https://github.com/tutorials/rnn/translate/translate.py throws TypeError: can't pickle _thread.lock objects

System information

• What is the top-level directory of the model you are using:
  models/tutorials/rnn/translate/
• Have I written custom code (as opposed to using a stock example script provided in TensorFlow):
  No
• OS Platform and Distribution (e.g., Linux Ubuntu 16.04):
  Windows 10
• TensorFlow installed from (source or binary):
  Pip install
• TensorFlow version (use command below):
  1.2.0
• Bazel version (if compiling from source):
  n/a
• CUDA/cuDNN version:
  Not GPU
• GPU model and memory:
  Not GPU
• Exact command to reproduce:
  Run translate.py

Describe the problem

When I run translate.py, it loads all the data from the internet on first run, then creates 3 layers of 1024 units, then crashes with the log below. I believe the issue lies with having a list of buckets with len > 1.

Source code / logs

https://pastebin.com/5PQeXx0J
Stack-Overflow with more details:
https://stackoverflow.com/questions/44855603/typeerror-cant-pickle-thread-lock-objects-in-seq2seq

[richardsun-voyager]

I ran into the same problem on my MacBook Pro, I thought perhaps there were some bugs in the new version 1.2.0, so I switched to Tensorflow 1.0.0 version and it worked. Will anybody give any clues to solve it?

[matteo-brucato]

I'm having the same issue.

[alive-zk]

@panyx0718, any thoughts on this?

@Ali01 Update: I actually get the issue on my own project too. It appears to only occur when a bucket of length > 1 is fed to the model ([100, 100), (200, 200)] fails while ([100, 100]) works fine.

For more details, I linked the Stackoverflow in the main issue post.

[richardsun-voyager]

Well, I created a virtual environment by conda prompt on my Mac, installed python3.6 and tensorflow 1.0.0 along with kernel. Then I switched to tensorflow 1.0.0 kernel in my jupyter notebook.

[dxteam2024]

try to delete the file named 'pycache' in your python workspace , i have the same problem and fixed in this way

[Jacobh2]

I get the same problem as OP, also get the exact same error log.

I'm also running Windows 10, and have tried both with and without gpu support on Tensorflow 1.2.0.

I get it to work if I change the list of buckets (_bucket) to length 1 and makes sure the __pycache__ folder is deleted (if I've done any failing runs before)

[DongjunLee]

also i have the same issue.

[guotong1988]

same problem ...

[maxim5]

@DongjunLee @guotong1988

Try the answer suggested on StackOverflow:

setattr(tf.contrib.rnn.GRUCell, '__deepcopy__', lambda self, _: self)
setattr(tf.contrib.rnn.BasicLSTMCell, '__deepcopy__', lambda self, _: self)
setattr(tf.contrib.rnn.MultiRNNCell, '__deepcopy__', lambda self, _: self)