Open chucklesb opened 5 months ago
This looks good, appreciate the contribution! I'll take a look at it this weekend and get it setup on my wsl2 env as I don't have a native linux installation.
This looks good, appreciate the contribution! I'll take a look at it this weekend and get it setup on my wsl2 env as I don't have a native linux installation.
OK, let me know if you have trouble. I should be able to assist via Discord if needed, but as long as you have Python >= 3.11.0rc2, CUDA Toolkit 12.4, and build dependencies installed, you should be good to go. I installed Python 3.11.9 and build dependencies like this:
sudo apt update && sudo apt upgrade
sudo apt install build-essential gdb lcov pkg-config \
libbz2-dev libffi-dev libgdbm-dev libgdbm-compat-dev liblzma-dev \
libncurses5-dev libreadline6-dev libsqlite3-dev libssl-dev \
lzma lzma-dev tk-dev uuid-dev zlib1g-dev
wget https://www.python.org/ftp/python/3.11.9/Python-3.11.9.tgz
tar xf Python-3.11.9.tgz
cd Python-3.11.9
./configure
make
sudo make altinstall
CUDA Toolkit 12.4 installation:
wget https://developer.download.nvidia.com/compute/cuda/12.4.0/local_installers/cuda_12.4.0_550.54.14_linux.run
chmod +x cuda_12.4.0_550.54.14_linux.run
sudo ./cuda_12.4.0_550.54.14_linux.run --silent --toolkit
It looks like there are a couple of other things I need to address before merge:
There are other improvements/fixes we can make (with both Windows and Linux setup scripts), but I think getting the Linux script on-par with the Windows script is a prerequisite step.
Oh, and I guess I should take a look at the update script.
@chucklesb I made a few modifications to the way python 3.11 is checked as when I ran it as is without 3.11, it skipped over that section.
There are a few more things that I gotta look into, but most of it is working except for deepspeed right now where I'm getting:
nvcc fatal : Unsupported gpu architecture 'compute_89'
Might need to rebuild deepspeed, but working on figuring that out
Sounds good to me. I am not around my PC at the moment, but I'll look closer when I can. First thing to check is that your CUDA Toolkit version is correct.
On Tue, Apr 16, 2024, 3:25 PM Jarod Mica @.***> wrote:
@chucklesb https://github.com/chucklesb I made a few modifications to the way python 3.11 is checked as when I ran it as is without 3.11, it skipped over that section.
There are a few more things that I gotta look into, but most of it is working except for deepspeed right now where I'm getting:
nvcc fatal : Unsupported gpu architecture 'compute_89'
Might need to rebuild deepspeed, but working on figuring that out
— Reply to this email directly, view it on GitHub https://github.com/JarodMica/ai-voice-cloning/pull/77#issuecomment-2059946506, or unsubscribe https://github.com/notifications/unsubscribe-auth/AIDEIDMPSSRZVG53E6Q73C3Y5WJKXAVCNFSM6AAAAABF4EQSPSVHI2DSMVQWIX3LMV43OSLTON2WKQ3PNVWWK3TUHMZDANJZHE2DMNJQGY . You are receiving this because you were mentioned.Message ID: @.***>
I've modified setup-cuda.sh and created setup-cuda-cpu.sh to reflect recent changes related to Python 3.11 and dependency updates. Briefly tested on WSL2/Ubuntu 22.04. There are other changes that should be made (we can discuss further if you like, but some of the scripts should be consolidated, and some, like the docker scripts, should probably be removed entirely), but this is a good first step to re-establish Linux compatibility.
https://github.com/JarodMica/ai-voice-cloning/issues/31 https://github.com/JarodMica/ai-voice-cloning/discussions/42