error running aocr

[yanpozka]

After installing aocr (pip install aocr) in both my OS and the official tensor docker container(gcr.io/tensorflow/tensorflow) I'm getting this error even when I run aocr --help:

Traceback (most recent call last):
  File "/usr/local/bin/aocr", line 7, in <module>
    from aocr.__main__ import main
  File "/usr/local/lib/python2.7/dist-packages/aocr/__main__.py", line 15, in <module>
    from .model.model import Model
  File "/usr/local/lib/python2.7/dist-packages/aocr/model/model.py", line 20, in <module>
    from .seq2seq_model import Seq2SeqModel
  File "/usr/local/lib/python2.7/dist-packages/aocr/model/seq2seq_model.py", line 27, in <module>
    from .seq2seq import model_with_buckets
  File "/usr/local/lib/python2.7/dist-packages/aocr/model/seq2seq.py", line 74, in <module>
    linear = rnn_cell._linear  # pylint: disable=protected-access
AttributeError: 'module' object has no attribute '_linear'

# uname -a
Linux 71e18aa0cfc4 4.10.0-42-generic #46~16.04.1-Ubuntu SMP Mon Dec 4 15:57:59 UTC 2017 x86_64 x86_64 x86_64 GNU/Linux

[AdamWP]

In some versions of TF _linear was moved from rnn_cell to rnn_cell_impl.

You need to modify /usr/local/lib/python2.7/dist-packages/aocr/model/seq2seq.py

Change line 71 from "from tensorflow.contrib.rnn.python.ops import rnn, rnn_cell"
to
"from tensorflow.contrib.rnn.python.ops import rnn
from tensorflow.python.ops import rnn_cell_impl"

Change line 74 "linear = rnn_cell._linear" to "rnn_cell_impl._linear"

[emedvedev]

Yup, looks like it's been moved in 1.4, so aocr is lagging a little behind.

@AdamWP would you mind submitting a PR? Ideally with a conditional import, to keep compatibility with the older TF versions.

[AdamWP]

@emedvedev Done!

[emedvedev]

Thanks! Closing the issue.

[avielas]

@emedvedev but how did you solve that?

[emedvedev]

@avielas #51

[avielas]

thanks so much!!

[avielas]

@emedvedev #51 didn't work for me. I used tensorflow version 1.4.0 on ubuntu 16, python version 2.7. my seq2seq.py:

from tensorflow.python.ops import nn_ops
from tensorflow.contrib.rnn.python.ops import rnn
from tensorflow.python.ops import variable_scope

try:
from tensorflow.contrib.rnn.python.ops import rnn_cell_impl
except ImportError:
from tensorflow.python.ops import rnn_cell_impl

try:
linear = rnn_cell_impl._linear # pylint: disable=protected-access
except AttributeError:
from tensorflow.contrib.rnn.python.ops import core_rnn_cell
linear = core_rnn_cell._linear # pylint: disable=protected-access

Any ideas?

[emedvedev]

Can you paste the full error log you’re getting? Maybe there’s been another change between TF versions.

…

On Apr 25, 2018, 23:55 +0200, avielas ***@***.***>, wrote: @emedvedev #51 didn't work for me. I used tensorflow version 1.4.0 on ubuntu 16. my seq2seq.py: from tensorflow.python.ops import nn_ops from tensorflow.contrib.rnn.python.ops import rnn from tensorflow.python.ops import variable_scope try: from tensorflow.contrib.rnn.python.ops import rnn_cell_impl except ImportError: from tensorflow.python.ops import rnn_cell_impl try: linear = rnn_cell_impl._linear # pylint: disable=protected-access except AttributeError: from tensorflow.contrib.rnn.python.ops import core_rnn_cell linear = core_rnn_cell._linear # pylint: disable=protected-access Any ideas? — You are receiving this because you were mentioned. Reply to this email directly, view it on GitHub, or mute the thread.

[avielas]

sure:
(virtualenv_attention) @:~/DL/attention-ocr$ sudo aocr dataset ./datasets/annotations-training.txt ./datasets/training.tfrecords
Traceback (most recent call last):
File "/usr/local/bin/aocr", line 7, in
from aocr.main import main
File "/usr/local/lib/python2.7/dist-packages/aocr/main.py", line 15, in
from .model.model import Model
File "/usr/local/lib/python2.7/dist-packages/aocr/model/model.py", line 20, in
from .seq2seq_model import Seq2SeqModel
File "/usr/local/lib/python2.7/dist-packages/aocr/model/seq2seq_model.py", line 27, in
from .seq2seq import model_with_buckets
File "/usr/local/lib/python2.7/dist-packages/aocr/model/seq2seq.py", line 83, in
linear = core_rnn_cell._linear # pylint: disable=protected-access
AttributeError: 'module' object has no attribute '_linear'

[avielas]

I tried tf 1.2 1.4 1.8. what exactly version should I use?

[avielas]

update error(tf 1.2):
(virtualenv_aocr_0_7_3) @computer-name:~/DL/aocr-0.7.3$ sudo aocr dataset ./datasets/annotations-training.txt ./datasets/training.tfrecords
Traceback (most recent call last):
File "/usr/local/bin/aocr", line 9, in
load_entry_point('aocr==0.7.3', 'console_scripts', 'aocr')()
File "/usr/lib/python2.7/dist-packages/pkg_resources/init.py", line 542, in load_entry_point
return get_distribution(dist).load_entry_point(group, name)
File "/usr/lib/python2.7/dist-packages/pkg_resources/init.py", line 2569, in load_entry_point
return ep.load()
File "/usr/lib/python2.7/dist-packages/pkg_resources/init.py", line 2229, in load
return self.resolve()
File "/usr/lib/python2.7/dist-packages/pkg_resources/init.py", line 2235, in resolve
module = import(self.module_name, fromlist=['name'], level=0)
File "build/bdist.linux-x86_64/egg/aocr/main.py", line 15, in
File "build/bdist.linux-x86_64/egg/aocr/model/model.py", line 20, in
File "build/bdist.linux-x86_64/egg/aocr/model/seq2seq_model.py", line 27, in
File "build/bdist.linux-x86_64/egg/aocr/model/seq2seq.py", line 83, in
AttributeError: 'module' object has no attribute '_linear'

[emedvedev]

Thanks! One more thing: which aocr version are you running, and is there a difference in the error logs between the latest version in pip and the github master?

…

On Apr 26, 2018, 00:18 +0200, avielas ***@***.***>, wrote: update error(tf 1.2): (virtualenv_aocr_0_7_3) @computer-name:~/DL/aocr-0.7.3$ sudo aocr dataset ./datasets/annotations-training.txt ./datasets/training.tfrecords Traceback (most recent call last): File "/usr/local/bin/aocr", line 9, in load_entry_point('aocr==0.7.3', 'console_scripts', 'aocr')() File "/usr/lib/python2.7/dist-packages/pkg_resources/init.py", line 542, in load_entry_point return get_distribution(dist).load_entry_point(group, name) File "/usr/lib/python2.7/dist-packages/pkg_resources/init.py", line 2569, in load_entry_point return ep.load() File "/usr/lib/python2.7/dist-packages/pkg_resources/init.py", line 2229, in load return self.resolve() File "/usr/lib/python2.7/dist-packages/pkg_resources/init.py", line 2235, in resolve module = import(self.module_name, fromlist=['name'], level=0) File "build/bdist.linux-x86_64/egg/aocr/main.py", line 15, in File "build/bdist.linux-x86_64/egg/aocr/model/model.py", line 20, in File "build/bdist.linux-x86_64/egg/aocr/model/seq2seq_model.py", line 27, in File "build/bdist.linux-x86_64/egg/aocr/model/seq2seq.py", line 83, in AttributeError: 'module' object has no attribute '_linear' — You are receiving this because you were mentioned. Reply to this email directly, view it on GitHub, or mute the thread.