Dan Wood

[Feature Request]: Support Torch 2.0 now that it is GA.

31

### Is there an existing issue for this? - [X] I have searched the existing issues and checked the recent builds/commits ### What would your feature do ? Either switch...

[Feature Request]: The greatest feature request in history.

4

### Is there an existing issue for this? - [X] I have searched the existing issues and checked the recent builds/commits ### What would your feature do ? Please, for...

## Description Stable Diffusion models usually are fp32 or fp16. So are TensorRT accelerated models. To get from SD to TRT one goes through ONNX. While most float weights appear...

I tried to do inference but got the above error. I'm using python3.10 on Ubuntu 22.04. i9-13900K + 4090 I fixed this by changing return cls(**checkpoint_dict) to return cls(**vars(checkpoint_dict)) ```...

[Feature]: model cache stored in VRam

1

### Feature description There is an existing model cache that stores models in system memory. Even with this the time to unload the current one and load the other from...

I got to 294 images per second with sdxs 512 but it would be nice to play with 1024 to see how fast i can push it. [https://x.com/Dan50412374/status/1772832044848169229](https://x.com/Dan50412374/status/1772832044848169229)