# tacotron-2-support
- e
Eugene
10/21/2022, 10:09 PMI got this error in 4. Begin training! - r
Reclezon
10/21/2022, 10:09 PMThe rest of the notebook will work without it. The tensor cell is just for monitering progress - r
Reclezon
10/21/2022, 10:10 PMI'd use the actually synthesis nb for testing progress. - e
Eugene
10/21/2022, 10:11 PMHow do I solve this? 🤔 - r
Reclezon
10/21/2022, 10:18 PMNot sure but I'd assume @mega b (sorry) might? Either them or Gdrive just yeeted into the void... - r
Reclezon
10/21/2022, 10:18 PMI am getting errors for Gdrive for some reason - u
{K EY1} (Kei)
10/21/2022, 10:26 PM@Eugene try adding
to to the top of the cell maybe?import uberduck_ml_dev - e
Eugene
10/21/2022, 10:27 PMLike here? - u
{K EY1} (Kei)
10/21/2022, 10:27 PMYeah - e
Eugene
10/21/2022, 10:29 PMHm, no, still a similar error - u
{K EY1} (Kei)
10/21/2022, 10:29 PMOk make a new cell - u
{K EY1} (Kei)
10/21/2022, 10:30 PMWait question You ran the setup cell right? - u
{K EY1} (Kei)
10/21/2022, 10:31 PMIf you did and it's just not working, add this to a new cell and run it:!pip install git+https://github.com/uberduck-ai/uberduck-ml-dev@b4246a8830ed4bbd21ce3b2b82df24b61a128d56 - e
Eugene
10/21/2022, 10:31 PMHmm.. I did, and there's a green checkmark next to it which I believe means it went successfully. However, there is some text in red in the output there. - u
{K EY1} (Kei)
10/21/2022, 10:32 PMOh i see what happened The connection timed out while you were installing it No problem, just either make the new cell and run that command or re-run the setup cell - e
Eugene
10/21/2022, 10:33 PMAhh okay, I'll try that :) - u
{K EY1} (Kei)
10/21/2022, 10:34 PMWatch to see if the red text appears again cuz if it does there's some sort of problem - e
Eugene
10/21/2022, 10:37 PMThis is the only red text this time. - u
{K EY1} (Kei)
10/21/2022, 10:38 PMOk good! It should work now - e
Eugene
10/21/2022, 10:39 PMRunning 4. Begin Training again - e
Eugene
10/21/2022, 10:39 PMIt is certainly doing something! - e
Eugene
10/21/2022, 10:40 PMBut what should I do now? If I've understood it correctly, this training will just keep on forever? And normally TensorBoard should give a visual indication of how far it is and when it's a good time to stop the training? - e
Eugene
10/21/2022, 10:40 PMBut TensorBoard doesn't work now. So I don't know for how long I should run this. - e
Eugene
10/21/2022, 10:41 PMOr how I can test some text to speech stuff after that. - u
{K EY1} (Kei)
10/21/2022, 10:41 PMTrain until it gets to a certain val loss - u
{K EY1} (Kei)
10/21/2022, 10:41 PMOne sec - r
Reclezon
10/21/2022, 10:43 PMSince tensor doesn't work for me either I test models using the Synth nb. Also a minor but early indicator if a model needs a custom hifigan - r
Reclezon
10/21/2022, 10:43 PMShould also be in #841437191073955920 - u
{K EY1} (Kei)
10/21/2022, 10:43 PMIt'll say every once in a while "Validation loss: 0.something" Train until it hits a low enough validation loss - e
Eugene
10/21/2022, 10:43 PMThe val loss started at 0.52. Now, a few minutes in, it's at 0.28. I have 94 files so then I should wait till the val loss is 0.09?