LoRA wrapper delegation fixes for encoder_call and init_cache

gemma/gm/nn/_lora.py

@@ -62,14 +62,30 @@ def __post_init__(self):
   @nn.compact
   def __call__(self, *args, **kwargs):
     """Calls the model."""
+    with self._lora_interceptor():
+      return self.model(*args, **kwargs)
+
+  def _lora_interceptor(self):
+    """Returns the LoRA ModuleInterceptor context manager."""
     replace_module_fn = functools.partial(
         _replace_by_lora,
         rank=self.rank,
         dtype=self.dtype,
         verbose=self.verbose,
     )
-    with peft.ModuleInterceptor(replace_module_fn):
-      return self.model(*args, **kwargs)
+    return peft.ModuleInterceptor(replace_module_fn)
+
+  @nn.compact
+  def encoder_call(self, *args, **kwargs):
+    """Calls the model's encoder_call with LoRA adapters active."""
+    with self._lora_interceptor():
+      return self.model.encoder_call(*args, **kwargs)
+
+  @nn.compact
+  def init_cache(self, *args, **kwargs):
+    """Calls the model's init_cache with LoRA adapters active."""
+    with self._lora_interceptor():
+      return self.model.init_cache(*args, **kwargs)
 
   def __kontext_keys__(self) -> dict[str, str]:
     """Kauldron keys when calling `kontext.get_from_keys_obj`."""

gemma/gm/nn/_lora_test.py

@@ -0,0 +1,76 @@
+# Copyright 2026 DeepMind Technologies Limited.
+#
+# Licensed under the Apache License, Version 2.0 (the "License");
+# you may not use this file except in compliance with the License.
+# You may obtain a copy of the License at
+#
+#     http://www.apache.org/licenses/LICENSE-2.0
+#
+# Unless required by applicable law or agreed to in writing, software
+# distributed under the License is distributed on an "AS IS" BASIS,
+# WITHOUT WARRANTIES OR CONDITIONS OF ANY KIND, either express or implied.
+# See the License for the specific language governing permissions and
+# limitations under the License.
+
+"""Tests for LoRA wrapper delegation."""
+
+from absl.testing import absltest
+from flax import linen as nn
+from gemma.gm.nn import _lora
+import jax
+import jax.numpy as jnp
+
+
+class DummyModel(nn.Module):
+  some_attr: int = 42
+
+  @nn.compact
+  def __call__(self, x):
+    return nn.Dense(features=4, name='dense')(x)
+
+  @nn.compact
+  def encoder_call(self, x):
+    return nn.Dense(features=4, name='dense')(x)
+
+  @nn.compact
+  def init_cache(self, x):
+    return nn.Dense(features=4, name='dense')(x)
+
+
+class LoRATest(absltest.TestCase):
+
+  def test_call_intercepts_dense(self):
+    model = _lora.LoRA(rank=2, model=DummyModel())
+    params = model.init(jax.random.key(0), jnp.zeros((1, 4)))['params']
+    self.assertIn('dense', params)
+    self.assertIn('lora', params['dense'])
+    self.assertIn('a', params['dense']['lora'])
+    self.assertIn('b', params['dense']['lora'])
+
+  def test_encoder_call_intercepts_dense(self):
+    model = _lora.LoRA(rank=2, model=DummyModel())
+    params = model.init(
+        jax.random.key(0), jnp.zeros((1, 4)), method=model.encoder_call
+    )['params']
+    self.assertIn('dense', params)
+    self.assertIn('lora', params['dense'])
+    self.assertIn('a', params['dense']['lora'])
+    self.assertIn('b', params['dense']['lora'])
+
+  def test_init_cache_intercepts_dense(self):
+    model = _lora.LoRA(rank=2, model=DummyModel())
+    params = model.init(
+        jax.random.key(0), jnp.zeros((1, 4)), method=model.init_cache
+    )['params']
+    self.assertIn('dense', params)
+    self.assertIn('lora', params['dense'])
+    self.assertIn('a', params['dense']['lora'])
+    self.assertIn('b', params['dense']['lora'])
+
+  def test_getattr_forwarding(self):
+    model = _lora.LoRA(rank=2, model=DummyModel())
+    self.assertEqual(model.some_attr, 42)
+
+
+if __name__ == '__main__':
+  absltest.main()