[linen] allows multiple compact methods

flax/linen/module.py

@@ -1037,7 +1037,6 @@ def __init_subclass__(cls, kw_only: bool = False, **kwargs: Any) -> None:
     cls._customized_dataclass_transform(kw_only)
     # We wrap user-defined methods including setup and __call__ to enforce
     # a number of different checks and to provide clear error messages.
-    cls._verify_single_or_no_compact()
     cls._find_compact_name_scope_methods()
     cls._wrap_module_attributes()
     # Set empty class defaults.
@@ -1116,20 +1115,6 @@ def _customized_dataclass_transform(cls, kw_only: bool):
 
     cls.__hash__ = _wrap_hash(cls.__hash__)  # type: ignore[method-assign]
 
-  @classmethod
-  def _verify_single_or_no_compact(cls):
-    """Statically verifies that at most a single method is labelled compact."""
-    methods = [m[0] for m in inspect.getmembers(cls, predicate=callable)]
-    n_compact_fns = len(
-      [
-        method_name
-        for method_name in methods
-        if hasattr(getattr(cls, method_name), 'compact')
-      ]
-    )
-    if n_compact_fns > 1:
-      raise errors.MultipleMethodsCompactError()
-
   @classmethod
   def _find_compact_name_scope_methods(cls):
     """Finds all compact_name_scope methods in the class."""

tests/linen/linen_module_test.py

@@ -513,18 +513,29 @@ def __call__(self):
 
     Foo({'a': ()}).apply({})
 
-  def test_only_one_compact_method(self):
-    msg = 'Only one method per class can be @compact'
-    with self.assertRaisesRegex(errors.MultipleMethodsCompactError, msg):
-
-      class MultipleCompactMethods(nn.Module):
-        @compact
-        def call1(self):
-          pass
-
-        @compact
-        def call2(self):
-          pass
+  def test_multiple_compact_methods(self):
+    """Test that multiple methods with the @compact decorator can be used.
+
+    NOTE: in the near future we might want to have compact methods reset the
+    autoname_cursor such that Dense would be reused in the second method.
+    """
+
+    class MultipleCompactMethods(nn.Module):
+      @compact
+      def __call__(self, x):
+        x = nn.Dense(1)(x)
+        return self.method(x)
+
+      @compact
+      def method(self, x):
+        x = nn.Dense(1)(x)
+        return x
+
+    m = MultipleCompactMethods()
+    variables = m.init(random.key(0), jnp.ones((1, 1)))
+    params = variables['params']
+    self.assertIn('Dense_0', params)
+    self.assertIn('Dense_1', params)
 
   def test_only_one_compact_method_subclass(self):
     class Dummy(nn.Module):

tests/linen/linen_transforms_test.py

@@ -2089,14 +2089,15 @@ def body_fn(mdl, c):
         rngs={'params': random.key(1), 'loop': random.key(2)},
     )
     self.assertEqual(vars['state']['acc'], x)
-    np.testing.assert_array_equal(
-        vars['state']['rng_params'][0], vars['state']['rng_params'][1]
+    self.assertTrue(
+      jnp.equal(vars['state']['rng_params'][0], vars['state']['rng_params'][1])
     )
     with jax_debug_key_reuse(False):
-      np.testing.assert_array_compare(
-          operator.__ne__,
+      self.assertFalse(
+        jnp.equal(
           vars['state']['rng_loop'][0],
           vars['state']['rng_loop'][1],
+        )
       )
 
   def test_cond(self):