# machine-learning
- u
user
05/05/2021, 1:38 PMSorry if I sound dumb but how do you do this? - p
pi
05/05/2021, 1:39 PMas a parameter when loading the model - u
user
05/05/2021, 1:39 PMso here? - p
pi
05/05/2021, 1:39 PMgoogle forpytorch fp16 inference - u
user
05/05/2021, 1:44 PMModuleNotFoundError: No module named 'tacotron2.layers' - p
pi
05/05/2021, 1:46 PM@User you could also do cpu inference - u
user
05/05/2021, 1:46 PMooh - u
user
05/05/2021, 1:46 PMnice - p
pi
05/05/2021, 1:46 PMplease read the readme 🤦 - u
user
05/05/2021, 1:46 PMoh wait - u
user
05/05/2021, 1:46 PMI only have tpu and gpu - p
pi
05/05/2021, 1:46 PMyou can set it to none - u
user
05/05/2021, 1:47 PMah - p
pi
05/05/2021, 1:47 PMand you need to remove the model.cuda() stuff - u
user
05/05/2021, 1:49 PMso I just remove this? - p
pi
05/05/2021, 1:50 PMonly the .cuda() - u
user
05/05/2021, 1:50 PMthis makes no senseCopy codeexport PYTHONPATH="$PYTHONPATH:/path/to/flask-tacotron2-tts-web-app" - p
pi
05/05/2021, 1:51 PMwhy? - u
user
05/05/2021, 1:51 PMwhat should i do with that - p
pi
05/05/2021, 1:51 PMplease don't - p
pi
05/05/2021, 1:51 PMyou'll be pwned in no time - u
user
05/05/2021, 1:52 PMIt did not like that - p
pi
05/05/2021, 1:52 PMbut it tells you the fix - u
user
05/05/2021, 1:52 PMohh - u
user
05/05/2021, 1:52 PMyeah - u
user
05/05/2021, 1:52 PMsorry - u
user
05/05/2021, 1:52 PMI read it - p
pi
05/05/2021, 1:53 PMI don't have a colab in front of me right now - p
pi
05/05/2021, 1:53 PMso I can't check - p
pi
05/05/2021, 1:54 PMit's an environment variable