Updated DLRM experiment to work with V6 chip

Author: ajkv-google

recml/core/data/iterator.py

@@ -21,6 +21,7 @@
 from etils import epath
 import numpy as np
 import tensorflow as tf
+import jax
 
 
 Iterator = clu_data.DatasetIterator
@@ -67,17 +68,22 @@ def _maybe_to_numpy(
     ) -> np.ndarray | tf.SparseTensor | tf.RaggedTensor:
       if isinstance(x, (tf.SparseTensor, tf.RaggedTensor, np.ndarray)):
         return x
+      # FIX: Check for attribute existence to avoid crashes on non-Tensor objects
       if hasattr(x, "_numpy"):
         numpy = x._numpy()  # pylint: disable=protected-access
-      else:
+      elif hasattr(x, "numpy"):
         numpy = x.numpy()
+      else:
+        return x  # Return as-is if it can't be converted
+
       if isinstance(numpy, np.ndarray):
         # `numpy` shares the same underlying buffer as the `x` Tensor.
         # Tensors are expected to be immutable, so we disable writes.
         numpy.setflags(write=False)
       return numpy
 
-    return tf.nest.map_structure(_maybe_to_numpy, batch)
+    # FIX: Use jax.tree.map instead of tf.nest.map_structure
+    return jax.tree.map(_maybe_to_numpy, batch)
 
   @property
   def element_spec(self) -> clu_data.ElementSpec:
@@ -109,7 +115,8 @@ def _to_element_spec(
         )
       return clu_data.ArraySpec(dtype=x.dtype, shape=tuple(x.shape))
 
-    element_spec = tf.nest.map_structure(_to_element_spec, batch)
+    # element_spec = tf.nest.map_structure(_to_element_spec, batch)
+    element_spec = jax.tree.map(_to_element_spec, batch)
     self._element_spec = element_spec
     return element_spec
 

recml/core/ops/embedding_ops.py

@@ -38,7 +38,7 @@ class SparsecoreParams:
   """Embedding parameters."""
 
   feature_specs: Nested[FeatureSpec]
-  mesh: jax.sharding.Mesh | jax.sharding.AbstractMesh
+  mesh: jax.sharding.Mesh 
   data_axes: Sequence[str | None]
   embedding_axes: Sequence[str | None]
   sharding_strategy: str

recml/core/training/mesh_context.py

@@ -0,0 +1,10 @@
+"""Simple global context for mesh, replacing jax.experimental.maps."""
+
+_GLOBAL_MESH = None
+
+def set_global_mesh(mesh):
+  global _GLOBAL_MESH
+  _GLOBAL_MESH = mesh
+
+def get_global_mesh():
+  return _GLOBAL_MESH

recml/core/training/partitioning.py

@@ -11,6 +11,7 @@
 # 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.
+
 """Utilities for partitioning."""
 
 import abc
@@ -21,6 +22,8 @@
 import flax.linen as nn
 import jax
 import numpy as np
+# FIXED: Use the public experimental module available in JAX 0.4.30
+from recml.core.training import mesh_context
 
 
 PyTree = Any
@@ -67,7 +70,8 @@ class DataParallelPartitioner(Partitioner):
   """Data parallel partitioner."""
 
   def __init__(self, data_axis: str = "batch"):
-    self.mesh = jax.make_mesh((jax.device_count(),), (data_axis,))
+    devices = jax.devices()
+    self.mesh = jax.sharding.Mesh(devices, (data_axis,))
     self.data_sharding = jax.sharding.NamedSharding(
         self.mesh, jax.sharding.PartitionSpec(data_axis)
     )
@@ -107,8 +111,10 @@ def _shard(x: np.ndarray) -> jax.Array:
   def partition_init(
       self, init_fn: CreateStateFn, *, abstract_batch: PyTree | None = None
   ) -> CreateStateFn:
-    with jax.sharding.use_mesh(self.mesh):
+    # FIXED: Use 'with self.mesh'
+    with self.mesh:
       if abstract_batch is not None:
+        mesh_context.set_global_mesh(self.mesh)
         abstract_state = jax.eval_shape(init_fn, abstract_batch)
         specs = nn.get_partition_spec(abstract_state)
         self.state_sharding = jax.tree.map(
@@ -117,7 +123,8 @@ def partition_init(
       init_fn = jax.jit(init_fn, out_shardings=self.state_sharding)
 
     def _wrapped_init(batch: PyTree) -> State:
-      with jax.sharding.use_mesh(self.mesh):
+      # FIXED: Use 'with self.mesh'
+      with self.mesh:
         state = init_fn(batch)
         state = _maybe_unbox_state(state)
         return state
@@ -130,15 +137,18 @@ def partition_step(self, fn: StepFn, *, training: bool = False) -> StepFn:
       jit_kws["out_shardings"] = (self.state_sharding, None)
       jit_kws["donate_argnums"] = (1,)
 
-    with jax.sharding.use_mesh(self.mesh):
+    # FIXED: Use 'with self.mesh' and legacy bridge
+    with self.mesh:
+      mesh_context.set_global_mesh(self.mesh)
       step_fn = jax.jit(
           fn,
           in_shardings=(self.data_sharding, self.state_sharding),
           **jit_kws,
       )
 
     def _wrapped_step(batch: PyTree, state: State) -> Any:
-      with jax.sharding.use_mesh(self.mesh):
+      # FIXED: Use 'with self.mesh'
+      with self.mesh:
         return step_fn(batch, state)
 
     return _wrapped_step
@@ -190,7 +200,8 @@ def __init__(
     if axis_sizes[0] == -1:
       axis_sizes[0] = len(devices) // math.prod(axis_sizes[1:])
 
-    self.mesh = jax.make_mesh(axis_sizes, axis_names, devices=devices)
+    # self.mesh = jax.make_mesh(axis_sizes, axis_names, devices=devices)
+    self.mesh = jax.sharding.Mesh(devices, axis_names)
     self.rules = rules
     self.aot_compile = aot_compile
     self.options = options
@@ -213,12 +224,6 @@ def __init__(
     self.abstract_batch = None
     self.abstract_state = None
 
-  @property
-  def mesh_context_manager(
-      self,
-  ) -> Callable[[jax.sharding.Mesh], ContextManager[None]]:
-    return jax.sharding.use_mesh
-
   def shard_inputs(self, inputs: PyTree) -> PyTree:
     def _shard(x: np.ndarray) -> jax.Array:
       return jax.make_array_from_process_local_data(self.data_sharding, x)
@@ -234,7 +239,10 @@ def partition_init(
           " model parallel partitioner."
       )
 
-    with self.mesh_context_manager(self.mesh):
+    # FIXED: Use 'with self.mesh' directly
+    with self.mesh:
+      # FIXED: Legacy bridge
+      mesh_context.set_global_mesh(self.mesh)
       abstract_state = jax.eval_shape(init_fn, abstract_batch)
       specs = nn.get_partition_spec(abstract_state)
 
@@ -247,7 +255,8 @@ def partition_init(
       compiled_init_fn = jax.jit(init_fn, out_shardings=state_sharding)
 
     def _init(batch: PyTree) -> State:
-      with self.mesh_context_manager(self.mesh):
+      # FIXED: Use 'with self.mesh' directly
+      with self.mesh:
         state = compiled_init_fn(batch)
         state = _maybe_unbox_state(state)
       return state
@@ -265,7 +274,9 @@ def partition_step(self, fn: StepFn, *, training: bool = False) -> StepFn:
     else:
       jit_kws["out_shardings"] = None
 
-    with self.mesh_context_manager(self.mesh):
+    # FIXED: Use 'with self.mesh' directly and legacy bridge
+    with self.mesh:
+      mesh_context.set_global_mesh(self.mesh)
       step_fn = jax.jit(
           fn,
           in_shardings=(self.data_sharding, self.state_sharding),
@@ -286,7 +297,8 @@ def partition_step(self, fn: StepFn, *, training: bool = False) -> StepFn:
       )
 
     def _step(batch: PyTree, state: State) -> Any:
-      with self.mesh_context_manager(self.mesh):
+      # FIXED: Use 'with self.mesh' directly
+      with self.mesh:
         return step_fn(batch, state)
 
     return _step
@@ -302,4 +314,4 @@ def _maybe_unbox(x: Any) -> Any:
       _maybe_unbox,
       x,
       is_leaf=lambda k: isinstance(k, nn.Partitioned),
-  )
+  )

recml/examples/dlrm_experiment_test.py

@@ -13,6 +13,12 @@
 # limitations under the License.
 """Tests for the DLRM experiment."""
 
+import sys
+import os
+# Add the RecML folder to the system path
+sys.path.append(os.path.join(os.getcwd(), "../../../RecML"))
+os.environ["KERAS_BACKEND"] = "jax"
+
 from absl.testing import absltest
 import fiddle as fdl
 from fiddle import selectors
@@ -32,8 +38,8 @@ def test_dlrm_experiment(self):
 
     experiment = dlrm_experiment.experiment()
 
-    experiment.task.train_data.global_batch_size = 4
-    experiment.task.eval_data.global_batch_size = 4
+    experiment.task.train_data.global_batch_size = 128
+    experiment.task.eval_data.global_batch_size = 128
     experiment.trainer.train_steps = 12
     experiment.trainer.steps_per_loop = 4
     experiment.trainer.steps_per_eval = 4

recml/layers/linen/sparsecore.py

@@ -28,10 +28,12 @@
 from recml.core.ops import embedding_ops
 import tensorflow as tf
 
+from recml.core.training import mesh_context
+
 
 with epy.lazy_imports():
   # pylint: disable=g-import-not-at-top
-  from jax_tpu_embedding.sparsecore.lib.flax import embed
+  from jax_tpu_embedding.sparsecore.lib.flax.linen import embed
   from jax_tpu_embedding.sparsecore.lib.nn import embedding
   from jax_tpu_embedding.sparsecore.lib.nn import embedding_spec
   from jax_tpu_embedding.sparsecore.lib.nn import table_stacking
@@ -369,16 +371,28 @@ class SparsecoreEmbed(nn.Module):
   sparsecore_config: SparsecoreConfig
   mesh: jax.sharding.Mesh | jax.sharding.AbstractMesh | None = None
 
-  def get_mesh(self) -> jax.sharding.Mesh | jax.sharding.AbstractMesh:
-    if self.mesh is not None:
-      return self.mesh
-    abstract_mesh = jax.sharding.get_abstract_mesh()
-    if not abstract_mesh.shape_tuple:
+  # def get_mesh(self) -> jax.sharding.Mesh | jax.sharding.AbstractMesh:
+  #   if self.mesh is not None:
+  #     return self.mesh
+  #   abstract_mesh = jax.sharding.get_abstract_mesh()
+  #   if not abstract_mesh.shape_tuple:
+  #     raise ValueError(
+  #         'No abstract mesh shape was set with `jax.sharding.use_mesh`. Make'
+  #         ' sure to set the mesh when calling the sparsecore module.'
+  #     )
+  #   return abstract_mesh
+  
+  def get_mesh(self) -> jax.sharding.Mesh:
+    # Try to get the mesh from our custom global context
+    mesh = mesh_context.get_global_mesh()
+
+    if mesh is None:
       raise ValueError(
-          'No abstract mesh shape was set with `jax.sharding.use_mesh`. Make'
-          ' sure to set the mesh when calling the sparsecore module.'
+          "No global mesh found. Make sure to call "
+          "`partitioning.partition_init` (which sets the mesh) "
+          "before initializing SparseCore."
       )
-    return abstract_mesh
+    return mesh
 
   def get_sharding_axis(
       self, mesh: jax.sharding.Mesh | jax.sharding.AbstractMesh

requirements.txt

@@ -63,7 +63,7 @@ platformdirs==4.3.7
 pluggy==1.5.0
 pre-commit==4.2.0
 promise==2.3
-protobuf==5.29.4
+# protobuf==5.29.4
 psutil==7.0.0
 pyarrow==19.0.1
 pygments==2.19.1