dragonn icon indicating copy to clipboard operation
dragonn copied to clipboard

Anaconda installation ships with incorrect keras installation

Open rmovva opened this issue 7 years ago • 3 comments

Hi,

I installed dragonn with "conda install dragonn -c kundajelab", but I got the following error message upon trying to train the example model:

Traceback (most recent call last): File “/Library/Frameworks/Python.framework/Versions/2.7/bin/dragonn”, line 6, in from pkg_resources import load_entry_point File “/Library/Frameworks/Python.framework/Versions/2.7/lib/python2.7/site-packages/pkg_resources/init.py”, line 2953, in @_call_aside File “/Library/Frameworks/Python.framework/Versions/2.7/lib/python2.7/site-packages/pkg_resources/init.py”, line 2939, in _call_aside f(*args, **kwargs) File “/Library/Frameworks/Python.framework/Versions/2.7/lib/python2.7/site-packages/pkg_resources/init.py”, line 2966, in _initialize_master_working_set working_set = WorkingSet._build_master() File “/Library/Frameworks/Python.framework/Versions/2.7/lib/python2.7/site-packages/pkg_resources/init.py”, line 637, in _build_master return cls._build_from_requirements(requires) File “/Library/Frameworks/Python.framework/Versions/2.7/lib/python2.7/site-packages/pkg_resources/init.py”, line 650, in _build_from_requirements dists = ws.resolve(reqs, Environment()) File “/Library/Frameworks/Python.framework/Versions/2.7/lib/python2.7/site-packages/pkg_resources/init.py”, line 829, in resolve raise DistributionNotFound(req, requirers) pkg_resources.DistributionNotFound: The ‘keras==0.3.3’ distribution was not found and is required by dragonn

Looks like keras v0.3.2 ships with the install but keras v0.3.3 is necessary for training.

Thanks, Rajiv

rmovva avatar Apr 19 '17 04:04 rmovva

@rmovva: Thank you for reporting this issue. I repackaged dragonn v0.1.2 with keras v0.3.3. Can you try conda install dragonn -c kundajelab again?

jisraeli avatar Apr 21 '17 02:04 jisraeli

The anaconda (python 2) installation now works cleanly on a linux-64 system (dragonn-0.1.2, keras-0.3.3). However, on osx-64, the installation now seems to default to a deprecated version of dragonn (v0.1.0).

rmovva avatar Apr 25 '17 07:04 rmovva

Installing DragoNN on a machine with Keras already installed fails with this error: "cannot import name 'TimeDistributedDense'" TimeDistributedDense was deprecated a long time ago, any plans to refresh Keras version?

gatapia avatar May 01 '18 03:05 gatapia