I managed to build a wheel for windows using deepspeed-0.11.1 for Python 3.10 and torch 2.1. 0.11.1 allows me to change models without having to restart. Performance is pretty decent but for my…
After more testing some I seem to hit OOM quite often. Think I might try WSL
The reason for the logs was to demonstrate how long it took between each iteration using different drivers using the same data. You can see the timestamps between the iterations are much smaller…
Thanks for the reply. I think you're right, I was checking dedicated GPU Memory and Shared GPU memory while using the newer drivers and Shared GPU memory seemed to be much much higher than I…
Run update.bat, wait for it to finish and then run start.bat again. The issue above was fixed in commit 805d7d35e8
All confirmed working, this is turning out great. Thank you
Thank you for doing this... There's one line in utils.py line 3480 missing the :
if autoregressive_model_path == "auto": autoregressive_model_path = deduce_autoregressive_model() else …
Thank you, it's always nice to have the extra options.
I've generated with all new options deleted cond_latents_d1f79232.pth each time(not sure if it was necessary) and they all sound good. So…
I copied a new folder with the wav files again, this time trying voice chunks of 1 rather than 5 this made him sound lower again. 10 was better but not quite there. 16 was similar.
I'll try…
I've just copied the voice latents file from vanilla tortoise-tts. I had to rename the file to cond_latents_d1f79232.pth before it would work. The sound is much much better, it actually sounds…
Sorry I should add, i've recorded 5 wave files for a new voice called Don. If it's not something like the sample rate, something is making it sound more generic, there are hints of his voice, but…