omniopt2 8331__tar.gz → 8341__tar.gz

{omniopt2-8331 → omniopt2-8341}/.omniopt.py

@@ -2133,25 +2133,25 @@ def try_saving_to_db() -> None:
     except Exception as e:
         print_debug(f"Failed trying to save sqlite3-DB: {e}")
 
-def merge_with_job_infos(pd_frame: pd.DataFrame) -> pd.DataFrame:
+def merge_with_job_infos(df: pd.DataFrame) -> pd.DataFrame:
     job_infos_path = os.path.join(get_current_run_folder(), "job_infos.csv")
     if not os.path.exists(job_infos_path):
-        return pd_frame
+        return df
 
     job_df = pd.read_csv(job_infos_path)
 
-    if 'trial_index' not in pd_frame.columns or 'trial_index' not in job_df.columns:
+    if 'trial_index' not in df.columns or 'trial_index' not in job_df.columns:
         raise ValueError("Both DataFrames must contain a 'trial_index' column.")
 
-    job_df_filtered = job_df[job_df['trial_index'].isin(pd_frame['trial_index'])]
+    job_df_filtered = job_df[job_df['trial_index'].isin(df['trial_index'])]
 
-    new_cols = [col for col in job_df_filtered.columns if col != 'trial_index' and col not in pd_frame.columns]
+    new_cols = [col for col in job_df_filtered.columns if col != 'trial_index' and col not in df.columns]
 
     job_df_reduced = job_df_filtered[['trial_index'] + new_cols]
 
-    merged = pd.merge(pd_frame, job_df_reduced, on='trial_index', how='left')
+    merged = pd.merge(df, job_df_reduced, on='trial_index', how='left')
 
-    old_cols = [col for col in pd_frame.columns if col != 'trial_index']
+    old_cols = [col for col in df.columns if col != 'trial_index']
 
     new_order = ['trial_index'] + new_cols + old_cols
 
@@ -2159,47 +2159,6 @@ def merge_with_job_infos(pd_frame: pd.DataFrame) -> pd.DataFrame:
 
     return merged
 
-def reindex_trials(df: pd.DataFrame) -> pd.DataFrame:
-    """
-    Ensure trial_index is sequential and arm_name unique.
-    Keep arm_name unless all parameters except 'order', 'hostname', 'queue_time' match.
-    """
-    if "trial_index" not in df.columns or "arm_name" not in df.columns:
-        return df
-
-    # Sort by something stable (queue_time if available)
-    sort_cols = ["queue_time"] if "queue_time" in df.columns else df.columns.tolist()
-    df = df.sort_values(by=sort_cols, ignore_index=True)
-
-    # Mapping from "parameter signature" to assigned arm_name
-    seen_signatures = {}
-    new_arm_names = []
-
-    for new_idx, row in df.iterrows():
-        # Create signature without 'order', 'hostname', 'queue_time', 'trial_index', 'arm_name'
-        ignore_cols = {"order", "hostname", "queue_time", "trial_index", "arm_name"}
-        signature = tuple((col, row[col]) for col in df.columns if col not in ignore_cols)
-
-        if signature in seen_signatures:
-            # Collision → make a unique name
-            base_name = seen_signatures[signature]
-            suffix = 1
-            new_name = f"{base_name}_{suffix}"
-            while new_name in new_arm_names:
-                suffix += 1
-                new_name = f"{base_name}_{suffix}"
-            new_arm_names.append(new_name)
-        else:
-            # First occurrence → use new_idx as trial index in name
-            new_name = f"{new_idx}_0"
-            seen_signatures[signature] = f"{new_idx}_0"
-            new_arm_names.append(new_name)
-
-        df.at[new_idx, "trial_index"] = new_idx
-        df.at[new_idx, "arm_name"] = new_name
-
-    return df
-
 def save_results_csv() -> Optional[str]:
     if args.dryrun:
         return None
@@ -2214,8 +2173,8 @@ def save_results_csv() -> Optional[str]:
     save_checkpoint()
 
     try:
-        pd_frame = fetch_and_prepare_trials()
-        write_csv(pd_frame, pd_csv)
+        df = fetch_and_prepare_trials()
+        write_csv(df, pd_csv)
         write_json_snapshot(pd_json)
         save_experiment_to_file()
 
@@ -2237,10 +2196,21 @@ def get_results_paths() -> tuple[str, str]:
 def fetch_and_prepare_trials() -> pd.DataFrame:
     ax_client.experiment.fetch_data()
     df = ax_client.get_trials_data_frame()
+
+    #print("========================")
+    #print("BEFORE merge_with_job_infos:")
+    #print(df["generation_node"])
     df = merge_with_job_infos(df)
-    return reindex_trials(df)
+    #print("AFTER merge_with_job_infos:")
+    #print(df["generation_node"])
+
+    return df
 
 def write_csv(df, path: str) -> None:
+    try:
+        df = df.sort_values(by=["trial_index"], kind="stable").reset_index(drop=True)
+    except KeyError:
+        pass
     df.to_csv(path, index=False, float_format="%.30f")
 
 def write_json_snapshot(path: str) -> None:
@@ -5051,8 +5021,6 @@ def end_program(_force: Optional[bool] = False, exit_code: Optional[int] = None)
 
     abandon_all_jobs()
 
-    save_results_csv()
-
     if exit_code:
         _exit = exit_code
 
@@ -7590,8 +7558,6 @@ def execute_evaluation(_params: list) -> Optional[int]:
         trial_counter += 1
 
         progressbar_description("started new job")
-
-        save_results_csv()
     except submitit.core.utils.FailedJobError as error:
         handle_failed_job(error, trial_index, new_job)
         trial_counter += 1
@@ -7882,8 +7848,6 @@ def get_batched_arms(nr_of_jobs_to_get: int) -> list:
 
     print_debug(f"get_batched_arms: Finished with {len(batched_arms)} arm(s) after {attempts} attempt(s).")
 
-    save_results_csv()
-
     return batched_arms
 
 def fetch_next_trials(nr_of_jobs_to_get: int, recursion: bool = False) -> Tuple[Dict[int, Any], bool]:
@@ -7938,8 +7902,6 @@ def generate_trials(n: int, recursion: bool) -> Tuple[Dict[int, Any], bool]:
                     cnt += 1
                     trials_dict[trial_index] = arm.parameters
 
-        save_results_csv()
-
         return _finalize_generation(trials_dict, cnt, n, start_time)
 
     except Exception as e:
@@ -8820,10 +8782,6 @@ def execute_trials(
         index_param_list.append(_args)
         i += 1
 
-        save_results_csv()
-
-    save_results_csv()
-
     start_time = time.time()
 
     cnt = 0
@@ -9655,7 +9613,6 @@ def load_experiment_state() -> None:
     state_path = get_current_run_folder("experiment_state.json")
 
     if not os.path.exists(state_path):
-        print(f"State file {state_path} does not exist, starting fresh")
         return
 
     if args.worker_generator_path:

{omniopt2-8331 → omniopt2-8341}/PKG-INFO

@@ -1,6 +1,6 @@
 Metadata-Version: 2.4
 Name: omniopt2
-Version: 8331
+Version: 8341
 Summary: Automatic highly parallelized hyperparameter optimizer based on Ax/Botorch
 Home-page: https://scads.ai/transfer-2/verfuegbare-software-dienste-en/omniopt/
 Author: Norman Koch

{omniopt2-8331 → omniopt2-8341}/omniopt

@@ -81,7 +81,7 @@
 
 	function end_all_bg_processes {
 		for bg_job_id in $(jobs -p | sed -e 's#.*][[:space:]]*+[[:space:]]*##' -e 's#[[:space:]].*##'); do
-			kill $bg_job_id 2>/dev/null >/dev/null
+			kill "$bg_job_id" 2>/dev/null >/dev/null
 		done
 	}
 
@@ -110,8 +110,8 @@
 
 		if (command -v sbatch >/dev/null && [[ -n "$SLURM_JOB_ID" ]]) || ! command -v sbatch >/dev/null; then
 			already_logging_this_command=1
-			exec 1> >(tee -ia $bash_logname)
-			exec 2> >(tee -ia $bash_logname >& 2)
+			exec 1> >(tee -ia "$bash_logname")
+			exec 2> >(tee -ia "$bash_logname" >& 2)
 		fi
 	else
 		echo "uuidgen is not installed. It's recommended you install it." >&2
@@ -124,7 +124,7 @@
 		ram_children=0
 
 		for pid in $(pgrep -P $$); do
-			child_ram=$(grep VmRSS /proc/$pid/status 2>/dev/null | awk '{print $2 / 1024}')
+			child_ram=$(grep VmRSS "/proc/$pid/status" 2>/dev/null | awk '{print $2 / 1024}')
 			ram_children=$(awk -v a="$ram_children" -v b="$child_ram" 'BEGIN {print a + b}')
 		done
 
@@ -170,7 +170,7 @@
 		while true; do
 			date_str=$(date +"%Y-%m-%d %H:%M:%S")
 			echo -e "\n\n$date_str -> $(show_ram)\n\n" >&2
-			sleep $n
+			sleep "$n"
 		done
 	}
 
@@ -400,7 +400,7 @@
 			fi
 		fi
 
-		exit $CODE
+		exit "$CODE"
 	}
 
 	function get_anon_user_id {
@@ -1188,7 +1188,7 @@ EOF
 								myexit 211
 							}
 
-							bash omniopt $*
+							bash omniopt "$*"
 							exit_code=$?
 
 							myexit $exit_code
@@ -1882,8 +1882,8 @@ EOF
 						tail_log_file
 
 						if [[ $already_logging_this_command -eq 0 ]]; then
-							exec 1> >(tee -ia $bash_logname)
-							exec 2> >(tee -ia $bash_logname >& 2)
+							exec 1> >(tee -ia "$bash_logname")
+							exec 2> >(tee -ia "$bash_logname" >& 2)
 						fi
 
 
@@ -1894,13 +1894,15 @@ EOF
 						fi
 
 						exit_code_sed=$(echo "$exit_code_lines" | sed -e 's#Exit-Code:*[[:space:]]*##i' -e 's#,.*##')
-						if [ $? -ne 0 ] || [ -z "$exit_code_sed" ]; then
+						exit_code_sed_sed=$?
+						if [ $exit_code_sed_sed -ne 0 ] || [ -z "$exit_code_sed" ]; then
 							echo "WARN: sed failed or no data after sed."
 							exit_code_sed=""
 						fi
 
 						exit_code_tail=$(echo "$exit_code_sed" | tail -n1)
-						if [ $? -ne 0 ] || [ -z "$exit_code_tail" ]; then
+						exit_code_tail_tail=$?
+						if [ $exit_code_tail_tail -ne 0 ] || [ -z "$exit_code_tail" ]; then
 							echo "WARN: tail failed or no data after tail."
 							exit_code_tail=""
 						fi

{omniopt2-8331 → omniopt2-8341}/omniopt2.egg-info/PKG-INFO

@@ -1,6 +1,6 @@
 Metadata-Version: 2.4
 Name: omniopt2
-Version: 8331
+Version: 8341
 Summary: Automatic highly parallelized hyperparameter optimizer based on Ax/Botorch
 Home-page: https://scads.ai/transfer-2/verfuegbare-software-dienste-en/omniopt/
 Author: Norman Koch

{omniopt2-8331 → omniopt2-8341}/pyproject.toml

@@ -5,7 +5,7 @@ authors = [
   {email = "norman.koch@tu-dresden.de"},
   {name = "Norman Koch"}
 ]
-version = "8331"
+version = "8341"
 
 readme = "README.md"
 dynamic = ["dependencies"]