# tacotron-2-support
- c
Cris140
08/14/2022, 3:56 PMThe tensorboard error means that there isn't any training data to analyze, probably because the training of the model hasn't started - d
Dingusmcbingus
08/14/2022, 3:57 PMok - c
Cris140
08/14/2022, 3:57 PMYou transcription probably has a blank line at the end - d
Dingusmcbingus
08/14/2022, 3:57 PMok - d
Dingusmcbingus
08/14/2022, 4:20 PMum the training just finished but the model didn't save to my drive in the tacotron folder - u
{K EY1} (Kei)
08/14/2022, 6:13 PMYou need to set the output to your drive, idk if it's there by default - u
{K EY1} (Kei)
08/14/2022, 6:13 PMAnd if you stopped the training before it saved once, or set the save interval to really high, it wouldn't have saved - c
Cris140
08/14/2022, 7:52 PMI don't think the training has started, you're the one who finishes it, it doesn't stop training automatically - s
SamSchultz1950
08/15/2022, 1:45 AMDamn it - u
{K EY1} (Kei)
08/15/2022, 2:06 AMPretty sure that's an old notebook Use one of the ones in #841437191073955920 - s
SamSchultz1950
08/15/2022, 6:57 AMnope, still getting nothing - h
haru0l
08/15/2022, 7:27 AMthat hifigan link looks broken - u
{K EY1} (Kei)
08/15/2022, 11:11 AMI mean... you didn't put anything in the boxes - a
AppData 174
08/15/2022, 6:42 PMik this is probably a pretty stupid question but im new to making models so can you load an existing trained model from Google Drive to train it again later? you need to just mount gdrive and start the train model cell? - h
haru0l
08/15/2022, 7:49 PMyup, as long as you save and point to the same directory later on it should auto resume from the most recent checkpoint - u
(Dawn) Will Draw Fictional Women
08/15/2022, 7:55 PMif you’re asking about tacotron PLEASE clarify if its legacy or pipeline it saves us tons of headache - u
(Dawn) Will Draw Fictional Women
08/15/2022, 7:55 PMjust general knowledge - l
ldmtwo
08/15/2022, 9:44 PMI've been trying to get a voice to pronounce "A" (aye) as in AI, but it will not make that sound no matter how I write it. I got close with "hey eye". This was tested using Spy voice (https://app.uberduck.ai/speak#mode=tts-basic&voice=spy) Any idea because this is a deal breaker if it doesn't work? - u
(Dawn) Will Draw Fictional Women
08/15/2022, 9:50 PMits arpabet - u
(Dawn) Will Draw Fictional Women
08/15/2022, 9:51 PM{EY1} - h
hecko
08/15/2022, 10:55 PMfor arpabet models plain text
should work tooay i - h
hecko
08/15/2022, 10:55 PMarpabetn't models are hit-or-miss in general - u
(Dawn) Will Draw Fictional Women
08/15/2022, 11:07 PMmanual arpa is the goat - d
DiamondBrickZ
08/16/2022, 3:38 AMagain, this is from the very first notebook in the notebooks channel, legacy tacotron 2, i tried doing all the things i was supposed to but alas no more - d
DiamondBrickZ
08/16/2022, 3:38 AMi even did the thing kei told me to, yet nothing - d
DiamondBrickZ
08/16/2022, 3:39 AMoh also i just restarted from the beginning rather than continuing the old session, it's been a few days and i am not experienced to go back in and fix what may have been fucked up - h
haru0l
08/16/2022, 4:31 AMi believe the hifigan link in that is just outdated - h
haru0l
08/16/2022, 4:31 AMtry replacing the drive id of that link to whats used in the pipeline synthesizer - a
AppData 174
08/16/2022, 6:06 AMOK thanks, also I'm using legacy - a
AppData 174
08/16/2022, 6:27 AMYou need to also import the WAVs and the transcript?