diff --git a/torch2jax/__init__.py b/torch2jax/__init__.py
index 905a8ba..02ba559 100644
--- a/torch2jax/__init__.py
+++ b/torch2jax/__init__.py
@@ -136,11 +136,11 @@ def div_(self, other):
 coerce = lambda x: Torchish(x).value
 
 
-def implements(torch_function, JAXishify_output=True):
+def implements(torch_function, Torchishify_output=True):
   """Register a torch function override for Torchish"""
 
   def decorator(func):
-    func1 = (lambda *args, **kwargs: Torchish(func(*args, **kwargs))) if JAXishify_output else func
+    func1 = (lambda *args, **kwargs: Torchish(func(*args, **kwargs))) if Torchishify_output else func
     functools.update_wrapper(func1, torch_function)
     HANDLED_FUNCTIONS[torch_function] = func1
     return func1
@@ -571,7 +571,7 @@ def scaled_dot_product_attention(query, key, value, attn_mask=None, dropout_p=0.
 
 # NOTE: the "torch.Tensor" type annotations here are a lie, or at least an approximation: In reality, they can be
 # anything coerce-able.
-@implements(torch.nn.functional.multi_head_attention_forward, JAXishify_output=False)
+@implements(torch.nn.functional.multi_head_attention_forward, Torchishify_output=False)
 def multi_head_attention_forward(
   query: torch.Tensor,
   key: torch.Tensor,