Rasmus Toivanen

Hi, Any more detailed instructions coming? I have been able to get this working on WSL2. Now I tried with compiling with the instructions above using Ubunt22.04 in WSL2. First...

@Topping1 sudo apt install g++-mingw-w64-x86-64-posix sudo apt install gcc-mingw-w64-x86-64-posix For me these run into error with Ubuntu 20.04 

But I did not yet test with fresh install. Trying that now

Ended up using Debian. Still got that same libwinpthread-1.dll is missing error but after putting that file in the same folder as the main.exe got it working. Windows 11, BUILD...

1.14.1 I also tried the non-custom examples and it also ran into error.  wsl version  Using Ubuntu 20.04LTS as my WSL2 distro 

Thanks @chensuyue for your continued efforts on this. I will try this again once I have time.

Any ideas? Has there someone been able to finetune the TTS model with self prepared train_manifest.json?

It is defined here https://github.com/facebookresearch/seamless_communication/tree/main/scripts/m4t/finetune

I am also looking into finetuning this for Finnish TTS using one device (RTX 4080). Is that possible? Some system info: Windows 11 Using WSL 2, Ubuntu 22.04 Cuda 11.8...

Ok, It seems it is trying to use devices 0 and 1 but I only have 1. Added some env variables to ~/.bashrc and now I see something happening ...