# tacotron-2-support
- c
CmanflipYT
07/15/2022, 2:35 AMI understand. I can't stop replaying this though, this got me dying XD - t
TheRoyalRuby2000
07/15/2022, 2:35 AMOh yeah Cmanflip, make sure the learning rate for Pipeline training is 0.0007 - c
CmanflipYT
07/15/2022, 2:35 AMAh okay I'll note that out - u
(Dawn) Will Draw Fictional Women
07/15/2022, 2:37 AMwhat - t
TheRoyalRuby2000
07/15/2022, 2:37 AMWhat's the other one? Bit size or something like that? - t
TheRoyalRuby2000
07/15/2022, 2:40 AMIs it batch size or something? But that should be set to 1 - t
TheRoyalRuby2000
07/15/2022, 2:40 AMAnd learning rate to 0.0007 - t
TheRoyalRuby2000
07/15/2022, 2:40 AMThat's what I was told by Pikachu - r
Radak
07/15/2022, 4:36 AMhttps://drive.google.com/file/d/1Y8uro8oDeJ5FsatNDLIyk5gL1KPJmbw1/view?usp=drivesdk is this the right server for dataset - c
Carson
07/15/2022, 4:37 AM#835647732453605376 - g
Gosmokeless28
07/15/2022, 9:40 AMThere's literally a datasets channel bro - u
(Dawn) Will Draw Fictional Women
07/15/2022, 11:11 AMthis message is staying pinned 😎 - u
{K EY1} (Kei)
07/15/2022, 6:24 PMI use it because it uses arpabet conversion Which makes it pronounce based on phonetics instead of based on the characters in the words Arpabet training prevents the model from mispronouncing words outside of the dataset (such as some models pronouncing 'sky' as 'ski') - u
{K EY1} (Kei)
07/15/2022, 6:25 PMBatch size, and it shouldn't always be set to 1 - t
TheRoyalRuby2000
07/15/2022, 6:26 PMWell I set it to one whenever I train using Pipeline whether it's AITCH or Crust. - u
{K EY1} (Kei)
07/15/2022, 6:27 PMIf you have more than like... 2 wavs that's not necessarily the best idea - g
Gosmokeless28
07/15/2022, 6:48 PMWhy is that? - u
(Dawn) Will Draw Fictional Women
07/15/2022, 9:00 PMim gonna keep doing it - u
(Dawn) Will Draw Fictional Women
07/15/2022, 9:00 PMunpin it all you like - u
(Dawn) Will Draw Fictional Women
07/15/2022, 9:00 PMits gonna go back up - u
(Dawn) Will Draw Fictional Women
07/15/2022, 9:01 PMnow are you gonna quit coping or am i gonna have to make you - u
{K EY1} (Kei)
07/15/2022, 9:07 PMCuz it'll take forever to train - u
(Dawn) Will Draw Fictional Women
07/15/2022, 9:36 PMquit seething it aint healthy - m
magicdogstudios
07/15/2022, 9:41 PMI'm getting this error: - m
magicdogstudios
07/15/2022, 9:41 PMUnpicklingError: invalid load key, '<'. - m
magicdogstudios
07/15/2022, 10:03 PMgee, thanks for the help - u
(Dawn) Will Draw Fictional Women
07/15/2022, 10:07 PMmaybe that isnt enough context (dont know what nb it is, what step, what led to it) - u
(Dawn) Will Draw Fictional Women
07/15/2022, 10:07 PMno need to be passive aggressive about it - m
magicdogstudios
07/15/2022, 10:28 PMIt's an error a lot of people are getting and it seems the creators aren't resolving it, but some people seem to be without it - m
magicdogstudios
07/15/2022, 10:28 PM918 "functionality.") 919 --> 920 magic_number = pickle_module.load(f, **pickle_load_args) 921 if magic_number != MAGIC_NUMBER: 922 raise RuntimeError("Invalid magic number; corrupt file?") UnpicklingError: invalid load key, '<'