no-op

PiperOrigin-RevId: 430987175

Author: paulgc17

tensorflow_data_validation/api/stats_api.py

@@ -104,8 +104,6 @@ def expand(
 def _sample_at_rate(example: pa.RecordBatch, sample_rate: float
                    ) -> Generator[pa.RecordBatch, None, None]:
   """Sample examples at input sampling rate."""
-  # TODO(pachristopher): Revisit this to decide if we need to fix a seed
-  # or add an optional seed argument.
   if random.random() <= sample_rate:
     yield example
 

tensorflow_data_validation/api/validation_api.py

@@ -129,7 +129,6 @@ def _may_be_set_legacy_flag(schema: schema_pb2.Schema):
     schema.generate_legacy_feature_spec = False
 
 
-# TODO(pachristopher): Add support for updating only a subset of features.
 def update_schema(schema: schema_pb2.Schema,
                   statistics: statistics_pb2.DatasetFeatureStatisticsList,
                   infer_feature_shape: Optional[bool] = True,

tensorflow_data_validation/statistics/generators/mutual_information.py

@@ -529,8 +529,6 @@ def compute(
           result[feature_path] = {self._custom_stats_key: 0.0}
       return stats_util.make_dataset_feature_stats_proto(result)
 
-    # TODO(pachristopher): Currently encoded examples operate on lists. Consider
-    # using ndarrays and vectorizing the operations.
     encoded_examples = _encode_examples(examples_record_batch,
                                         self._multivalent_features,
                                         self._categorical_features,

tensorflow_data_validation/statistics/stats_impl.py

@@ -83,8 +83,6 @@ def expand(
               slicing_util.GenerateSlicesSqlDoFn(
                   slice_sqls=self._options.experimental_slice_sqls)))
     else:
-      # TODO(pachristopher): Remove this special case if this doesn't give any
-      # performance improvement.
       dataset = (dataset
                  | 'KeyWithVoid' >> beam.Map(lambda v: (None, v)))
     _ = dataset | 'TrackDistinctSliceKeys' >> _TrackDistinctSliceKeys()  # pylint: disable=no-value-for-parameter
@@ -551,9 +549,6 @@ def __init__(
     else:
       self._desired_batch_size = constants.DEFAULT_DESIRED_INPUT_BATCH_SIZE
 
-    # TODO(pachristopher): Understand the cost of incrementing beam counters
-    # for every input batch. The other option is to update the counters during
-    # teardown.
     # Metrics
     self._combine_batch_size = beam.metrics.Metrics.distribution(
         constants.METRICS_NAMESPACE, 'combine_batch_size')

tensorflow_data_validation/utils/batch_util.py

@@ -28,7 +28,7 @@
 from tfx_bsl.coders import batch_util
 
 
-# TODO(pachristopher): Deprecate this.
+# TODO(b/221152546): Deprecate this.
 @beam.ptransform_fn
 def BatchExamplesToArrowRecordBatches(
     examples: beam.PCollection[types.Example],

tensorflow_data_validation/utils/stats_gen_lib.py

@@ -225,8 +225,7 @@ def generate_statistics_from_dataframe(
     merged_partial_stats = _generate_partial_statistics_from_df(
         dataframe, stats_options, stats_generators)
   else:
-    # TODO(pachristopher): Investigate why we don't observe linear speedup after
-    # a certain number of processes.
+    # TODO(b/144580609): Consider using Beam for inmemory mode as well.
     splits = np.array_split(dataframe, n_jobs)
     partial_stats = Parallel(n_jobs=n_jobs)(
         delayed(_generate_partial_statistics_from_df)(