so upcoming version of tfp will not support tf 2.0 no more ?
Just wondering: where did you get this information from? Is there a comfortable way to stay up-to-date on this things?
Just wondering: where did you get this information from? Is there a comfortable way to stay up-to-date on this things?
from error
Sorry, which error? Could you point me to it?
in tf 2.0 and tfp>=0.9 case, import tensorflow then import tfp will throw such kind of error.
Can you maybe post the error here? Doesn't it say that it needs 2.1 or higher?
as here: https://github.com/tensorflow/probability/blob/master/tensorflow_probability/__init__.py#L56
I use TF 2.1 and TFP 0.9 and don't see any error.
TFP is developed against TF head; our current practice is that every time TF cuts a new release, we cut a new TFP release. So in general every TFP version is compatible with the most recent TF version at the time of its release: TFP 0.9 was released in sync with TF 2.1, and TFP 0.10 will (most likely) be released shortly after TF 2.2.
We don't in general have the resources to backport TFP to older TF versions, though of course we're happy to accept patches if there's an incompatibility that's easily fixable.
Also, we aim to keep tfp-nightly compatible with tf-nightly and tf-nightly-cpu.
Most helpful comment
TFP is developed against TF head; our current practice is that every time TF cuts a new release, we cut a new TFP release. So in general every TFP version is compatible with the most recent TF version at the time of its release: TFP 0.9 was released in sync with TF 2.1, and TFP 0.10 will (most likely) be released shortly after TF 2.2.
We don't in general have the resources to backport TFP to older TF versions, though of course we're happy to accept patches if there's an incompatibility that's easily fixable.