chore: add guides on supporting custom models and objectives. (#153)

* chore: moved parametric function definition inside for modularity

* chore: added guide on how to support custom models and objectives.

README.md

@@ -45,6 +45,12 @@ optional arguments:
   --apply               Apply optimized configuration
 ```
 
+## Using custom models and objectives
+To explore the effects of other objectives and models, you need to modify the `src/objective/objective.py` and 
+`src/objective/parametric_function.py` files.
+If you need to change the sampling process, e.g. the initial samples, you need to modify the `src/sampling/sampler.py` 
+file. 
+
 ## Acknowledgments
 
 This work was supported by the Natural Sciences and Engineering Research Council of Canada (NSERC), Mitacs, and The University of British Columbia (UBC).

src/objective/objective.py

@@ -20,12 +20,12 @@ def __init__(
     def termination_value(self) -> float:
         """computes a value that indicates that we are confident"""
         knowledge_values = self.get_knowledge(self.memory_space)
-        y = self.param_function(self.memory_space)
+        y = self.param_function(self.memory_space) * self.memory_space
         return knowledge_values[np.argmin(y)]
 
     def get_values(self, memories: np.ndarray) -> np.ndarray:
         """Computes the objective values of the memories in input."""
-        real_cost = self.param_function(memories)
+        real_cost = self.param_function(memories) * memories
         knowledge = self.get_knowledge(memories)
         return real_cost * knowledge
 

src/objective/parametric_function.py

@@ -18,8 +18,8 @@ class ParametricFunction:
         bounds (tuple): Lower and upper bounds on parameters.
     """
 
-    function: callable
-    bounds: tuple
+    function: callable = lambda x, a0, a1, a2: a0 + a1 * np.exp(-x / a2)
+    bounds: tuple = ([-np.inf, -np.inf, -np.inf], [np.inf, np.inf, np.inf])
     params: any = None
 
     def __call__(self, x: int or np.ndarray):
@@ -41,14 +41,14 @@ def fit(self, sample: Sample) -> None:
         self.params = curve_fit(
             f=self.function,
             xdata=sample.memories,
-            ydata=sample.costs,
+            ydata=sample.durations,
             maxfev=int(1e8),
             p0=self.params,
             bounds=self.bounds,
         )[0]
 
     def minimize(
-        self, memory_space: np.ndarray, execution_time_threshold: int = None
+            self, memory_space: np.ndarray, execution_time_threshold: int = None
     ) -> int:
         """Minimizes the cost function and returns the corresponding memory configuration.
 
@@ -59,7 +59,7 @@ def minimize(
         Returns:
             int: Memory configuration that minimizes the cost function.
         """
-        costs = self.__call__(memory_space)
+        costs = self.__call__(memory_space) * memory_space
 
         # Handling execution threshold constraint
         if execution_time_threshold:
@@ -73,11 +73,11 @@ def minimize(
         min_index = np.argmin(costs)
         return memory_space[min_index]
 
+    @staticmethod
     def _filter_execution_time_constraint(
-        self,
-        memory_space: np.ndarray,
-        costs: np.ndarray,
-        execution_time_threshold: int = None,
+            memory_space: np.ndarray,
+            costs: np.ndarray,
+            execution_time_threshold: int = None,
     ) -> tuple:
         filtered_memories = np.array([])
         filtered_costs = np.array([])

src/parrotfish.py

@@ -36,10 +36,7 @@ def __init__(self, config: any):
                 credentials=credentials,
             )
 
-        self.param_function = ParametricFunction(
-            function=lambda x, a0, a1, a2: a0 * x + a1 * np.exp(-x / a2) * x,
-            bounds=([0, 0, 0], [np.inf, np.inf, np.inf]),
-        )
+        self.param_function = ParametricFunction()
 
         self.objective = Objective(
             param_function=self.param_function,

src/sampling/sampler.py

@@ -96,7 +96,7 @@ def update_sample(self, memory_mb: int) -> None:
         self.sample.update(subsample)
 
         logger.info(
-            f"Finished sampling {memory_mb} with billed duration results: {subsample_durations} is ms."
+            f"Finished sampling {memory_mb} with billed duration results: {subsample_durations} in ms."
         )
 
     def _explore_dynamically(self, durations: list) -> list:

tests/recommendation/test_parametric_function.py

@@ -8,10 +8,7 @@
 
 @pytest.fixture
 def param_function():
-    bounds = ([0, 0, 0], [np.inf, np.inf, np.inf])
-    return ParametricFunction(
-        function=lambda x, a, b, c: a * x + b * np.exp(-x / c) * x, bounds=bounds
-    )
+    return ParametricFunction()
 
 
 class TestFittingFunction: