tensorflow 2.1 doesn't require separate pip installes for gpu and cpu about openl3 HOT 8 CLOSED

Thanks for your interest! :)

Currently, the most recent version supports Tensorflow 1.0 versions. We haven't explicitly looked at 2.0 (and 2.1) versions yet, but if or when we decide to drop support for 1.0 then we can explicitly add the requirement.

from openl3.

@jtcramer sure. feel free to close this then

from openl3.

@jtcramer and @faroit : I tried using it with tensorflow 2.1 on Colab and it gave me a compatibility error.

Note that pip install tensorflow installs tensorflow2.1 now, maybe force it to be pip install tensorflow==1.14.0

from openl3.

@jtcramer we should update the instructions in the readme per^^

from openl3.

FYI: @justinsalamon @jtcramer pip install openl3 is now force installing tf2.0, even though I had first pip install tensorflow==1.14.0 before pip install openl3

My previous openl3 installation (based on pip install openl3) from 9 days ago still works fine.

from openl3.

Thanks @happypanda5 for the heads up, we'll look into this as soon as we can.

from openl3.

Per offline discussion with @justinsalamon, we'll plan on updating the dependencies to ensure that keras < 2.3.0 (2.3.0 uses TF 2.0) and add a note to the README that TF 2.* is not yet supported and make a v0.3.1 release. We'll aim to move to TF 2.0 (and thus tf.keras) in a future release at some point.

from openl3.

Closing this issue, follow #42 for more on the installation problems.

from openl3.