sigma default

keysongkang · Aug 20, 2024 · 1ecdf8a · 1ecdf8a
1 parent d5d6565
commit 1ecdf8a
Show file tree

Hide file tree

Showing 4 changed files with 10 additions and 2 deletions.
diff --git a/libs/lib_criteria.py b/libs/lib_criteria.py
@@ -495,8 +495,10 @@ def get_criteria_prob(inputs, Epot_step, uncerts, criteria):
         Prob = np.exp((-1) * (Epot_step / inputs.NumAtoms) / beta)
         Prob_upper_limit = np.exp(
             (-1) * ((criteria.Epotential_avg + criteria.Epotential_std) / inputs.NumAtoms) / beta)
+        # Prob_lower_limit = np.exp(
+        #     (-1) * ((criteria.Epotential_avg + criteria.Epotential_std*1.8) / inputs.NumAtoms) / beta)
         Prob_lower_limit = np.exp(
-            (-1) * ((criteria.Epotential_avg + criteria.Epotential_std*1.8) / inputs.NumAtoms) / beta)
+            (-1) * ((criteria.Epotential_avg - criteria.Epotential_std*2.0) / inputs.NumAtoms) / beta)
 
         # Get relative probability of the canomical ensemble
         criteria_Prob_inter = Prob / Prob_upper_limit;

diff --git a/libs/lib_input.py b/libs/lib_input.py
@@ -126,7 +126,11 @@ def __init__(self, input_file='input.in'):
 
         self.temp_factor = 0.0
 
+        # Set the testing data
+        # test / train
         self.npz_test = 'test'
+
+        # Usage of energy criteria
         self.criteria_energy = True
 
         ##[runMD default inputs]

diff --git a/libs/lib_npz.py b/libs/lib_npz.py
@@ -121,6 +121,8 @@ def generate_npz_DFT_init(inputs, traj, workpath):
 
         if inputs.npz_sigma:
             arrays_to_save['sigma'] = np.array(sigma_train_store)
+        else:
+            arrays_to_Save['sigma'] = np.zeros_like(np.array(E_train_store))
         if inputs.ensemble == 'NPTisoiso' or inputs.train_stress:
             arrays_to_save['stress'] = np.array(stress_train_store)
 

diff --git a/tutorial/ALmoMD_tutorial.md b/tutorial/ALmoMD_tutorial.md
@@ -208,7 +208,7 @@ conda activate almomd
 #[Active learning types]
 calc_type     : active             # (active) sampling up to the number of training data needed (period) sampling during the assigned period
 al_type       : force_max          # Uncertainty type: force_max (Maximum atomic force uncertainty)
-uncert_type   : absolute           # Absolute or relative uncertainty
+uncert_type   : absolute           # Absolute or relative uncertainty (absolute / relative)
 output_format : trajectory.son     # File format of the output file (trajectory.son / aims.out / nequip)
 device        : cuda               # Calating core type (cpu / cuda)