# machine-learning
- j
jmcb
06/22/2021, 4:13 PMthen I would have thought you'd understand the idea - p
pi
06/22/2021, 4:13 PMagain: the model learns a grapheme to phoneme mapping - p
pi
06/22/2021, 4:13 PMnothing more - j
jmcb
06/22/2021, 4:14 PMok - j
jmcb
06/22/2021, 4:15 PMI guess I didn't understand how it worked in as much detail so I didn't immediately realise why my idea wouldn't work - m
mega b
06/22/2021, 4:23 PMI didn't manually extract each time he was talking, i simply took every 12 seconds of audio in the video and I transcribed all the chunks - m
mega b
06/22/2021, 4:23 PMits bad practice, but it shows that even with zero effort, you can make a quite decent model - p
pi
06/22/2021, 4:26 PMand what did you use for transcriptions - p
pi
06/22/2021, 4:26 PMaws transcribe? - m
mega b
06/22/2021, 4:26 PMone moment - m
mega b
06/22/2021, 4:26 PM - m
mega b
06/22/2021, 4:27 PMit's really well done, should check it out 👍 - p
pi
06/22/2021, 4:27 PMoh ok - p
pi
06/22/2021, 4:27 PMi used it before - m
mega b
06/22/2021, 4:27 PMoh nice - p
pi
06/22/2021, 4:28 PMit's very nice for working on low resource languages too - m
mega b
06/22/2021, 4:32 PMRight - r
recker
06/22/2021, 6:57 PMyo - s
Scramble
06/22/2021, 6:58 PMdoes anyone when I should stop training on talknet? - r
recker
06/22/2021, 6:59 PMhow do you train? - s
Scramble
06/22/2021, 7:00 PMI'm not talking about tacotron2 (what uberduck uses) - a
Awesome Face (GTX 1660 Ti)
06/22/2021, 7:00 PMyes - a
Awesome Face (GTX 1660 Ti)
06/22/2021, 7:00 PMi heard someone said you have to stop learning rate at 1 - a
Awesome Face (GTX 1660 Ti)
06/22/2021, 7:00 PMor else you are over training it - a
Awesome Face (GTX 1660 Ti)
06/22/2021, 7:00 PMi saw that some pictures - s
Scramble
06/22/2021, 7:00 PMlearning rate? - s
Scramble
06/22/2021, 7:01 PMOHHHH - a
Awesome Face (GTX 1660 Ti)
06/22/2021, 7:01 PMtalking about this - r
recker
06/22/2021, 7:01 PMbro... I'm basically asking how do I train - a
Awesome Face (GTX 1660 Ti)
06/22/2021, 7:01 PMuh... #841422801965416538