Internal

PiperOrigin-RevId: 417830373

tensorflow_datasets/core/dataset_builder.py

@@ -591,7 +591,7 @@ def _build_single_dataset(
       shuffle_files,
       batch_size,
       decoders: Optional[TreeDict[decode.partial_decode.DecoderArg]],
-      read_config,
+      read_config: read_config_lib.ReadConfig,
       as_supervised,
   ):
     """as_dataset for a single split."""
@@ -810,7 +810,7 @@ def _as_dataset(
       self,
       split,
       decoders: Optional[TreeDict[decode.partial_decode.DecoderArg]] = None,
-      read_config=None,
+      read_config: Optional[read_config_lib.ReadConfig] = None,
       shuffle_files=False,
   ):
     """Constructs a `tf.data.Dataset`.

tensorflow_datasets/core/tfrecords_reader.py

@@ -178,7 +178,7 @@ def make_file_instructions(
 
   # TODO(epot): Should try to merge the instructions together as well as
   # performing additional validation. For example, should raise an error
-  # if there is overlapp between splits (`train[:50]+train[:25]`)
+  # if there is overlap between splits (`train[:50]+train[:25]`)
   # If there is a single shard, `train[:25]+train[50:75]` could be optimized
   # into a single `ds.take(25).skip(50-25).take(75-50)`
 
@@ -471,8 +471,9 @@ def read_files(
       raise ValueError(msg)
 
     # Prepend path to filename
+    path = self._path
     file_instructions = [
-        f.replace(filename=os.path.join(self._path, f.filename))
+        f.replace(filename=os.path.join(path, f.filename))
         for f in file_instructions
     ]
 

tensorflow_datasets/core/utils/shard_utils.py

@@ -21,21 +21,20 @@
  sharding needs.
 """
 
-from typing import Any, List, Sequence
-
 import dataclasses
+from typing import Any, List, Sequence
 
 
 @dataclasses.dataclass(eq=True, frozen=True)
 class FileInstruction(object):
   """Instruction to read a single shard/file.
 
   Attributes:
-    filename: The filenames contains the relative path, not absolute.
+    filename: The filename containing the relative path, not absolute.
     skip: Indicates which example read in the shard (`ds.skip().take()`). `0` if
-      no skipping
-    take: Indicates how many examples to read (`-1` to read all)
-    num_examples: `int`, The total number of examples
+      no skipping.
+    take: Indicates how many examples to read (`-1` to read all).
+    num_examples: `int`, The total number of examples.
   """
   filename: str
   skip: int