#Help the chinese RVC dev solve gptsovits pip install issue

1 messages · Page 1 of 1 (latest)

unborn charm
#

Existing problem: cannot using pip install -r requirements to install gpt-sovits repo

#

Debug attempt 1: did pip install cmake first then pip install -r requirements

#

result: failed

hasty harness
#

how is it failing

unborn charm
#

pyopenjtalk

hasty harness
#

oh i was getting ssl errors with torch but nothing with openjtalk

unborn charm
#

send log

#

i will send him

unborn charm
unborn charm
#

problem solved, need visual studio 2017 ( i installed the enterprise version), vscode 2022 failed because problems with pyopenjtalk

glass pollen
#

was it possibly because of fairseq?

#

afaik you need visual studio c++ tools to install that

unborn charm
#

thats the dev

unborn charm
#

ok he forked this person's solution, which is more convenient

shut sapphire
#

bc thats linux only, forcibly using conda, and, assuming the end user is using bash

#

which, most people use bash

#

but, it is super stupid to just not generically call a shell

unborn charm
#

whats ur take

shut sapphire
#

conda is dumb

#

to me

#

conda is essentially third party, download, install, it stores environment somewhere else

unborn charm
#

i have to install vs 2017

shut sapphire
#

to use the prezip?

unborn charm
#

no

#

pip install

shut sapphire
#

im talking about that linux quick install

unborn charm
#

prezip is fine for my computer

shut sapphire
#

above

#

the amount of people that force conda

#

is cringe

#

virtual environments, are native to python

unborn charm
#

conda is meta rn

shut sapphire
#

meta these nuts

unborn charm
#

every project i see use conda

shut sapphire
#

i hate conda

#

all it takes

#

is one fucking dumb mistake

#

of not activating

#

or wrong env

#

and congrats, globally install

#

whereas with a venv

#

you have the script activate the venv

#

launch

unborn charm
#

lol venv

shut sapphire
#

or activate, git pull, install reqs

unborn charm
#

i just paste your venv tensorboard bat to every python machine learning repo and it works every time

shut sapphire
#

venvs are nice because, one, its first party native to python

#

two, the venv is also located in the project folder

unborn charm
#

yeah fax

shut sapphire
#

three, you can automate it, and, not have the end user fuck up

#

a prime example

#

is automatic-1111's stable diffusion webui

#

name one person that has accidentally installed SD to their global python install

#

not a single person

#

because the venv is automated

unborn charm
#

i need to learn software dev real quick

#

to an extent

#

otherwise ik theory alot not practical

shut sapphire
#

the best theory u can have as a dev

#

assume everyone is dumb, and make ur life simpler and easier to fix issues when people run into them

#

not that people ARE dumb

#

but if you take the scissors away from the toddler, and lock them up

#

you no longer have to worry about the toddler having scissors, or running with scissors

unborn charm
#

yeah

#

ty

#

so he should rewrite the thing from venv?

#

instead of conda

serene mango
#

@shut sapphire how would you deal with multiple versions of python when not using conda?

shut sapphire
#

ik 3.11 had tiny issues for SD, but, ive had zero issues with 3.10

#

i understand conda can force certain versions of python envs, i just have disdain to conda because all conda using devs

#

just go "activate ur conda, then run this"

#

instead of making a script to auto create the conda env, and activate it, and self install

serene mango
#

running conda activate <env> every time does indeed get annoying

shut sapphire
#

which again

#

can be in the start script

#

you can literally have the start script, activate conda, if the conda env fails, then creates one, installs reqs

#

etc

#

then at the end, activate the project, but no, everyone wants u to do it manually

serene mango
#

xD

shut sapphire
#

like i get, people wanna name their envs themselves, but, me personally, i dont care to, but, i have made the mistake of installing globally once, bc forgot conda