keras-team
Model serializer path is incompatible with Windows
System information.
- OS Platform and Distribution: Windows Version 10.0.19044
- TensorFlow version: 2.7
- Python version: 3.9.12
Describe the problem.
Both the serializer and deserializer construct the temp_dir path using the hard coded prefix "ram://". This works for Unix-based systems, but the prefix is not valid on Windows.
Describe the current behavior.
Produces the following error, which stems from the fact that f (L75) is invalid because dest_path (L74) is not a valid memory address.
Traceback (most recent call last):
File "C:\Users\Hope\Anaconda3\envs\Association\lib\site-packages\keras\saving\pickle_utils.py", line 77, in serialize_model_as_bytecode
info.size = f.size()
File "C:\Users\Hope\Anaconda3\envs\Association\lib\site-packages\tensorflow\python\lib\io\file_io.py", line 99, in size
return stat(self.__name).length
File "C:\Users\Hope\Anaconda3\envs\Association\lib\site-packages\tensorflow\python\lib\io\file_io.py", line 910, in stat
return stat_v2(filename)
File "C:\Users\Hope\Anaconda3\envs\Association\lib\site-packages\tensorflow\python\lib\io\file_io.py", line 926, in stat_v2
return _pywrap_file_io.Stat(compat.path_to_str(path))
tensorflow.python.framework.errors_impl.NotFoundError
Suggested fix.
Implement something like the _get_temp_folder() method from scikeras
- the gist doesn't actually create a model and serialize/deserialize it, so you wouldn't get the error
- you need to run the serializer on Windows. "ram://" is not valid on Windows.
tagging @adriangb since he's worked on this file and scikeras
The lack of support for ram:// on windows was a known issue at the time we wrote that module.
To be quite honest, I've somewhat lost track of the work on that in the ~ 1 year since I did the implementation. I believe there were still some issues, e.g. https://github.com/tensorflow/tensorflow/issues/48086.
But at a high level, what is your goal? Do you need to pickle your models, or can you used the SavedModel API directly?
This method gets called by RLlib's internal serialization (for passing environments to worker processes), so unfortunately its usage is unavoidable
Gotcha. Did that just not work at all before this module was implemented in TF?
@adriangb is this issue purely a side effect of https://github.com/tensorflow/tensorflow/issues/48086 then? Meaning that when https://github.com/tensorflow/tensorflow/issues/48086 is fixed, this issue will automatically be fixed.
Some context: https://github.com/keras-team/keras/pull/14748#discussion_r654980437
I think we could ask to @mihaimaruseac It they have someone to comment in the upstream ticket. Probably we don't want to lost all the analsys we spent to debug that bug.
