diff --git a/cleverhans/tf2/attacks/fast_gradient_method.py b/cleverhans/tf2/attacks/fast_gradient_method.py
index ae3210f20..4f5279fb4 100644
--- a/cleverhans/tf2/attacks/fast_gradient_method.py
+++ b/cleverhans/tf2/attacks/fast_gradient_method.py
@@ -55,6 +55,9 @@ def fast_gradient_method(
     if clip_max is not None:
         asserts.append(tf.math.less_equal(x, clip_max))
 
+    # cast to tensor if provided as numpy array
+    x = tf.cast(x, tf.float32)
+
     if y is None:
         # Using model predictions as ground truth to avoid label leaking
         y = tf.argmax(model_fn(x), 1)