gradboard 2.0.0__tar.gz → 2.1.0__tar.gz

{gradboard-2.0.0 → gradboard-2.1.0}/PKG-INFO

@@ -1,6 +1,6 @@
 Metadata-Version: 2.3
 Name: gradboard
-Version: 2.0.0
+Version: 2.1.0
 Summary: Easily snowboard down gnarly loss gradients
 License: MIT
 Author: Nicholas Bailey

{gradboard-2.0.0 → gradboard-2.1.0}/gradboard/scheduler.py

@@ -159,8 +159,23 @@ class PASS:
         """
         range_test_results = self._smoothed_range_test(self.range_test_results)
         minimum = min(range_test_results, key=lambda x: x[1])
-        self.max_lr = minimum[0] / 2
-        self.cool_point = self.max_lr * 0.1
+        max_left_of_min = (0, 0)
+        for r in range_test_results:
+            if r[0] > minimum[0]:
+                break
+            elif r[1] > max_left_of_min[1]:
+                max_left_of_min = r
+            else:
+                continue
+        difference = max_left_of_min[1] - minimum[1]
+        for r in range_test_results:
+            if r[1] > minimum[1] + 0.66 * difference:
+                self.cool_point = r[0]
+            elif r[1] > minimum[1] + 0.33 * difference:
+                self.max_lr = r[0]
+            else:
+                continue
+        self.cool_point = min(self.cool_point, self.max_lr / 4)
         print("High LR", self.max_lr)
         print("Cool point", self.cool_point)
 

{gradboard-2.0.0 → gradboard-2.1.0}/pyproject.toml

@@ -1,6 +1,6 @@
 [project]
 name = "gradboard"
-version = "2.0.0"
+version = "2.1.0"
 description = "Easily snowboard down gnarly loss gradients"
 authors = [
     {name = "Nicholas Bailey"}