ivoryos 1.3.6__py3-none-any.whl → 1.3.7__py3-none-any.whl

ivoryos/optimizer/nimo_optimizer.py

@@ -0,0 +1,162 @@
+### ivoryos/optimizers/nimo_optimizer.py
+import itertools
+import os
+
+
+from ivoryos.optimizer.base_optimizer import OptimizerBase
+
+
+class NIMOOptimizer(OptimizerBase):
+    def __init__(self, experiment_name:str, parameter_space: list, objective_config: list, optimizer_config: dict, datapath:str):
+        """
+        :param experiment_name: arbitrary name
+        :param parameter_space: list of parameter names
+        [
+            {"name": "param_1", "type": "range", "bounds": [1.0, 2.0], "value_type": "float"},
+            {"name": "param_2", "type": "choice", "bounds": ["a", "b", "c"], "value_type": "str"},
+            {"name": "param_3", "type": "range", "bounds": [0 10], "value_type": "int"},
+        ]
+        :param objective_config: objective configuration
+                [
+            {"name": "obj_1", "minimize": True, "weight": 1},
+            {"name": "obj_2", "minimize": False, "weight": 1}
+        ]
+        :param optimizer_config: optimizer configuration
+        optimizer_config={
+            "step_1": {"model": "Random", "num_samples": 10},
+            "step_2": {"model": "BOTorch"}
+        }
+        """
+        self.current_step = 0
+        self.experiment_name = experiment_name
+        self.parameter_space = parameter_space
+        self.objective_config = objective_config
+        self.optimizer_config = optimizer_config
+
+        super().__init__(experiment_name, parameter_space, objective_config, optimizer_config, datapath)
+
+        step_1 = optimizer_config.get("step_1", {})
+        step_2 = optimizer_config.get("step_2", {})
+        self.step_1_generator = step_1.get("model", "RE")
+        self.step_1_batch_num = step_1.get("num_samples", 1)
+        self.step_2_generator = step_2.get("model", "PDC")
+        self.candidates = os.path.join(self.datapath, f"{self.experiment_name}_candidates.csv")
+        self.proposals = os.path.join(self.datapath, f"{self.experiment_name}_proposals.csv")
+        self.n_objectives = len(self.objective_config)
+        self._create_candidates_csv()
+
+
+    def _create_candidates_csv(self):
+        # Extract parameter names and their possible values
+        import pandas as pd
+        import nimo
+        if os.path.exists(self.candidates) and nimo.history(self.candidates, self.n_objectives):
+            return
+        param_names = [p["name"] for p in self.parameter_space]
+
+        param_values = []
+        for p in self.parameter_space:
+            if p["type"] == "choice" and isinstance(p["bounds"], list):
+                param_values.append(p["bounds"])
+            elif p["type"] == "range" and len(p["bounds"]) == 2:
+                low, high = p["bounds"]
+                num_points = 10  # you can customize this granularity
+                step = (high - low) / (num_points - 1)
+                param_values.append([round(low + i * step, 4) for i in range(num_points)])
+            else:
+                raise ValueError(f"Unsupported parameter format: {p}")
+
+        # Generate all possible combinations
+        combos = list(itertools.product(*param_values))
+
+        # Create a DataFrame with parameter columns
+        df = pd.DataFrame(combos, columns=param_names)
+
+        # Add empty objective columns
+        for obj in self.objective_config:
+            df[obj["name"]] = ""
+
+        # Save to CSV
+        df.to_csv(self.candidates, index=False)
+
+
+    def suggest(self, n=1):
+        import pandas as pd
+        import nimo
+        method = self.step_1_generator if self.current_step <= self.step_1_batch_num else self.step_2_generator
+        nimo.selection(method = method,
+                       input_file = self.candidates,
+                       output_file = self.proposals,
+                       num_objectives = self.n_objectives,
+                       num_proposals = n)
+        self.current_step += 1
+        # Read proposals from CSV file
+        proposals_df = pd.read_csv(self.proposals)
+        # Get parameter names
+        param_names = [p["name"] for p in self.parameter_space]
+        # Convert proposals to list of parameter dictionaries
+        proposals = []
+        for _, row in proposals_df.iterrows():
+            proposal = {name: row[name] for name in param_names}
+            proposals.append(proposal)
+        return proposals[0] if n == 1 else proposals
+
+    def _convert_observation_to_list(self, obs: dict) -> list:
+        obj_names = [o["name"] for o in self.objective_config]
+        return [obs.get(name, None) for name in obj_names]
+
+    def observe(self, results: dict):
+        """
+        observe single output, nimo obj input is [1,2,3] or [[1, 2], [1, 2], [1, 2]] for MO
+        :param results: {"objective_name": "value"}
+        """
+        import nimo
+        nimo_objective_values = [self._convert_observation_to_list(results)]
+
+        nimo.output_update(input_file=self.proposals,
+                           output_file=self.candidates,
+                           num_objectives=self.n_objectives,
+                           objective_values=nimo_objective_values)
+
+    def append_existing_data(self, existing_data):
+        # TODO, history is part of the candidate file, we probably won't need this
+        pass
+
+
+    @staticmethod
+    def get_schema():
+        return {
+            "parameter_types": ["choice"],
+            "multiple_objectives": True,
+            "optimizer_config": {
+                "step_1": {"model": ["RE", "ES", "PDC"], "num_samples": 5},
+                "step_2": {"model": ["PHYSBO", "BLOX", "PTR", "SLESA", "BOMP", "COMBI"]}
+            },
+        }
+
+
+
+
+if __name__ == "__main__":
+    parameter_space = [
+        {"name": "silica", "type": "choice", "bounds": [100], "value_type": "float"},
+        {"name": "water", "type": "choice", "bounds": [900, 800, 750, 700, 650, 600, 550, 500], "value_type": "float"},
+        {"name": "PVA", "type": "choice", "bounds": [0, 0.005, 0.0075, 0.01, 0.05, 0.075, 0.1], "value_type": "float"},
+        {"name": "SDS", "type": "choice", "bounds": [0], "value_type": "float"},
+        {"name": "DTAB", "type": "choice", "bounds": [0, 0.005, 0.0075, 0.01, 0.05, 0.075, 0.1], "value_type": "float"},
+        {"name": "PVP", "type": "choice", "bounds": [0], "value_type": "float"},
+    ]
+    objective_config = [
+        {"name": "objective", "minimize": False, "weight": 1},
+
+    ]
+    optimizer_config = {
+        "step_1": {"model": "RE", "num_samples": 10},
+        "step_2": {"model": "PDC"}
+    }
+
+    nimo_optimizer = NIMOOptimizer(experiment_name="example_experiment", optimizer_config=optimizer_config, parameter_space=parameter_space, objective_config=objective_config)
+    nimo_optimizer.suggest(n=1)
+    nimo_optimizer.observe(
+        results={"objective": 1.0}
+    )

ivoryos/version.py

@@ -1 +1 @@
-__version__ = "1.3.6"
+__version__ = "1.3.7"

{ivoryos-1.3.6.dist-info → ivoryos-1.3.7.dist-info}/METADATA

@@ -1,6 +1,6 @@
 Metadata-Version: 2.4
 Name: ivoryos
-Version: 1.3.6
+Version: 1.3.7
 Summary: an open-source Python package enabling Self-Driving Labs (SDLs) interoperability
 Author-email: Ivory Zhang <ivoryzhang@chem.ubc.ca>
 License: MIT

{ivoryos-1.3.6.dist-info → ivoryos-1.3.7.dist-info}/RECORD

@@ -3,10 +3,11 @@ ivoryos/app.py,sha256=G6kzEOVzCduj7Fc2r1rMbMFHgDzZQV0lC20Oxps7RSM,4839
 ivoryos/config.py,sha256=y3RxNjiIola9tK7jg-mHM8EzLMwiLwOzoisXkDvj0gA,2174
 ivoryos/server.py,sha256=Idr41Boa8rPsh3bVJ8xy2Flwfxa1Xu6f2T0oMp4qBEk,7121
 ivoryos/socket_handlers.py,sha256=VWVWiIdm4jYAutwGu6R0t1nK5MuMyOCL0xAnFn06jWQ,1302
-ivoryos/version.py,sha256=5ZbAQtod5QalTI1C2N07edlxplzG_Q2XvGOSyOok4uA,22
+ivoryos/version.py,sha256=9peaXOar2qezOPJEKG6cD_A0aaXrzdVN8h-v6fBoBEk,22
 ivoryos/optimizer/ax_optimizer.py,sha256=k7RzhNMjh3MseOE-_FADVX72lYLDyXz-XE3juSBRWTk,7116
 ivoryos/optimizer/base_optimizer.py,sha256=gNBX9m3RxoYklwEBkqeDGZlU7c6e9e5kN0ZeTtnxnCE,2044
 ivoryos/optimizer/baybe_optimizer.py,sha256=EdrrRiYO-IOx610cPXiQhH4qG8knUP0uiZ0YoyaGIU8,7954
+ivoryos/optimizer/nimo_optimizer.py,sha256=J1nR1gMI4LqOkKqqdc5Pq7r3e3R8QQTJ-TZ9l9GVPQE,6696
 ivoryos/optimizer/registry.py,sha256=dLMo5cszcwa06hfBxdINQIGpkHtRe5-J3J7t76Jq6X0,306
 ivoryos/routes/__init__.py,sha256=47DEQpj8HBSa-_TImW-5JCeuQeRkm5NMpJWZG3hSuFU,0
 ivoryos/routes/api/api.py,sha256=97Y7pqTwOaWgZgI5ovEPxEBm6Asrt0Iy0VhBkVp2xqA,2304
@@ -96,8 +97,8 @@ ivoryos/utils/script_runner.py,sha256=gIYdTEFNWZFmPf9cf31y3wAUKUeJk_qnjohYDKevWN
 ivoryos/utils/serilize.py,sha256=lkBhkz8r2bLmz2_xOb0c4ptSSOqjIu6krj5YYK4Nvj8,6784
 ivoryos/utils/task_runner.py,sha256=xiMzK8gQ0mHsg0A1Ah8fmXe3azpaJh4hJiQJLHA11ZQ,3682
 ivoryos/utils/utils.py,sha256=dIc4BO55eS3lCA0nhbIncS5d7sLaKZy5hJS1I_Sm45o,14949
-ivoryos-1.3.6.dist-info/licenses/LICENSE,sha256=p2c8S8i-8YqMpZCJnadLz1-ofxnRMILzz6NCMIypRag,1084
-ivoryos-1.3.6.dist-info/METADATA,sha256=wUfp7SCz5Vtw7CrpuoKeirxfuf71F6deGzjBlx2Zuj8,8516
-ivoryos-1.3.6.dist-info/WHEEL,sha256=_zCd3N1l69ArxyTb8rzEoP9TpbYXkqRFSNOD5OuxnTs,91
-ivoryos-1.3.6.dist-info/top_level.txt,sha256=FRIWWdiEvRKqw-XfF_UK3XV0CrnNb6EmVbEgjaVazRM,8
-ivoryos-1.3.6.dist-info/RECORD,,
+ivoryos-1.3.7.dist-info/licenses/LICENSE,sha256=p2c8S8i-8YqMpZCJnadLz1-ofxnRMILzz6NCMIypRag,1084
+ivoryos-1.3.7.dist-info/METADATA,sha256=b4vaTs88sF290jix6Yiwlw0RbwIVUoNiDUg2KfjTWXM,8516
+ivoryos-1.3.7.dist-info/WHEEL,sha256=_zCd3N1l69ArxyTb8rzEoP9TpbYXkqRFSNOD5OuxnTs,91
+ivoryos-1.3.7.dist-info/top_level.txt,sha256=FRIWWdiEvRKqw-XfF_UK3XV0CrnNb6EmVbEgjaVazRM,8
+ivoryos-1.3.7.dist-info/RECORD,,