Merge pull request #5 from arcann-chem/develop

Update

README.md

@@ -392,7 +392,7 @@ Since `initialization` has only a single `start` *phase*, which is self-explanat
 
 | Phase | Description |
 | --- | --- |
-| `prepare` | Prepares folders and files for training the user-defined number of independent NNPs to be used in the next iteration. |
+| `prepare` | Prepares folders and files for training and the user-defined number of independent NNPs to be used in the next iteration. |
 | `launch` | Submits the training calculations using the `dp train` code from DeePMD-kit. |
 | `check` | Verifies whether the training has completed successfully. If any traoining ended abruptly, they need to be resubmitted manually to ensure the training finishes correctly. |
 | `freeze` | Freezes the NN parameters into a binary file (`.pb` extension for TensorFlow back-end) usable with LAMMPS and Python. This is done with the `dp freeze` code from DeePMD-kit. |

arcann_training/exploration/check.py

@@ -6,7 +6,7 @@
 #   SPDX-License-Identifier: AGPL-3.0-only                                                           #
 #----------------------------------------------------------------------------------------------------#
 Created: 2022/01/01
-Last modified: 2024/05/15
+Last modified: 2024/05/26
 """
 
 # Standard library modules
@@ -20,7 +20,7 @@
 
 # Local imports
 from arcann_training.common.json import load_json_file, write_json_file, get_key_in_dict, load_default_json_file, backup_and_overwrite_json_file
-from arcann_training.common.list import textfile_to_string_list
+from arcann_training.common.list import textfile_to_string_list, string_list_to_textfile
 from arcann_training.common.check import validate_step_folder, check_vmd, check_dcd_is_valid, check_nc_is_valid
 
 
@@ -111,6 +111,7 @@ def main(
     completed_count = 0
     skipped_count = 0
     forced_count = 0
+    failed_explorations_list = []
 
     for system_auto_index, system_auto in enumerate(main_json["systems_auto"]):
         # Counters
@@ -173,6 +174,7 @@ def main(
                         timings.append(float(timings_str[0].split(" ")[3]))
                         del timings_str
                     else:
+                        failed_explorations_list.append(f"{lammps_output_file.parent}")
                         arcann_logger.critical(f"'{lammps_output_file}' failed. Check manually.")
                     del lammps_output, lammps_output_file, traj_file, model_deviation_filename
 
@@ -213,6 +215,7 @@ def main(
                         timings.append(float(matches[0]))
                         del timings_str
                     else:
+                        failed_explorations_list.append(f"{lammps_output_file.parent}")
                         arcann_logger.critical(f"'{sander_emle_output_file}' failed. Check manually.")
                     del sander_emle_ouput, sander_emle_output_file, traj_file, model_deviation_filename
 
@@ -241,6 +244,7 @@ def main(
                         timings.append(np.average(np.asarray(ipi_time2, dtype="float32")))
                         del ipi_time, ipi_time2
                     else:
+                        failed_explorations_list.append(f"{lammps_output_file.parent}")
                         arcann_logger.critical(f"'{ipi_output_file}' failed. Check manually.")
                         del ipi_output
                     del ipi_output_file
@@ -285,7 +289,10 @@ def main(
         del timings, average_per_step, system_count
 
     del system_auto, it_nnp, it_number
-
+    if failed_explorations_list:
+        string_list_to_textfile((current_path / "failed_explorations.txt"), failed_explorations_list, read_only=True)
+        arcann_logger.critical(f"Failed explorations are listed in 'failed_explorations.txt'.")
+
     arcann_logger.info(f"-" * 88)
     # Update the booleans in the exploration JSON
     if (completed_count + skipped_count + forced_count) == (exploration_json["nnp_count"] * sum([exploration_json["systems_auto"][_]["traj_count"] for _ in exploration_json["systems_auto"]])):

arcann_training/exploration/prepare.py

@@ -6,7 +6,7 @@
 #   SPDX-License-Identifier: AGPL-3.0-only                                                           #
 #----------------------------------------------------------------------------------------------------#
 Created: 2022/01/01
-Last modified: 2024/05/15
+Last modified: 2024/05/17
 """
 
 # Standard library modules
@@ -389,7 +389,7 @@ def main(
                     system_nb_steps = system_exp_time_ps / system_timestep_ps
                 # Auto value
                 else:
-                    system_nb_steps = update_system_nb_steps_factor(previous_exploration_json, system_auto)
+                    system_nb_steps = update_system_nb_steps_factor(previous_exploration_json, system_auto) / system_timestep_ps
                     # Update if over Max value
                     if system_nb_steps > system_max_exp_time_ps / system_timestep_ps:
                         system_nb_steps = system_max_exp_time_ps / system_timestep_ps
@@ -559,7 +559,7 @@ def main(
                     system_nb_steps = system_exp_time_ps / system_timestep_ps
                 # Auto value
                 else:
-                    system_nb_steps *= update_system_nb_steps_factor(previous_exploration_json, system_auto)
+                    system_nb_steps *= update_system_nb_steps_factor(previous_exploration_json, system_auto) / system_timestep_ps
                     # Update if over Max value
                     if system_nb_steps > (system_max_exp_time_ps / system_timestep_ps):
                         system_nb_steps = system_max_exp_time_ps / system_timestep_ps

arcann_training/exploration/utils.py

@@ -6,7 +6,7 @@
 #   SPDX-License-Identifier: AGPL-3.0-only                                                           #
 #----------------------------------------------------------------------------------------------------#
 Created: 2022/01/01
-Last modified: 2024/05/15
+Last modified: 2024/05/17
 
 Functions
 ---------
@@ -761,6 +761,7 @@ def get_last_frame_number(model_deviation: np.ndarray, sigma_high_limit: float,
     return last_frame
 
 
+# TODO: Sould be renamed because it is not returning a factor or a number of steps but a time of simulation
 # Unittested
 @catch_errors_decorator
 def update_system_nb_steps_factor(previous_json: Dict, system_auto_index: int) -> int:

arcann_training/labeling/prepare.py

@@ -6,7 +6,7 @@
 #   SPDX-License-Identifier: AGPL-3.0-only                                                           #
 #----------------------------------------------------------------------------------------------------#
 Created: 2022/01/01
-Last modified: 2024/05/15
+Last modified: 2024/05/26
 """
 
 # Standard library modules
@@ -360,15 +360,15 @@ def main(
             first_job_input_t = deepcopy(system_first_job_input)
             first_job_input_t = replace_substring_in_string_list(first_job_input_t, "_R_PADDEDSTEP_", padded_labeling_step)
             if labeling_program == "cp2k":
-                first_job_input_t = replace_substring_in_string_list(first_job_input_t, "_R_CELL_", " ".join([str(_) for _ in [cell_info[0][i] for i in [0, 4, 8]]]))
+                first_job_input_t = replace_substring_in_string_list(first_job_input_t, "_R_CELL_", " ".join([str(_) for _ in [cell_info[labeling_step][i] for i in [0, 4, 8]]]))
 
             string_list_to_textfile(labeling_step_path / f"1_labeling_{padded_labeling_step}.inp", first_job_input_t)
             del first_job_input_t
 
             if labeling_program == "cp2k":
                 second_job_input_t = deepcopy(system_second_job_input)
                 second_job_input_t = replace_substring_in_string_list(second_job_input_t, "_R_PADDEDSTEP_", padded_labeling_step)
-                second_job_input_t = replace_substring_in_string_list(second_job_input_t, "_R_CELL_", " ".join([str(_) for _ in [cell_info[0][i] for i in [0, 4, 8]]]))
+                second_job_input_t = replace_substring_in_string_list(second_job_input_t, "_R_CELL_", " ".join([str(_) for _ in [cell_info[labeling_step][i] for i in [0, 4, 8]]]))
 
                 string_list_to_textfile(labeling_step_path / f"2_labeling_{padded_labeling_step}.inp", second_job_input_t)
                 del second_job_input_t
@@ -416,13 +416,13 @@ def main(
 
                 first_job_input_t = deepcopy(system_first_job_input)
                 first_job_input_t = replace_substring_in_string_list(first_job_input_t, "_R_PADDEDSTEP_", padded_labeling_step)
-                first_job_input_t = replace_substring_in_string_list(first_job_input_t, "_R_CELL_", " ".join([str(_) for _ in [cell_info[0][i] for i in [0, 4, 8]]]))
+                first_job_input_t = replace_substring_in_string_list(first_job_input_t, "_R_CELL_", " ".join([str(_) for _ in [cell_info[labeling_step][i] for i in [0, 4, 8]]]))
                 string_list_to_textfile(labeling_step_path / f"1_labeling_{padded_labeling_step}.inp", first_job_input_t)
                 del first_job_input_t
                 if labeling_program == "cp2k":
                     second_job_input_t = deepcopy(system_second_job_input)
                     second_job_input_t = replace_substring_in_string_list(second_job_input_t, "_R_PADDEDSTEP_", padded_labeling_step)
-                    second_job_input_t = replace_substring_in_string_list(second_job_input_t, "_R_CELL_", " ".join([str(_) for _ in [cell_info[0][i] for i in [0, 4, 8]]]))
+                    second_job_input_t = replace_substring_in_string_list(second_job_input_t, "_R_CELL_", " ".join([str(_) for _ in [cell_info[labeling_step][i] for i in [0, 4, 8]]]))
                     string_list_to_textfile(labeling_step_path / f"2_labeling_{padded_labeling_step}.inp", second_job_input_t)
                     del second_job_input_t
 

tools/force_explorations.sh

@@ -0,0 +1,19 @@
+#!/bin/bash
+#----------------------------------------------------------------------------------------------------#
+#   ArcaNN: Automatic training of Reactive Chemical Architecture with Neural Networks                #
+#   Copyright 2022-2024 ArcaNN developers group <https://github.com/arcann-chem>                     #
+#                                                                                                    #
+#   SPDX-License-Identifier: AGPL-3.0-only                                                           #
+#----------------------------------------------------------------------------------------------------#
+# Created: 2024/05/26
+# Last modified: 2024/05/26
+#----------------------------------------------
+
+# Read the file line by line
+while IFS= read -r directory; do
+    # Remove leading/trailing whitespaces
+    directory=$(echo "$directory" | xargs)
+
+    # Create the "force" file in the director
+    touch "$directory"/force && echo "touch ${directory}/force"
+done < failed_explorations.txt