# machine-learning
- h
heaven inspector
06/17/2021, 3:50 PMWell I got 10 minutes and 211 voice clips for Hawkodile, I hope despite what the epochs are showing it turns out well - s
SuperLuig
06/17/2021, 3:58 PMin hindsight getting 19 minutes of data for mao mao probably was overkill - s
SuperLuig
06/17/2021, 3:58 PM10 or 11 would have been fine - i
Isoar
06/17/2021, 4:09 PMA ramp like that is what you want for your model - i
Isoar
06/17/2021, 4:10 PM(at least thats what ive been told, and its worked for my 2 :) - u
user
06/17/2021, 4:34 PMDid some training for seven - h
heaven inspector
06/17/2021, 4:57 PMHmm this is odd, Hawkodile is only 40 MBs instead of 322. Maybe I'll try to retrain him again - j
Justin
06/17/2021, 5:00 PMOoff you gotta retrain it - j
Justin
06/17/2021, 5:01 PMModel didn't save at the correct time or you stopped too early or something - t
TGG
06/17/2021, 8:37 PMBarney has a stroke - g
Gushiavo_N2
06/17/2021, 9:55 PMHow many epochs and batch size for Hawkodile but it please to unblock friend - j
Jara
06/17/2021, 10:59 PMit's because hewent through puberty over the show, and died too young - m
mega b
06/17/2021, 11:00 PMmake sure you use my colab, and the session doesnt crash - h
heaven inspector
06/17/2021, 11:15 PMI already got him trained and I put it in #835647711305793551 - u
user
06/17/2021, 11:26 PMmega b - u
user
06/18/2021, 5:43 AMI'm new to tacotron and while I'm checking the data it says [WARNING] wavs/1.wav in filelist while expecting .npy . - u
user
06/18/2021, 5:43 AMit says the same thing for every file - i
IBob012
06/18/2021, 5:44 AMthats the error jacob got when he was first starting out, i helped him fix that - i
IBob012
06/18/2021, 5:46 AMactually i think you should try to find the small conversation i had with him here, he sent images so its easier to understand what he was doing - i
IBob012
06/18/2021, 5:48 AMactually? im going to give you the code i was planning to give to jacob, ill tell you where to place it - u
user
06/18/2021, 5:49 AMgenerating mels doesnt fix it - i
IBob012
06/18/2021, 5:49 AMreplace the generating mels code with this !sed -i -- 's,.wav|,.npy|,g' {hparams.training_files} !sed -i -- 's,.wav|,.npy|,g' {hparams.validation_files} check_dataset(hparams) - m
melody
06/18/2021, 5:50 AMhey im new here if i were looking to create a dataset for a voice, what format would that ideally be in? like how should i arrange the voice lines/text or whatever to submit - u
user
06/18/2021, 5:50 AMIt worked - u
user
06/18/2021, 5:50 AMThank you very much - i
IBob012
06/18/2021, 5:50 AMhell yeah thank cookiegalaxy for that he gave me that code that saved my ass - i
IBob012
06/18/2021, 5:51 AMthe video pinned here is a good start i think - m
melody
06/18/2021, 5:51 AMis it just me or is the mf in this video totally using a voice changer 🤨 - m
melody
06/18/2021, 5:52 AMalso no external emote perms https://tenor.com/view/money-wallet-broke-gif-7855913 - i
IBob012
06/18/2021, 5:52 AMnah its not only you lol