Checks whether a V1 or V2 checkpoint exists with the specified prefix.

This is the recommended way to check if a checkpoint exists, since it takes into account the naming difference between V1 and V2 formats.


  • checkpoint_prefix: the prefix of a V1 or V2 checkpoint, with V2 taking priority. Typically the result of or that of tf.train.latest_checkpoint(), regardless of sharded/non-sharded or V1/V2. Returns: A bool, true iff a checkpoint referred to by checkpoint_prefix exists.

Defined in tensorflow/python/training/