DL Python Network Creator - File path error even after enabling Win32 ong paths

Hi there, I am trying to create DL workflows with the different options (DL4J, Keras and DL Python).

Can anyone help with the file generation error in the DL Python Network Creator I keep getting? Thanks!

KNIME 4.3.2 with all extensions up to date.

I hoped setting the ‘Enable Win32 long paths’ to enabled would fix it (as per Installation Guide Known Issues), but after setting that and system restart I get the same error:

Execute failed: Failed to create a NewWriteableFile: C:\Users\codrin_kruijne\AppData\Local\Temp\knime_Deep Learning P18093\fs-DL_Py_0-95-18095\000\000\4b4c3522-0e8c-4285-8248-944d6799c42c\variables\variables_temp_d312bb449a8241afa3b87eb249cd68af/part-00000-of-00001.data-00000-of-00001.tempstate647386528939771127 : The system cannot find the path specified.
; No such process [Op:SaveV2]
Traceback (most recent call last):
File “”, line 3, in
File “C:\Program Files\KNIME 4.2.3\plugins\org.knime.dl.tensorflow2_4.3.1.v202101261634\py\TF2Network.py”, line 150, in save
model.save(path)
File “C:\Users\codrin_kruijne.conda\envs\py3_knime_tf2_torch\lib\site-packages\tensorflow\python\keras\engine\training.py”, line 1978, in save
save.save_model(self, filepath, overwrite, include_optimizer, save_format,
File “C:\Users\codrin_kruijne.conda\envs\py3_knime_tf2_torch\lib\site-packages\tensorflow\python\keras\saving\save.py”, line 133, in save_model
saved_model_save.save(model, filepath, overwrite, include_optimizer,
File “C:\Users\codrin_kruijne.conda\envs\py3_knime_tf2_torch\lib\site-packages\tensorflow\python\keras\saving\saved_model\save.py”, line 80, in save
save_lib.save(model, filepath, signatures, options)
File “C:\Users\codrin_kruijne.conda\envs\py3_knime_tf2_torch\lib\site-packages\tensorflow\python\saved_model\save.py”, line 984, in save
object_saver.save(utils_impl.get_variables_path(export_dir),
File “C:\Users\codrin_kruijne.conda\envs\py3_knime_tf2_torch\lib\site-packages\tensorflow\python\training\tracking\util.py”, line 1199, in save
save_path, new_feed_additions = self._save_cached_when_graph_building(
File “C:\Users\codrin_kruijne.conda\envs\py3_knime_tf2_torch\lib\site-packages\tensorflow\python\training\tracking\util.py”, line 1145, in _save_cached_when_graph_building
save_op = saver.save(file_prefix, options=options)
File “C:\Users\codrin_kruijne.conda\envs\py3_knime_tf2_torch\lib\site-packages\tensorflow\python\training\saving\functional_saver.py”, line 295, in save
return save_fn()
File “C:\Users\codrin_kruijne.conda\envs\py3_knime_tf2_torch\lib\site-packages\tensorflow\python\training\saving\functional_saver.py”, line 269, in save_fn
sharded_saves.append(saver.save(shard_prefix, options))
File “C:\Users\codrin_kruijne.conda\envs\py3_knime_tf2_torch\lib\site-packages\tensorflow\python\training\saving\functional_saver.py”, line 78, in save
return io_ops.save_v2(file_prefix, tensor_names, tensor_slices, tensors)
File “C:\Users\codrin_kruijne.conda\envs\py3_knime_tf2_torch\lib\site-packages\tensorflow\python\ops\gen_io_ops.py”, line 1728, in save_v2
return save_v2_eager_fallback(
File “C:\Users\codrin_kruijne.conda\envs\py3_knime_tf2_torch\lib\site-packages\tensorflow\python\ops\gen_io_ops.py”, line 1749, in save_v2_eager_fallback
_result = _execute.execute(b"SaveV2", 0, inputs=_inputs_flat, attrs=_attrs,
File “C:\Users\codrin_kruijne.conda\envs\py3_knime_tf2_torch\lib\site-packages\tensorflow\python\eager\execute.py”, line 59, in quick_execute
tensors = pywrap_tfe.TFE_Py_Execute(ctx._handle, device_name, op_name,
tensorflow.python.framework.errors_impl.NotFoundError: Failed to create a NewWriteableFile: C:\Users\codrin_kruijne\AppData\Local\Temp\knime_Deep Learning P18093\fs-DL_Py_0-95-18095\000\000\4b4c3522-0e8c-4285-8248-944d6799c42c\variables\variables_temp_d312bb449a8241afa3b87eb249cd68af/part-00000-of-00001.data-00000-of-00001.tempstate647386528939771127 : The system cannot find the path specified.
; No such process [Op:SaveV2]

Hi @Codrin

I am sorry for the late reply. We are aware that enabling long file paths does not always resolve the issue. In fact, one of our developers has filed an issue over at the Tensorflor repository. In some cases this issue has also been magically resolved, so I’m sorry to say that I don’t have a reliable solution right now.

If you happen to find a solution, we (and the guys over at Tensorflow surely too) would be happy to know!

Kind regards
Marvin