I am wondering if it is possible to update the TensorFlow version (currently v1.4.0). The reason is that the more recent keras releases (e.g. 2.2.2) aren't guaranteed to be compatible with older TensorFlow versions [1]. To be more specific, I think that the most recent stable v1.* release would solve the issue I'm seeing -- i.e., I'm not asking to upgrade to tensorflow 2.0.
I included more details on why this became an issue for me in [2].
Thanks,
Sam
[1] https://github.com/keras-team/keras/issues/10440
[2] I tried using one of the GPUs on uaf-1 through a singularity container [3] today, and found a keras/tensorflow error when trying to run a DNN. This same script worked a few weeks ago, and I think the error may be due to the fact that keras was updated (?) and the newer version expects a newer tensorflow version than what is currently in the singularity container (1.4.0):
I found that other people [1] have found the same error I'm seeing [4] -- the solution that the keras team suggested is to upgrade tensorflow, as they only guarantee support for the most recent releases.
efajardo
commented
5 years ago
Hi,
I am wondering if it is possible to update the TensorFlow version (currently v1.4.0). The reason is that the more recent keras releases (e.g. 2.2.2) aren't guaranteed to be compatible with older TensorFlow versions [1]. To be more specific, I think that the most recent stable v1.* release would solve the issue I'm seeing -- i.e., I'm not asking to upgrade to tensorflow 2.0.
I included more details on why this became an issue for me in [2].
Thanks, Sam
[1] https://github.com/keras-team/keras/issues/10440 [2] I tried using one of the GPUs on uaf-1 through a singularity container [3] today, and found a keras/tensorflow error when trying to run a DNN. This same script worked a few weeks ago, and I think the error may be due to the fact that keras was updated (?) and the newer version expects a newer tensorflow version than what is currently in the singularity container (1.4.0):
I found that other people [1] have found the same error I'm seeing [4] -- the solution that the keras team suggested is to upgrade tensorflow, as they only guarantee support for the most recent releases.
[3]
singularity shell --bind /usr/lib64/nvidia:/host-libs /cvmfs/singularity.opensciencegrid.org/opensciencegrid/tensorflow-gpu:latest[4]TypeError: while_loop() got an unexpected keyword argument 'maximum_iterations'