Do not merge: min repro of CI failure after adding new tests.

jax_ai_stack/tests/test_nnx_with_optax2.py

@@ -0,0 +1,57 @@
+# Copyright 2024 The JAX Authors.
+#
+# 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
+#
+#     https://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.
+
+import unittest
+import jax
+import jax.numpy as jnp
+from flax import nnx
+import optax
+
+
+class SimpleModel(nnx.Module):
+
+  def __init__(self, rngs):
+    self.layer1 = nnx.Linear(2, 5, rngs=rngs)
+    self.layer2 = nnx.Linear(5, 3, rngs=rngs)
+
+  def __call__(self, x):
+    for layer in [self.layer1, self.layer2]:
+      x = layer(x)
+    return x
+
+
+class NNXOptaxTest(unittest.TestCase):
+
+  def test_nnx_optax(self):
+    key = jax.random.key(1701)
+    x = jax.random.normal(key, (1, 2))
+    y = jnp.ones((1, 3))
+
+    model = SimpleModel(nnx.Rngs(0))
+    optimizer = optax.adam(learning_rate=1e-3)
+    state = nnx.ModelAndOptimizer(model, optimizer)
+
+    def loss(model, x=x, y=y):
+      return jnp.mean((model(x) - y) ** 2)
+
+    initial_loss = loss(model)
+    grads = nnx.grad(loss)(state.model)
+    state.update(grads)
+    final_loss = loss(model)
+
+    self.assertNotAlmostEqual(initial_loss, final_loss)
+
+
+if __name__ == '__main__':
+  unittest.main()

jax_ai_stack/tests/test_nnx_with_optax3.py

@@ -0,0 +1,57 @@
+# Copyright 2024 The JAX Authors.
+#
+# 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
+#
+#     https://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.
+
+import unittest
+import jax
+import jax.numpy as jnp
+from flax import nnx
+import optax
+
+
+class SimpleModel(nnx.Module):
+
+  def __init__(self, rngs):
+    self.layer1 = nnx.Linear(2, 5, rngs=rngs)
+    self.layer2 = nnx.Linear(5, 3, rngs=rngs)
+
+  def __call__(self, x):
+    for layer in [self.layer1, self.layer2]:
+      x = layer(x)
+    return x
+
+
+class NNXOptaxTest(unittest.TestCase):
+
+  def test_nnx_optax(self):
+    key = jax.random.key(1701)
+    x = jax.random.normal(key, (1, 2))
+    y = jnp.ones((1, 3))
+
+    model = SimpleModel(nnx.Rngs(0))
+    optimizer = optax.adam(learning_rate=1e-3)
+    state = nnx.ModelAndOptimizer(model, optimizer)
+
+    def loss(model, x=x, y=y):
+      return jnp.mean((model(x) - y) ** 2)
+
+    initial_loss = loss(model)
+    grads = nnx.grad(loss)(state.model)
+    state.update(grads)
+    final_loss = loss(model)
+
+    self.assertNotAlmostEqual(initial_loss, final_loss)
+
+
+if __name__ == '__main__':
+  unittest.main()