dumped best params into serialisable and fixed optuna to version 2.3.0 (

#225)

data/optuna_fitting/03-optuna_deltag_fitting_best_params.json

@@ -0,0 +1,13 @@
+{
+    "gd": -0.9076664181509553,
+    "ga": -16.13143317348325,
+    "g0": 1.8842791231644282,
+    "gs": 0.046698073586512054,
+    "gr": -3.6429426011852524,
+    "gpi1": -1.6022604035599135,
+    "gpi2": -1.1740853831959548,
+    "gi": 4.999560447432514,
+    "F": 0.5143359987215719,
+    "expd": 0.5041243951898653,
+    "expa": 0.3436880638589098
+}

notebooks/gerber_deltag_validation.ipynb

@@ -9,6 +9,7 @@
     "import os\n",
     "import sys\n",
     "import bz2\n",
+    "import json\n",
     "import mpld3\n",
     "import pandas as pd\n",
     "import _pickle as cPickle\n",
@@ -72,11 +73,10 @@
     "input_data = bz2.BZ2File(data_filepath, 'rb')\n",
     "input_data = cPickle.load(input_data)\n",
     "\n",
-    "# Open the study\n",
-    "study_filepath = os.path.join(optuna_path, config.OPTUNA_DELTAG_STUDY_FILENAME)\n",
-    "study = bz2.BZ2File(study_filepath, 'rb')\n",
-    "study = cPickle.load(study)\n",
-    "best_params = study.best_trial.params\n",
+    "# Open the parameters\n",
+    "params_filepath = os.path.join(optuna_path, config.OPTUNA_DELTAG_PARAMETERS_FILENAME)\n",
+    "with open(params_filepath) as f:\n",
+    "    best_params = json.load(f)\n",
     "best_params"
    ]
   },

notebooks/guthrie_deltag_validation.ipynb

@@ -9,6 +9,7 @@
     "import os\n",
     "import sys\n",
     "import bz2\n",
+    "import json\n",
     "import mpld3\n",
     "import pandas as pd\n",
     "import _pickle as cPickle\n",
@@ -64,14 +65,12 @@
     "sys.path.insert(0, opt_path)\n",
     "import config\n",
     "\n",
-    "# Open the study\n",
+    "# Open the parameters\n",
     "data_path = os.path.abspath(os.path.join(os.getcwd(), '..', config.DATA_PATH))\n",
     "optuna_path = os.path.join(data_path, config.OPTUNA_DIRNAME)\n",
-    "study_filepath = os.path.join(optuna_path, config.OPTUNA_DELTAG_STUDY_FILENAME)\n",
-    "study = bz2.BZ2File(study_filepath, 'rb')\n",
-    "study = cPickle.load(study)\n",
-    "\n",
-    "best_params = study.best_trial.params\n",
+    "params_filepath = os.path.join(optuna_path, config.OPTUNA_DELTAG_PARAMETERS_FILENAME)\n",
+    "with open(params_filepath) as f:\n",
+    "    best_params = json.load(f)\n",
     "best_params"
    ]
   },

optimisation/02-deltag_fitting.py

@@ -1,6 +1,7 @@
 """Parameter fitting of Jazzy using Optuna and sklearn."""
 # optimisation/02-deltag_fitting.py
 import config
+from helpers import dump_parameters_to_json
 from helpers import load_data_configuration
 from helpers import run_optimisation
 from sklearn.metrics import mean_absolute_error
@@ -106,10 +107,13 @@ def deltag_objective(trial):
     return loss
 
 
-# load input/ouput configuration
-input_data, study_filepath = load_data_configuration(
-    config.OPTUNA_DELTAG_STUDY_FILENAME
+# load input/output configuration
+input_data, study_filepath, params_filepath = load_data_configuration(
+    config.OPTUNA_DELTAG_STUDY_FILENAME, config.OPTUNA_DELTAG_PARAMETERS_FILENAME
 )
 
 # run optimisation
 study = run_optimisation(deltag_objective, study_filepath, verbose=config.VERBOSE)
+
+# dump best parameters into json
+dump_parameters_to_json(study_filepath, params_filepath)

optimisation/config.py

@@ -9,6 +9,7 @@
 ORIGINAL_DATA_FILENAME = "03-free_energies_gerber_curated.txt"
 PRECALCULATED_DATA_FILENAME = "01-precalculated_mol_data.pbz2"
 OPTUNA_DELTAG_STUDY_FILENAME = "02-optuna_deltag_fitting_study.pbz2"
+OPTUNA_DELTAG_PARAMETERS_FILENAME = "03-optuna_deltag_fitting_best_params.json"
 
 # logic configuration
 MINIMISATION_METHOD = "MMFF94"

optimisation/helpers.py

@@ -2,18 +2,20 @@
 # optimisation/helpers.py
 import _pickle as cpickle
 import bz2
+import json
 import os
 
 import config
 import optuna
 from optuna.samplers import TPESampler
 
 
-def load_data_configuration(study_filename):
+def load_data_configuration(study_filename, params_filename):
     """Abstraction for input and output loading and configuration setting.
 
-    The method only accepts the name of the output study file, whilst
-    the rest of the parameters must be defined in a config file.
+    The method only accepts the name of the output study and best parameters
+    files, whilst the rest of the parameters must be defined in a config file.
+
     """
     # load the input
     data_path = os.path.abspath(os.path.join(os.getcwd(), "..", config.DATA_PATH))
@@ -24,7 +26,8 @@ def load_data_configuration(study_filename):
 
     # configure output
     study_filepath = os.path.join(optuna_path, study_filename)
-    return input_data, study_filepath
+    params_filepath = os.path.join(optuna_path, params_filename)
+    return input_data, study_filepath, params_filepath
 
 
 def run_optimisation(objective, study_filepath, verbose=False):
@@ -80,3 +83,20 @@ def early_stopping_opt(study, trial):
     # write the results out
     with bz2.BZ2File(study_filepath, "w") as f:
         cpickle.dump(study, f)
+
+
+def dump_parameters_to_json(study_filepath, params_filepath):
+    """Helper for dumping the study best parameters.
+
+    Accepts the path to a study file, unpickles it,
+    and dumps the best parameters into a serialisable.
+
+    """
+    # read the parameters
+    with bz2.BZ2File(study_filepath, "rb") as f:
+        study = cpickle.load(f)
+    best_params = study.best_trial.params
+
+    # dump the parameters
+    with open(params_filepath, "w") as f:
+        json.dump(best_params, f, indent=4)