diff --git a/tests/jax/test_fused_attn.py b/tests/jax/test_fused_attn.py index 1dd92fe181..a657431fdc 100644 --- a/tests/jax/test_fused_attn.py +++ b/tests/jax/test_fused_attn.py @@ -957,7 +957,7 @@ def test_forward(self): customcall_args = [ # Put test data onto each GPU for distributed. - # TODO(mgoldfarb-nvidia): We will need to add reordering for bias, mas and + # TODO(mgoldfarb-nvidia): We will need to add reordering for bias, mask and # THD params once we support those features on CP. jax.device_put(self.cp_reorder_fn(self.q), self.qkvo_sharding), jax.device_put(self.cp_reorder_fn(self.k), self.qkvo_sharding), @@ -1089,7 +1089,7 @@ def grad_func( self.dropout_rng, ] customcall_args = [ - # TODO(mgoldfarb-nvidia): We will need to add reordering for bias, mas and + # TODO(mgoldfarb-nvidia): We will need to add reordering for bias, mask and # THD params once we support those features on CP. jax.device_put(self.cp_reorder_fn(self.q), self.qkvo_sharding), jax.device_put(self.cp_reorder_fn(self.k), self.qkvo_sharding),