Fix issues with randomgrayscale layer

keras-team · Dec 14, 2024 · 4c05e0c · 4c05e0c
1 parent d96f910
commit 4c05e0c
Showing 1 changed file with 3 additions and 2 deletions.
diff --git a/keras/src/layers/preprocessing/image_preprocessing/random_grayscale.py b/keras/src/layers/preprocessing/image_preprocessing/random_grayscale.py
@@ -45,7 +45,7 @@ class RandomGrayscale(BaseImagePreprocessingLayer):
         will have the same value.
     """
 
-    def __init__(self, factor=0.5, data_format=None, **kwargs):
+    def __init__(self, factor=0.5, data_format=None, seed=None, **kwargs):
         super().__init__(**kwargs)
         if factor < 0 or factor > 1:
             raise ValueError(
@@ -54,7 +54,8 @@ def __init__(self, factor=0.5, data_format=None, **kwargs):
             )
         self.factor = factor
         self.data_format = backend.standardize_data_format(data_format)
-        self.generator = self.backend.random.SeedGenerator()
+        self.seed = seed
+        self.generator = self.backend.random.SeedGenerator(seed)
 
     def get_random_transformation(self, images, training=True, seed=None):
         if seed is None: