search

tensorflow / fold

Deep learning with dynamic computation graphs in TensorFlow
Apache License 2.0
1.83k stars 266 forks source link

TFFold does not install on Tensorflow 1.0 #27

Open shivajid opened 7 years ago

shivajid commented 7 years ago

TFFold does not install on Tensorflow 1.0 instead it needs the rc0 mentioned here. I have TF 1.0 and it did not install. It said it is not supported for the environment.

moshelooks commented 7 years ago

Thanks for letting us know. Would you mind pasting in the full trace?

On Sun, Mar 5, 2017 at 10:19 PM, Shivaji Dutta notifications@github.com wrote:

TFFold does not install on Tensorflow 1.0 instead it needs the rc0 mentioned here. I have TF 1.0 and it did not install. It said it is not supported for the environment.

— You are receiving this because you are subscribed to this thread. Reply to this email directly, view it on GitHub https://github.com/tensorflow/fold/issues/27, or mute the thread https://github.com/notifications/unsubscribe-auth/AAsbjsMc1DPItvb1TFXnAbDpq9FXaLj3ks5ri6V-gaJpZM4MTw8J .

arwhirang commented 7 years ago

I installed the tensorflow 1.0.1 version, then install the fold library. Installation was completed without fuss. But when I tried to run the fold code, my ipython says 'dead kernel...' I found that when I downgraded the tensorflow to version 1.0, no problem occurs.

shivajid commented 7 years ago

@moshelooks - I have gone past the error, so i cannot give the details. It gave me a warning that the tffold is not compatible.

oychang commented 7 years ago

I had the same issue as @arwhirang with dying notebooks. In TensorFlow v1.0.1 there is a segmentation fault somewhere in the software stack that goes away when downgrading and using TF v1.0.0.

moshelooks commented 7 years ago

Because TensorFlow does not have a stable ABI you should to run Fold with exactly the same version of TensorFlow that Fold was compiled against or segfaults will happen. There's no way to mix and match different versions.

On Mon, Mar 27, 2017 at 11:59 AM, Oliver Chang notifications@github.com wrote:

I had the same issue as @arwhirang https://github.com/arwhirang with dying notebooks. In TensorFlow v1.0.1 there is a segmentation fault somewhere in the software stack that goes away when downgrading and using TF v1.0.0.

— You are receiving this because you were mentioned. Reply to this email directly, view it on GitHub https://github.com/tensorflow/fold/issues/27#issuecomment-289551446, or mute the thread https://github.com/notifications/unsubscribe-auth/AAsbjspnQY61Ik7UblH9rReZDyauyHWVks5rqAcPgaJpZM4MTw8J .