78Alpha
78Alpha
Yeah, that's what it should look like. All graphs filled is the sign of all parts working.
> @yl4579 is this repository still maintained? Most advancement is on the dagshub repo. This appears to be a research only repo.
Probably when it can be run with ease on a majority of computers. So not in the near future.
That is one of the known issues. - Black images - No dynamic shape by default (limited to the shape you setup with) - Dynamic shapes have limits (the combination...
It should, in they're, override the model you select with the manually selected TRT UNET. So the bully in VAE of the model should work as normal and any extra...
> update to dev commit The new 1.5 should work as well
As stated in the README, TensorRT does not work with controlnet. CKPT and SAFETENSOR can be altered on the fly without issue. TensorRT isn't as graceful. What you gain in...
For the most humanistic voice, this (used for tortoise) is the best you can get and properly train. For something that can just clone your voice and sound like text...
Seems pretty standard. A few dozen model trainings and stuff like that is always present. Sometime a full second. It could be that whatever dataset is being used is being...