signal: sustained potential-energy shifts, drifts, or localized departures define an abnormal job or time region. Lattice parameters, volume, temperature, pressure, and angles remain supporting diagnostics. The analysis uses the last `run` segment, or the last 80% when `run.in` has no `run` record; inspect the plot before making a final scientific judgment.
+`gpumd-thermo` writes `temperature.txt`, `potential-energy.txt`, `pressure.txt`, `lattice-parameters.txt`, `volume.txt`, `lattice-angles.txt`, and a headless `thermo.png` in every detected GPUMD job folder. Potential energy is the primary equilibrium signal. Detection requires a persistent change with a meaningful magnitude relative to the robust natural energy fluctuations, so ordinary correlated thermal noise and isolated spikes are not classified as abnormal. Lattice parameters, volume, temperature, pressure, and angles remain supporting diagnostics. The analysis uses the last `run` segment, or the last 80% when `run.in` has no `run` record; inspect the plot before making a final scientific judgment.
 
 `gpumd-trjcat` recursively scans job subfolders for `traj.xyz`, `thermo.out`, and `run.in`. Stable trajectories are merged in deterministic folder order into `train.xyz` (or `-o`). For an abnormal job, only frames whose times fall in potential-energy abnormal regions are written to `abnormal.xyz` (or `-ao`); frame times are obtained from `time_step` and `dump_xyz` in that job's `run.in`. Missing, unreadable, or unmappable jobs are skipped. When `<job>/model.xyz` contains `Config_type`, that value is applied to every exported frame while all other extended-XYZ bytes remain unchanged.
 

{pymdkit-1.4.2 → pymdkit-1.4.3}/README.md

@@ -149,7 +149,7 @@ When `pca` or `umap` receives `-it train.xyz`, it reads `Config_type` from every
 
 For folder conversion, `convert -if INPUT -of OUTPUT` writes `.xyz` by default. Set another target extension with `-oe`, for example `-oe cif` or `-oe vasp`. Single-file and trajectory modes infer the target format from `-o` and `-ot`.
 
-`gpumd-thermo` writes `temperature.txt`, `potential-energy.txt`, `pressure.txt`, `lattice-parameters.txt`, `volume.txt`, `lattice-angles.txt`, and a headless `thermo.png` in every detected GPUMD job folder. Potential energy is the primary equilibrium signal: sustained potential-energy shifts, drifts, or localized departures define an abnormal job or time region. Lattice parameters, volume, temperature, pressure, and angles remain supporting diagnostics. The analysis uses the last `run` segment, or the last 80% when `run.in` has no `run` record; inspect the plot before making a final scientific judgment.
+`gpumd-thermo` writes `temperature.txt`, `potential-energy.txt`, `pressure.txt`, `lattice-parameters.txt`, `volume.txt`, `lattice-angles.txt`, and a headless `thermo.png` in every detected GPUMD job folder. Potential energy is the primary equilibrium signal. Detection requires a persistent change with a meaningful magnitude relative to the robust natural energy fluctuations, so ordinary correlated thermal noise and isolated spikes are not classified as abnormal. Lattice parameters, volume, temperature, pressure, and angles remain supporting diagnostics. The analysis uses the last `run` segment, or the last 80% when `run.in` has no `run` record; inspect the plot before making a final scientific judgment.
 
 `gpumd-trjcat` recursively scans job subfolders for `traj.xyz`, `thermo.out`, and `run.in`. Stable trajectories are merged in deterministic folder order into `train.xyz` (or `-o`). For an abnormal job, only frames whose times fall in potential-energy abnormal regions are written to `abnormal.xyz` (or `-ao`); frame times are obtained from `time_step` and `dump_xyz` in that job's `run.in`. Missing, unreadable, or unmappable jobs are skipped. When `<job>/model.xyz` contains `Config_type`, that value is applied to every exported frame while all other extended-XYZ bytes remain unchanged.
 

{pymdkit-1.4.2 → pymdkit-1.4.3}/pyproject.toml

@@ -4,7 +4,7 @@ build-backend = "setuptools.build_meta"
 
 [project]
 name = "pymdkit"
-version = "1.4.2"
+version = "1.4.3"
 description = "A unified command-line toolkit for atomistic / MD structure workflows."
 readme = "README.md"
 requires-python = ">=3.9"

{pymdkit-1.4.2 → pymdkit-1.4.3}/src/pymdkit/__init__.py

@@ -1,3 +1,3 @@
 """pymdkit -- a unified CLI for atomistic / MD structure workflows."""
 
-__version__ = "1.4.2"
+__version__ = "1.4.3"

{pymdkit-1.4.2 → pymdkit-1.4.3}/src/pymdkit/commands/_gpumd.py

@@ -195,7 +195,7 @@ def _clean_mask(mask):
 
 
 def potential_abnormal_mask(values):
-    """Locate sustained potential-energy transitions or departures."""
+    """Locate persistent potential-energy transitions or abnormal plateaus."""
     values = np.asarray(values, dtype=float)
     size = len(values)
     if size < 20 or not np.all(np.isfinite(values)):
@@ -204,55 +204,46 @@ def potential_abnormal_mask(values):
     edge_count = max(5, size // 10)
     left_center, left_spread = _robust_spread(values[:edge_count])
     right_center, right_spread = _robust_spread(values[-edge_count:])
+    noise = max(left_spread, right_spread, 1e-12)
     difference = abs(right_center - left_center)
 
-    window = max(5, size // 50)
+    # Average over 1% of the segment so isolated thermal excursions cannot
+    # trigger a warning. Raw values are used again to refine the boundaries.
+    window = max(5, size // 100)
     if window % 2 == 0:
         window += 1
     radius = window // 2
-    smooth = np.asarray([
-        np.median(values[max(0, i - radius):min(size, i + radius + 1)])
-        for i in range(size)
-    ])
+    padded = np.pad(values, (radius, radius), mode="edge")
+    cumulative = np.cumsum(np.insert(padded, 0, 0.0))
+    smooth = (cumulative[window:] - cumulative[:-window]) / window
 
-    mean_threshold = max(
-        6.0 * np.hypot(left_spread, right_spread) / np.sqrt(edge_count),
+    practical_shift = max(
+        1.5 * noise,
         1e-6 * max(abs(left_center), abs(right_center), 1.0),
     )
-    if difference > mean_threshold:
-        # Both edge plateaus are valid equilibrium levels. Mark only samples
-        # between them, not either stable plateau.
-        plateau_tolerance = max(
-            4.0 * max(left_spread, right_spread) / np.sqrt(window),
-            0.02 * difference,
-            1e-12,
-        )
-        distance = np.minimum(
+    if difference > practical_shift:
+        # The ends are two meaningfully different stable plateaus. Keep only
+        # the persistent transition between them.
+        tolerance = max(0.75 * noise, 0.05 * difference)
+        smooth_distance = np.minimum(
             np.abs(smooth - left_center), np.abs(smooth - right_center)
         )
-        transition = _clean_mask(distance > plateau_tolerance)
-        if transition.any():
-            return transition
-
-        # An abrupt jump may contain no intermediate values. In that case,
-        # retain a narrow window around the change rather than one plateau.
-        score = np.zeros(size)
-        for index in range(window, size - window):
-            score[index] = abs(
-                np.mean(smooth[index:index + window])
-                - np.mean(smooth[index - window:index])
-            )
-        threshold = max(
-            0.1 * difference,
-            4.0 * max(left_spread, right_spread) / np.sqrt(window),
+        raw_distance = np.minimum(
+            np.abs(values - left_center), np.abs(values - right_center)
+        )
+        return _clean_mask(
+            (smooth_distance > tolerance) & (raw_distance > tolerance)
         )
-        return _clean_mask(score > threshold)
 
+    # The ends share one baseline. Detect only persistent excursions whose
+    # magnitude is at least twice the natural robust fluctuation scale.
     baseline = np.concatenate((values[:edge_count], values[-edge_count:]))
     center, spread = _robust_spread(baseline)
-    threshold = max(6.0 * spread, 1e-6 * max(abs(center), 1.0), 1e-12)
-    return _clean_mask(np.abs(smooth - center) > threshold)
-
+    threshold = max(2.0 * spread, 1e-6 * max(abs(center), 1.0), 1e-12)
+    return _clean_mask(
+        (np.abs(smooth - center) > threshold)
+        & (np.abs(values - center) > threshold)
+    )
 
 def mask_intervals(time, mask):
     """Return inclusive physical-time intervals for contiguous true mask regions."""

{pymdkit-1.4.2 → pymdkit-1.4.3/src/pymdkit.egg-info}/PKG-INFO

@@ -1,6 +1,6 @@
 Metadata-Version: 2.4
 Name: pymdkit
-Version: 1.4.2
+Version: 1.4.3
 Summary: A unified command-line toolkit for atomistic / MD structure workflows.
 Author-email: Yueda Wang <ydwang0608@ustc.edu.cn>
 License-Expression: GPL-3.0-or-later
@@ -175,7 +175,7 @@ When `pca` or `umap` receives `-it train.xyz`, it reads `Config_type` from every
 
 For folder conversion, `convert -if INPUT -of OUTPUT` writes `.xyz` by default. Set another target extension with `-oe`, for example `-oe cif` or `-oe vasp`. Single-file and trajectory modes infer the target format from `-o` and `-ot`.
 
-`gpumd-thermo` writes `temperature.txt`, `potential-energy.txt`, `pressure.txt`, `lattice-parameters.txt`, `volume.txt`, `lattice-angles.txt`, and a headless `thermo.png` in every detected GPUMD job folder. Potential energy is the primary equilibrium signal: sustained potential-energy shifts, drifts, or localized departures define an abnormal job or time region. Lattice parameters, volume, temperature, pressure, and angles remain supporting diagnostics. The analysis uses the last `run` segment, or the last 80% when `run.in` has no `run` record; inspect the plot before making a final scientific judgment.
+`gpumd-thermo` writes `temperature.txt`, `potential-energy.txt`, `pressure.txt`, `lattice-parameters.txt`, `volume.txt`, `lattice-angles.txt`, and a headless `thermo.png` in every detected GPUMD job folder. Potential energy is the primary equilibrium signal. Detection requires a persistent change with a meaningful magnitude relative to the robust natural energy fluctuations, so ordinary correlated thermal noise and isolated spikes are not classified as abnormal. Lattice parameters, volume, temperature, pressure, and angles remain supporting diagnostics. The analysis uses the last `run` segment, or the last 80% when `run.in` has no `run` record; inspect the plot before making a final scientific judgment.
 
 `gpumd-trjcat` recursively scans job subfolders for `traj.xyz`, `thermo.out`, and `run.in`. Stable trajectories are merged in deterministic folder order into `train.xyz` (or `-o`). For an abnormal job, only frames whose times fall in potential-energy abnormal regions are written to `abnormal.xyz` (or `-ao`); frame times are obtained from `time_step` and `dump_xyz` in that job's `run.in`. Missing, unreadable, or unmappable jobs are skipped. When `<job>/model.xyz` contains `Config_type`, that value is applied to every exported frame while all other extended-XYZ bytes remain unchanged.