bssrdf

South Polar Stereographic Projection not producing right plot

2

Hi, This may be a followup of a previous closed issue (https://github.com/matplotlib/basemap/issues/118). I am making polar stereographic projection pcolormesh plots of some sea ice data. It works fine with the...

Are lora residual weights critical to faithfully keep ID in the generated image?

1

HI my question is if it is necessary to apply lora residual weight from the model file during inference? In other words, will applying lora help generate images more faithfully...

fix VAE loading issue in train_dreambooth

9

# What does this PR do? This PR is similar to https://github.com/huggingface/diffusers/pull/3462 but fixes the same issue in ```train_dreambooth.py```. Fixes https://github.com/huggingface/diffusers/issues/7619 ## Before submitting - [ ] This PR fixes...

Hi, I am working on a side project off `stable-diffusion.cpp` which really needs the functionality of your clip implementation. What's the best way to use clip? Also it seems `ggml`...

[Dreambooth] number of channels error in train_dreambooth.py

17

### Describe the bug I am trying to use train_dreambooth.py to train a personalized model by following https://github.com/huggingface/diffusers/tree/main/examples/dreambooth. I got the following error: ``` File "C:\WPy64-31090\python-3.10.9.amd64\lib\site-packages\torch\nn\modules\conv.py", line 458, in _conv_forward...

Support for PhotoMaker Version 2

2

This PR added support for PhotoMaker V2 (PMV2). PMV2 requires/adds a face detection step via insightface. I couldn't find a C++ lib to do equivalent in SD. (There is a...

A specialized Winograd Conv2d op

29

This PR added a new conv2d op using Winograd algorithm. Currently ggml's conv2d operator uses im2col and GEMM. There have been efforts to speed up this process using other faster...

CUDA: add implicit conv3d

25

This PR adds an implicit conv3d op in CUDA backend, as a complement to IM2COL_3D+GEMM kernel currently used in SD.cpp for video models. It pretty much follows conv2d_implicit. Using tensor...

cuda : Add conv2d Implicit GEMM

30

This PR added another CUDA conv_2d op using implicit GEMM approach. It is only optimized for cuda cores and its performance is up to 10x of that of direct method...

I think for kernel 10 (warp tiling), additional constraints for block parameters may be needed: ``` const uint K10_NUM_THREADS = 128; const uint K10_BN = 256; const uint K10_BM =...