fixing lint and test

algorithmic_efficiency/workloads/imagenet_vit/imagenet_jax/workload.py

@@ -1,5 +1,4 @@
 """ImageNet workload implemented in Jax."""
-import copy
 from typing import Dict, Optional, Tuple
 
 from flax import jax_utils

tests/workloads/imagenet_resnet/imagenet_jax/workload_test.py

@@ -28,22 +28,20 @@ def test_forward_pass(self):
     first_input_batch = jax.random.normal(data_rngs[0], shape=input_shape)
     expected_logits_shape = (jax.local_device_count(), batch_size, 1000)
 
-    # static_broadcasted_argnums=(3, 7) will recompile each time we call it in
-    # this file because we call it with a different combination of those two
+    # static_broadcasted_argnums=(3, 5) will recompile each time we call it in
+    # this function because we call it with a different combination of those two
     # args each time. Can't call with kwargs.
     pmapped_model_fn = jax.pmap(
         workload.model_fn,
         axis_name='batch',
-        in_axes=(0, 0, 0, None, None, None, None, None),
-        static_broadcasted_argnums=(3, 7))
+        in_axes=(0, 0, 0, None, None, None),
+        static_broadcasted_argnums=(3, 5))
     logits, updated_batch_stats = pmapped_model_fn(
         model_params,
         {'inputs': first_input_batch},
         batch_stats,
         spec.ForwardPassMode.TRAIN,
         rng,
-        None,
-        None,
         True)
     self.assertEqual(logits.shape, expected_logits_shape)
     # Test that batch stats are updated.
@@ -58,8 +56,6 @@ def test_forward_pass(self):
         updated_batch_stats,
         spec.ForwardPassMode.TRAIN,
         rng,
-        None,
-        None,
         False)
     self.assertEqual(
         _pytree_total_diff(same_batch_stats, updated_batch_stats), 0.0)
@@ -71,8 +67,6 @@ def test_forward_pass(self):
         batch_stats,
         spec.ForwardPassMode.EVAL,
         rng,
-        None,
-        None,
         False)
     self.assertEqual(logits.shape, expected_logits_shape)