fix vector_grad bugs (#68)

brainunit/autograd/_vector_grad.py

@@ -57,9 +57,11 @@ def grad_fun(*args, **kwargs):
             assert len(leaves) == 1, 'The function must return a single array when unit_aware is True.'
         tangents = jax.tree.unflatten(tree, [jnp.ones(l.shape, dtype=l.dtype) for l in leaves])
         grads = vjp_fn(tangents)
+        if isinstance(argnums, int):
+            grads = grads[0]
         if unit_aware:
             args_to_grad = jax.tree.map(lambda i: args[i], argnums)
-            r_unit = get_unit(jax.tree.leaves(y, is_leaf=lambda x: isinstance(x, Quantity))[0])
+            r_unit = get_unit(y)
             grads = jax.tree.map(
                 lambda arg, grad: maybe_decimal(
                     Quantity(get_mantissa(grad), unit=r_unit / get_unit(arg))
@@ -68,8 +70,6 @@ def grad_fun(*args, **kwargs):
                 grads,
                 is_leaf=lambda x: isinstance(x, Quantity)
             )
-        if isinstance(argnums, int):
-            grads = grads[0]
         if has_aux:
             return (grads, y, aux) if return_value else (grads, aux)
         else:

brainunit/autograd/_vector_grad_test.py

@@ -39,6 +39,21 @@ def simple_function(x):
             assert u.math.allclose(grad, jnp.array([6.0, 8.0]) * unit)
 
 
+def test_vector_grad_simple2():
+    def simple_function(x):
+        return x ** 3
+
+    x = jnp.array([3.0, 4.0])
+    for unit in [None, u.ms, u.mvolt]:
+        vector_grad_fn = u.autograd.vector_grad(simple_function)
+        if unit is None:
+            grad = vector_grad_fn(x)
+            assert jnp.allclose(grad, 3 * x ** 2)
+        else:
+            grad = vector_grad_fn(x * unit)
+            assert u.math.allclose(grad, 3 * (x * unit) ** 2)
+
+
 def test_vector_grad_multiple_args():
     def multi_arg_function(x, y):
         return x * y