Merge pull request #7 from sradc/dev

move numgrads to smallpebble dir, add TODO

TODO

@@ -0,0 +1 @@
+- [ ] Update docstrings; use NumPy format

smallpebble/numerical_gradients.py

@@ -0,0 +1,52 @@
+# Copyright 2022 The SmallPebble Authors, Sidney Radcliffe
+#
+# Licensed under the Apache License, Version 2.0 (the "License");
+# you may not use this file except in compliance with the License.
+# You may obtain a copy of the License at
+#
+#     http://www.apache.org/licenses/LICENSE-2.0
+#
+# Unless required by applicable law or agreed to in writing, software
+# distributed under the License is distributed on an "AS IS" BASIS,
+# WITHOUT WARRANTIES OR CONDITIONS OF ANY KIND, either express or implied.
+# See the License for the specific language governing permissions and
+# limitations under the License.
+#
+"""Numerical gradients, used for debugging and tests.
+Computed using finite differences.
+"""
+from typing import Callable
+
+import smallpebble.array_library as np
+
+
+def numgrads(
+    func: Callable, args: list[np.ndarray], n: int = 1, delta: float = 1e-6
+) -> list[np.ndarray]:
+    "Numerical nth derivatives of func w.r.t. args."
+    gradients = []
+    for i, arg in enumerate(args):
+
+        def func_i(a):
+            new_args = [x for x in args]
+            new_args[i] = a
+            return func(*new_args)
+
+        gradfunc = lambda a: numgrad(func_i, a, delta)
+
+        for _ in range(1, n):
+            prev_gradfunc = gradfunc
+            gradfunc = lambda a: numgrad(prev_gradfunc, a, delta)
+
+        gradients.append(gradfunc(arg))
+    return gradients
+
+
+def numgrad(func: Callable, a: np.ndarray, delta: float = 1e-6) -> np.ndarray:
+    "Numerical gradient of func(a) at `a`."
+    grad = np.zeros(a.shape, a.dtype)
+    for index, _ in np.ndenumerate(grad):
+        delta_array = np.zeros(a.shape, a.dtype)
+        delta_array[index] = delta / 2
+        grad[index] = np.sum((func(a + delta_array) - func(a - delta_array)) / delta)
+    return grad

smallpebble/tests/test_smallpebble.py

@@ -15,10 +15,14 @@
 """Tests for SmallPebble.
 Check results, and derivatives against numerical derivatives.
 """
+from typing import Callable
+
 import pytest
-import smallpebble as sp
 import tensorflow as tf
 
+import smallpebble as sp
+from smallpebble.numerical_gradients import numgrads
+
 np = sp.np
 
 EPS = 1e-6
@@ -34,7 +38,13 @@ class NumericalError(Exception):
     pass
 
 
-def compare_results(args, sp_func, np_func, delta=1, eps=EPS):
+def compare_results(
+    args: list[np.ndarray],
+    sp_func: Callable,
+    np_func: Callable,
+    delta: int = 1,
+    eps: float = EPS,
+) -> None:
     """Compares:
     - SmallPebble function output against NumPy function output.
     - SmallPebble gradient against numerical gradient.
@@ -493,36 +503,6 @@ def test_sgd_step():
 # ---------------- UTIL
 
 
-def numgrads(func, args, n=1, delta=1e-6):
-    "Numerical nth derivatives of func w.r.t. args."
-    gradients = []
-    for i, arg in enumerate(args):
-
-        def func_i(a):
-            new_args = [x for x in args]
-            new_args[i] = a
-            return func(*new_args)
-
-        gradfunc = lambda a: numgrad(func_i, a, delta)
-
-        for _ in range(1, n):
-            prev_gradfunc = gradfunc
-            gradfunc = lambda a: numgrad(prev_gradfunc, a, delta)
-
-        gradients.append(gradfunc(arg))
-    return gradients
-
-
-def numgrad(func, a, delta=1e-6):
-    "Numerical gradient of func(a) at `a`."
-    grad = np.zeros(a.shape, a.dtype)
-    for index, _ in np.ndenumerate(grad):
-        delta_array = np.zeros(a.shape, a.dtype)
-        delta_array[index] = delta / 2
-        grad[index] = np.sum((func(a + delta_array) - func(a - delta_array)) / delta)
-    return grad
-
-
 def rmse(a: np.ndarray, b: np.ndarray):
     "Root mean square error."
     return np.sqrt(np.mean((a - b) ** 2))