diff --git a/python/tvm/relay/frontend/keras.py b/python/tvm/relay/frontend/keras.py
index 1913d4a2681a..ec960b9f0b12 100644
--- a/python/tvm/relay/frontend/keras.py
+++ b/python/tvm/relay/frontend/keras.py
@@ -254,6 +254,8 @@ def _convert_dense(
     weightList = keras_layer.get_weights()
     weight = etab.new_const(weightList[0].transpose([1, 0]))
     params = {"weight": weight, "units": weightList[0].shape[1]}
+    units = list(weightList[0].shape)[1]
+    assert units > 0, "The value of units must be a positive integer"
     if input_shape is None:
         input_shape = keras_layer.input_shape
     input_dim = len(input_shape)
@@ -1008,6 +1010,7 @@ def _convert_lstm(
     if keras_layer.go_backwards:
         in_data = _op.reverse(in_data, axis=1)
     units = list(weightList[0].shape)[1]
+    assert units > 0, "The value of units must be a positive integer"
     time_steps = in_shape[1]
     in_data = _op.squeeze(in_data, axis=[0])
     in_data = _op.split(in_data, indices_or_sections=time_steps, axis=0)
@@ -1051,6 +1054,7 @@ def _convert_simple_rnn(
     if keras_layer.use_bias:
         in_bias = etab.new_const(weightList[2])
     units = list(weightList[0].shape)[1]
+    assert units > 0, "The value of units must be a positive integer"
     in_data = _op.nn.batch_flatten(in_data)
     ixh = _op.nn.dense(in_data, kernel_weight, units=units)
     if keras_layer.use_bias:
@@ -1080,6 +1084,7 @@ def _convert_gru(
     if keras_layer.use_bias:
         in_bias = etab.new_const(weightList[2])
     units = list(weightList[0].shape)[1]
+    assert units > 0, "The value of units must be a positive integer"
     in_data = _op.nn.batch_flatten(in_data)
     matrix_x = _op.nn.dense(in_data, kernel_weight, units=units)
     if keras_layer.use_bias:
diff --git a/tests/python/frontend/keras/test_forward.py b/tests/python/frontend/keras/test_forward.py
index 50a0e9850559..2f6e1098df78 100644
--- a/tests/python/frontend/keras/test_forward.py
+++ b/tests/python/frontend/keras/test_forward.py
@@ -244,7 +244,7 @@ def test_forward_activations_except(self, keras_mod):
         ):
             act_funcs = [
                 keras_mod.layers.LeakyReLU(alpha=None),
-                keras_mod.layers.LEU(2, 3, 4),
+                keras_mod.layers.ELU(2, 3, 4),
                 keras_mod.layers.ReLU(threshold=None),
             ]
             data = keras_mod.layers.Input(shape=(2, 3, 4))