Upgrade to Tensorflow V 2.x?

[v2thegreat]

Hey, seeing as there's a new version out for TensorFlow, and there is a tool for upgrading from TF 1.x to 2.x

Is this something that can be done for this project? Especially seeing as the current version of TF that this project has been tested on is still running on 1.8, which is not even available on the latest version of Python 3.7 (due to being deprecated I think)

[el-tocino]

I think you'd have to DIY and submit a PR. The head repo also is currently incompatible: keithito#309
Mimic2 seems to be pretty low on the priority list for the mycroft team right now.

[v2thegreat]

Hmm, isn't that the point of the upgrade tool.
I'll run the tool and make sure all the tests pass. If it's not too much of a hassle to fix whatever issues come up next, I'll make a new PR

[PureTryOut]

Hey @v2thegreat did you have any luck with this?

[v2thegreat]

I'm sorry to say that I haven't had the time to explore this. Been busy with Uni and work and some other projects. Actually, I wasn't able to get this working too far in the first place. I might try to get back into getting mimic2 working and then porting over to V2, but no guarantees for now since I'm not very familiar with TensorFlow yet.

If someone more experienced than me wants to take a crack at it, go ahead

[catfishjw]

after running the fixing tool on the entire file, I can confirm that this does not work. While much of the project can use Tensorflow's built in compadability features, the project is too dependent on tf.contrib, which was entirely removed in Tensorflow 2.X. The only way for it to work would be to go in and replace all instances of tf.contrib with their supported functions. Im giving it a shot, but at the end of the day I'm not proficient enough at Tensorflow to fix the various issues this introduces. If I do eventually figure it out, I will post results here.

[el-tocino]

I'd skip working on this and go contribute to https://github.com/coqui-ai/TTS/ instead, which is already leaps ahead and actively being worked on.