d38e560e42
1. Add an option to allow users to use UNet in fp8/gguf but lora in fp16. 2. All FP16 loras do not need patch. Others will only patch again when lora weight change. 3. FP8 unet + fp16 lora are available (somewhat only available) in Forge now. This also solves some “LoRA too subtle” problems. 4. Significantly speed up all gguf models (in Async mode) by using independent thread (CUDA stream) to compute and dequant at the same time, even when low-bit weights are already on GPU. 5. View “online lora” as a module similar to ControlLoRA so that it is moved to GPU together with model when sampling, achieving significant speedup and perfect low VRAM management simultaneously.
Stable Diffusion WebUI Forge
Stable Diffusion WebUI Forge is a platform on top of Stable Diffusion WebUI (based on Gradio 
The name "Forge" is inspired from "Minecraft Forge". This project is aimed at becoming SD WebUI's Forge.
Forge is currently based on SD-WebUI 1.10.1 at this commit. (Because original SD-WebUI is almost static now, Forge will sync with original WebUI every 90 days, or when important fixes.)
News
Aug 15: Flux BNB NF4 / GGUF Q8_0/Q5_0/Q5_1/Q4_0/Q4_1 are all natively supported with GPU weight slider and Quene/Async Swap toggle and swap location toggle.
Aug 15: All Flux BNB NF4 / GGUF Q8_0/Q5_0/Q4_0 have LoRA support.
Installing Forge
Just use this one-click installation package (with git and python included).
>>> Click Here to Download One-Click Package (CUDA 12.1 + Pytorch 2.3.1) <<<
Some other CUDA/Torch Versions:
Forge with CUDA 12.1 + Pytorch 2.3.1 <- Recommended
Forge with CUDA 12.4 + Pytorch 2.4 <- Fastest, but MSVC may be broken, xformers may not work
Forge with CUDA 12.1 + Pytorch 2.1 <- the previously used old environments
After you download, you uncompress, use update.bat to update, and use run.bat to run.
Note that running update.bat is important, otherwise you may be using a previous version with potential bugs unfixed.
Advanced Install
If you are proficient in Git and you want to install Forge as another branch of SD-WebUI, please see here. In this way, you can reuse all SD checkpoints and all extensions you installed previously in your OG SD-WebUI, but you should know what you are doing.
If you know what you are doing, you can also install Forge using same method as SD-WebUI. (Install Git, Python, Git Clone the forge repo https://github.com/lllyasviel/stable-diffusion-webui-forge.git and then run webui-user.bat).
Previous Versions
You can download previous versions here.
Forge Status
Based on manual test one-by-one:
| Component | Status | Last Test |
|---|---|---|
| Basic Diffusion | Normal | 2024 July 27 |
| GPU Memory Management System | Normal | 2024 July 27 |
| LoRAs | Normal | 2024 July 27 |
| All Preprocessors | Normal | 2024 July 27 |
| All ControlNets | Normal | 2024 July 27 |
| All IP-Adapters | Normal | 2024 July 27 |
| All Instant-IDs | Normal | 2024 July 27 |
| All Reference-only Methods | Normal | 2024 July 27 |
| All Integrated Extensions | Normal | 2024 July 27 |
| Popular Extensions (Adetailer, etc) | Normal | 2024 July 27 |
| Gradio 4 UIs | Normal | 2024 July 27 |
| Gradio 4 Forge Canvas | Normal | 2024 July 27 |
| LoRA/Checkpoint Selection UI for Gradio 4 | Normal | 2024 July 27 |
| Photopea/OpenposeEditor/etc for ControlNet | Normal | 2024 July 27 |
| Wacom 128 level touch pressure support for Canvas | Normal | 2024 July 15 |
| Microsoft Surface touch pressure support for Canvas | Broken, pending fix | 2024 July 29 |
Feel free to open issue if anything is broken and I will take a look every several days. If I do not update this "Forge Status" then it means I cannot reproduce any problem. In that case, fresh re-install should help most.
UnetPatcher
Below are self-supported single file of all codes to implement FreeU V2.
See also extension-builtin/sd_forge_freeu/scripts/forge_freeu.py:
import torch
import gradio as gr
from modules import scripts
def Fourier_filter(x, threshold, scale):
# FFT
x_freq = torch.fft.fftn(x.float(), dim=(-2, -1))
x_freq = torch.fft.fftshift(x_freq, dim=(-2, -1))
B, C, H, W = x_freq.shape
mask = torch.ones((B, C, H, W), device=x.device)
crow, ccol = H // 2, W // 2
mask[..., crow - threshold:crow + threshold, ccol - threshold:ccol + threshold] = scale
x_freq = x_freq * mask
# IFFT
x_freq = torch.fft.ifftshift(x_freq, dim=(-2, -1))
x_filtered = torch.fft.ifftn(x_freq, dim=(-2, -1)).real
return x_filtered.to(x.dtype)
def patch_freeu_v2(unet_patcher, b1, b2, s1, s2):
model_channels = unet_patcher.model.diffusion_model.config["model_channels"]
scale_dict = {model_channels * 4: (b1, s1), model_channels * 2: (b2, s2)}
on_cpu_devices = {}
def output_block_patch(h, hsp, transformer_options):
scale = scale_dict.get(h.shape[1], None)
if scale is not None:
hidden_mean = h.mean(1).unsqueeze(1)
B = hidden_mean.shape[0]
hidden_max, _ = torch.max(hidden_mean.view(B, -1), dim=-1, keepdim=True)
hidden_min, _ = torch.min(hidden_mean.view(B, -1), dim=-1, keepdim=True)
hidden_mean = (hidden_mean - hidden_min.unsqueeze(2).unsqueeze(3)) / (hidden_max - hidden_min).unsqueeze(2).unsqueeze(3)
h[:, :h.shape[1] // 2] = h[:, :h.shape[1] // 2] * ((scale[0] - 1) * hidden_mean + 1)
if hsp.device not in on_cpu_devices:
try:
hsp = Fourier_filter(hsp, threshold=1, scale=scale[1])
except:
print("Device", hsp.device, "does not support the torch.fft.")
on_cpu_devices[hsp.device] = True
hsp = Fourier_filter(hsp.cpu(), threshold=1, scale=scale[1]).to(hsp.device)
else:
hsp = Fourier_filter(hsp.cpu(), threshold=1, scale=scale[1]).to(hsp.device)
return h, hsp
m = unet_patcher.clone()
m.set_model_output_block_patch(output_block_patch)
return m
class FreeUForForge(scripts.Script):
sorting_priority = 12 # It will be the 12th item on UI.
def title(self):
return "FreeU Integrated"
def show(self, is_img2img):
# make this extension visible in both txt2img and img2img tab.
return scripts.AlwaysVisible
def ui(self, *args, **kwargs):
with gr.Accordion(open=False, label=self.title()):
freeu_enabled = gr.Checkbox(label='Enabled', value=False)
freeu_b1 = gr.Slider(label='B1', minimum=0, maximum=2, step=0.01, value=1.01)
freeu_b2 = gr.Slider(label='B2', minimum=0, maximum=2, step=0.01, value=1.02)
freeu_s1 = gr.Slider(label='S1', minimum=0, maximum=4, step=0.01, value=0.99)
freeu_s2 = gr.Slider(label='S2', minimum=0, maximum=4, step=0.01, value=0.95)
return freeu_enabled, freeu_b1, freeu_b2, freeu_s1, freeu_s2
def process_before_every_sampling(self, p, *script_args, **kwargs):
# This will be called before every sampling.
# If you use highres fix, this will be called twice.
freeu_enabled, freeu_b1, freeu_b2, freeu_s1, freeu_s2 = script_args
if not freeu_enabled:
return
unet = p.sd_model.forge_objects.unet
unet = patch_freeu_v2(unet, freeu_b1, freeu_b2, freeu_s1, freeu_s2)
p.sd_model.forge_objects.unet = unet
# Below codes will add some logs to the texts below the image outputs on UI.
# The extra_generation_params does not influence results.
p.extra_generation_params.update(dict(
freeu_enabled=freeu_enabled,
freeu_b1=freeu_b1,
freeu_b2=freeu_b2,
freeu_s1=freeu_s1,
freeu_s2=freeu_s2,
))
return
See also Forge's Unet Implementation.
Under Construction
WebUI Forge is now under some constructions, and docs / UI / functionality may change with updates.