I have read similar questions - my error appears to be different since the solutions proposed do not solve my problem.
I am having trouble plotting graphs of keras models.
I have installed graphviz binaries using homebrew
I have installed graphviz python wrapper and pydot using pip (also tried with conda, since this seems to have been an issue in the past).
Using python 3.5
Running:
from keras.utils import plot_model
plot_model(cnn_model, to_file='cnn_model.png')
I get the error:
ImportError: Failed to import pydot. You must install pydot and graphviz for
pydotprint
to work.
with the trace:
---------------------------------------------------------------------------
AttributeError Traceback (most recent call last)
/Users/jusjosgra/anaconda/lib/python3.5/site-packages/keras/utils/vis_utils.py in _check_pydot()
26 # so no specific class can be caught.
---> 27 raise ImportError('Failed to import pydot. You must install pydot'
28 ' and graphviz for `pydotprint` to work.')
AttributeError: 'NoneType' object has no attribute 'Dot'
During handling of the above exception, another exception occurred:
ImportError Traceback (most recent call last)
<ipython-input-450-82ff54d9260b> in <module>()
1 from keras.utils import plot_model
----> 2 plot_model(cnn_model, to_file='cnn_model.png')
/Users/jusjosgra/anaconda/lib/python3.5/site-packages/keras/utils/vis_utils.py in plot_model(model, to_file, show_shapes, show_layer_names, rankdir)
133 if not extension:
134 extension = 'png'
--> 135 else:
136 extension = extension[1:]
137 dot.write(to_file, format=extension)
/Users/jusjosgra/anaconda/lib/python3.5/site-packages/keras/utils/vis_utils.py in model_to_dot(model, show_shapes, show_layer_names, rankdir)
54 dot.set('rankdir', rankdir)
55 dot.set('concentrate', True)
---> 56 dot.set_node_defaults(shape='record')
57
58 if isinstance(model, Sequential):
/Users/jusjosgra/anaconda/lib/python3.5/site-packages/keras/utils/vis_utils.py in _check_pydot()
29
30
---> 31 def model_to_dot(model,
32 show_shapes=False,
33 show_layer_names=True,
I can successfully import pydot and graphviz independently.
There appears to be a history of bugs between keras and graphviz. Any ideas on a solution?
The error message is ambiguous: the exception could be raised also when pydot
(or any of the forks mentioned in the module vis_utils
) is successfully imported, but calling pydot.Dot.create
fails. From https://github.com/keras-team/keras/blob/4eab0556d29f11ff41758d80c15d6457263f6a93/keras/utils/vis_utils.py:
def _check_pydot():
try:
# Attempt to create an image of a blank graph
# to check the pydot/graphviz installation.
pydot.Dot.create(pydot.Dot())
except Exception:
# pydot raises a generic Exception here,
# so no specific class can be caught.
raise ImportError('Failed to import pydot. You must install pydot'
' and graphviz for `pydotprint` to work.')
and the method pydot.Dot.create
attempts to call the executable dot
(installed by GraphViz):
https://github.com/erocarrera/pydot/blob/d6ac9e9244d1a882103422ac2b35ceef96f5dfe3/pydot.py#L1856
If dot
is not in the environment's PATH
variable, then it is invisible to pydot
, despite being present on the machine.
Importing packages in a Python interpreter means that they are available under site-packages
, or from wherever they were installed in development mode (e.g., with python setup.py develop
, or with pip install -e .
). Whether executables of GraphViz are on the path is a separate issue.
Also, the Python package graphviz
is unrelated to pydot
, and not needed for using GraphViz via pydot
. For more on this issue, please see: