[jax2tf] Fixed stale documentation about XLA metadata.

jax2tf does not yet support passing source location information
through to TF. The mechanism is partially implemented but disabled.
Here we remove misleading documentation that suggests the mechanism
is enabled.

jax/experimental/jax2tf/README.md

@@ -606,13 +606,6 @@ operations. There is support for `sharded_jit` and `pjit`.
 If you suspect that the SavedModel is larger than it should be, check first
 that you are not including the parameters as constants in the graph (see [above](#usage-saved-model)).
 
-Additionally, the SavedModel obtained from a `jax2tf.convert`-ed function may include source
-location information. This ensures that the debugging experience is similar
-for JAX with XLA vs. `jax2tf.convert` with XLA. However, this debugging information
-increases the size of the SavedModel, even possibly doubling it. You can
-disable the generation of this metadata with the parameter
-`include_xla_op_metadata`.
-
 ### SavedModel supports only first-order gradients
 
 The `jax2tf`-converted function supports higher-order gradients, but when the

jax/experimental/jax2tf/jax2tf.py

@@ -163,7 +163,8 @@ def __init__(self):
     self.shape_env: Sequence[Tuple[str, TfVal]] = ()
 
     # Whether to actually include XLA op metadata in the generated TF ops
-    self.include_xla_op_metadata = True
+    # TODO(b/189306134): implement support for XLA metadata
+    self.include_xla_op_metadata = False
 
     # A cache for the tf.convert_to_tensor for constants. We try to preserve
     # sharing for constants, to enable tf.Graph to take advantage of it.
@@ -415,6 +416,7 @@ def fix_in_ct(in_ct, arg_aval: core.ShapedArray):
       _thread_local_state.enable_xla = enable_xla
 
       prev_include_xla_op_metadata = _thread_local_state.include_xla_op_metadata
+      # TODO(b/189306134): implement support for XLA metadata
       _thread_local_state.include_xla_op_metadata = False
 
       _thread_local_state.shape_env = shape_env