change random seed to match the splits of `facebookresearch/vissl`

Caltech-101.py

@@ -259,7 +259,7 @@ class Caltech101(datasets.GeneratorBasedBuilder):
 
         is_train_split = split == "train"
 
-        rng = np.random.default_rng(1234)
+        rng = np.random.default_rng(0)
 
         for class_dir in img_dir.iterdir():
             class_name = class_dir.name