🚑 Fixed param_grid as optional and checkpoints

pyproject.toml

@@ -4,7 +4,7 @@ build-backend = "setuptools.build_meta"
 
 [project]
 name = "qcml"
-version = "0.1.2"
+version = "0.1.2.1"
 description = "A benchmarking library for quantum and classical machine learning, with specialized support for evaluating kernel methods."
 readme = "README.md"
 authors = [

qcml/bench/__init__.py

@@ -13,6 +13,14 @@
 # limitations under the License.
 
 from .grid_search import GridSearch
+from .model_evaluator import ModelEvaluator
+from .parameter_grid import ParameterGrid
+from .checkpoint import (
+    save_checkpoint,
+    load_checkpoint,
+    delete_checkpoints,
+    get_highest_batch_checkpoint,
+)
 from .grids.kernel_grid import (
     classical_kernel_grid,
     classical_kernel_param_map,
@@ -26,7 +34,13 @@
 from .grids.transformation_grid import get_kernel_transform
 
 __all__ = [
-    "grid_search",
+    "GridSearch",
+    "ModelEvaluator",
+    "ParameterGrid",
+    "save_checkpoint",
+    "load_checkpoint",
+    "delete_checkpoints",
+    "get_highest_batch_checkpoint",
     "classical_kernel_grid",
     "classical_kernel_param_map",
     "quantum_kernel_grid",

qcml/bench/checkpoint.py

@@ -12,15 +12,12 @@
 # See the License for the specific language governing permissions and
 # limitations under the License.
 
-# File: qic/bench/checkpoint.py
-
 import os
 import json
 import logging
 
 logger = logging.getLogger(__name__)
 
-
 def save_checkpoint(
     results,
     classifier_name,
@@ -29,26 +26,37 @@ def save_checkpoint(
     dataset_name,
     results_path="checkpoints/",
 ):
+    # Handle classifier_name being either a string or a list
+    if isinstance(classifier_name, list):
+        classifier_name_str = "+".join([clf.__name__ for clf in classifier_name])
+    else:
+        classifier_name_str = classifier_name
+
     checkpoint_file_name = (
-        f"qcml-{experiment_name}-{classifier_name}-{dataset_name}-batch{batch_idx}.json"
+        f"qcml-{experiment_name}-{classifier_name_str}-{dataset_name}-batch{batch_idx}.json"
     )
     checkpoint_file_path = os.path.join(results_path, checkpoint_file_name)
 
-    if not os.path.exists(results_path):
-        os.makedirs(results_path)
+    # Ensure the results_path directory exists
+    os.makedirs(results_path, exist_ok=True)
 
     checkpoint_data = {"results": results, "batch_idx": batch_idx}
 
-    with open(checkpoint_file_path, "w") as f:
-        json.dump(checkpoint_data, f)
-
-    logger.info(f"Checkpoint saved to {checkpoint_file_path}")
-
+    try:
+        with open(checkpoint_file_path, "w") as f:
+            json.dump(checkpoint_data, f)
+        logger.info(f"Checkpoint saved to {checkpoint_file_path}")
+    except IOError as e:
+        logger.error(f"Failed to save checkpoint: {e}")
 
 def load_checkpoint(
     classifier_name, experiment_name, dataset_name, results_path="checkpoints/"
 ):
-    classifier_name_str = "+".join([clf.__name__ for clf in classifier_name])
+    if isinstance(classifier_name, list):
+        classifier_name_str = "+".join([clf.__name__ for clf in classifier_name])
+    else:
+        classifier_name_str = classifier_name
+
     checkpoint_file_name = (
         f"qcml-{experiment_name}-{classifier_name_str}-{dataset_name}.json"
     )
@@ -58,40 +66,56 @@ def load_checkpoint(
         logger.info("No checkpoint found for this classifier, dataset, and experiment.")
         return None
 
-    with open(checkpoint_file_path, "r") as f:
-        checkpoint_data = json.load(f)
-
-    logger.info(f"Checkpoint loaded from {checkpoint_file_path}")
-    return checkpoint_data
-
+    try:
+        with open(checkpoint_file_path, "r") as f:
+            checkpoint_data = json.load(f)
+        logger.info(f"Checkpoint loaded from {checkpoint_file_path}")
+        return checkpoint_data
+    except IOError as e:
+        logger.error(f"Failed to load checkpoint: {e}")
+        return None
 
 def delete_checkpoints(
     classifier_name, experiment_name, dataset_name, results_path="checkpoints/"
 ):
-    classifier_name_str = "+".join([clf.__name__ for clf in classifier_name])
+    if isinstance(classifier_name, list):
+        classifier_name_str = "+".join([clf.__name__ for clf in classifier_name])
+    else:
+        classifier_name_str = classifier_name
+
     checkpoint_file_name = (
         f"qcml-{experiment_name}-{classifier_name_str}-{dataset_name}.json"
     )
     checkpoint_file_path = os.path.join(results_path, checkpoint_file_name)
     if os.path.exists(checkpoint_file_path):
-        os.remove(checkpoint_file_path)
-        logger.info(
-            f"Checkpoint {checkpoint_file_path} deleted after successful completion of the experiment."
-        )
+        try:
+            os.remove(checkpoint_file_path)
+            logger.info(
+                f"Checkpoint {checkpoint_file_path} deleted after successful completion of the experiment."
+            )
+        except OSError as e:
+            logger.error(f"Failed to delete checkpoint: {e}")
     else:
         logger.info(
             f"No checkpoint found to delete for {classifier_name_str} on {dataset_name}."
         )
 
-
 def get_highest_batch_checkpoint(
     classifier_name, experiment_name, dataset_name, results_path="checkpoints/"
 ):
-    classifier_name_str = "+".join([clf.__name__ for clf in classifier_name])
+    if isinstance(classifier_name, list):
+        classifier_name_str = "+".join([clf.__name__ for clf in classifier_name])
+    else:
+        classifier_name_str = classifier_name
+
     checkpoint_file_pattern = (
         f"qcml-{experiment_name}-{classifier_name_str}-{dataset_name}-batch"
     )
 
+    if not os.path.exists(results_path):
+        logger.debug(f"No checkpoints directory found at {results_path}.")
+        return None, None
+
     # List all files in the checkpoints directory that match the pattern
     checkpoint_files = [
         f
@@ -120,11 +144,15 @@ def get_highest_batch_checkpoint(
 
     if latest_checkpoint_file:
         checkpoint_file_path = os.path.join(results_path, latest_checkpoint_file)
-        with open(checkpoint_file_path, "r") as f:
-            checkpoint_data = json.load(f)
-        logger.debug(
-            f"Checkpoint found: {latest_checkpoint_file} with batch {highest_batch}."
-        )
-        return checkpoint_data, highest_batch
-
-    return None, None
+        try:
+            with open(checkpoint_file_path, "r") as f:
+                checkpoint_data = json.load(f)
+            logger.debug(
+                f"Checkpoint found: {latest_checkpoint_file} with batch {highest_batch}."
+            )
+            return checkpoint_data, highest_batch
+        except IOError as e:
+            logger.error(f"Failed to load checkpoint: {e}")
+            return None, None
+
+    return None, None

qcml/bench/grid_search.py

@@ -31,7 +31,7 @@ class GridSearch:
     def __init__(
         self,
         classifiers: List[Callable],
-        param_grid: Dict[str, List[Any]],
+        param_grid: Optional[Dict[str, List[Any]]] = None,
         combinations: Optional[
             List[Tuple[Dict[str, Any], Callable, Dict[str, Any]]]
         ] = None,
@@ -66,13 +66,18 @@ def __init__(
         # Initialize ModelEvaluator
         self.evaluator = ModelEvaluator(use_jax=self.use_jax)
 
+        # Validate input parameters
+        if self.combinations is None and self.param_grid is None:
+            raise ValueError("Either 'param_grid' or 'combinations' must be provided.")
+
         # Initialize ParameterGrid if combinations are not provided
         if self.combinations is None:
             self.param_grid_obj = ParameterGrid(
                 self.param_grid, self.transformations, self.transformation_params
             )
             self.combinations = self.param_grid_obj.combinations
 
+
     def run(
         self,
         datasets: Optional[List[Dict[str, Any]]] = None,