Reading a TFRecordfile from an Amazon S3 path

I am trying to read a TFRecord file directly from an Amazon S3 bucket using file path and tf.data.TFRecordDataset().

It seems tf.data.TFRecordDataset() only accepts filename in tf.string or tf.data.dataset format.

I tried following and it did not work.

filenames = ["s3://path_to_TFRecord"]
dataset = tf.data.TFRecordDataset(filenames)

I also tried using s3fs and a file handle to solve the issue but it fails.

Any solution is appreciated.