experiments/STT-tensorflow
1
0
Fork 0
Code Issues Pull Requests Projects Releases Wiki Activity

Issue:#46004

Browse Source 
Context:
Initial changes resulted in merge conflicts. As a result,I had to
come up with a different approach since _SHARDED_SUFFIX is a tensor now.

Solution:
Add os.path.normpath function to _temp/part instead

Additional:
Adding comments
This commit is contained in:
ather 2020-12-28 22:45:45 -05:00
parent a22a29347c
commit b69fa92a92
1 changed files with 3 additions and 2 deletions
Show Stats Download Patch File Download Diff File Expand all files Collapse all files

5
tensorflow/python/training/saver.py
View File

@ -228,7 +228,8 @@ class BaseSaverBuilder(object):
# Transformations:
# * Users pass in "save_path" in save() and restore(). Say "myckpt".
# * checkpoint_prefix gets fed <save_path><_SHARDED_SUFFIX>.
#
# * If checkpoint_prefix is a S3 bucket path ".part" is appended to it
# * Otherwise _temp/part is appended which is normalized relative to the OS
# Example:
# During runtime, a temporary directory is first created, which contains
# files
@ -254,7 +255,7 @@ class BaseSaverBuilder(object):
_SHARDED_SUFFIX = array_ops.where(
string_ops.regex_full_match(checkpoint_prefix, "^s3://.*"),
constant_op.constant(".part"),
constant_op.constant("_temp/part"))
constant_op.constant(os.path.normpath("_temp/part")))
tmp_checkpoint_prefix = string_ops.string_join(
[checkpoint_prefix, _SHARDED_SUFFIX])