Skip to content

Instantly share code, notes, and snippets.

View Clybius's full-sized avatar

Clybius

24 followers · 14 following
View GitHub Profile
@Clybius
Clybius / SnooC_CPU.py
Created October 24, 2025 21:26
SNOO: Step-K Nesterov Outer Optimizer Implementation (Clybius Variant) (CPU)
import torch
from torch.optim import Optimizer
"""
Usage: Wrap your optimizer of choice with SnooC, default params seem to work fine. Params and such are copied into SnooC's internal optimizer (SGD), which are placed on the CPU to save VRAM.
Example:
```
from SnooC_CPU import SnooC
optimizer = SnooC(torch.optim.AdamW([params], lr=0.001, weight_decay=0.01))
```
@Clybius
Clybius / XLOmniV2.4.py
Created July 16, 2024 05:02
XLOmniV2.4 Merge Recipe
import sd_mecha
from sd_mecha.hypers import Hyper
from sd_mecha.merge_methods import SameMergeSpace
from sd_mecha.extensions.merge_method import LiftFlag, convert_to_recipe
from sd_mecha.merge_space import MergeSpace
import torch
from torch import Tensor
from typing import Optional, Dict, Mapping, TypeVar
import functools
import math
@Clybius
Clybius / model_merger.py
Created February 9, 2024 17:16
ComfyUI Model Merger Custom Nodes
import torch
from tqdm.auto import trange
import tqdm
def train_difference(a, b, c, key):
a = a[key][0]
b = b[key][0]
c = c[key][0]
merged = []
atype = a.dtype
@Clybius
Clybius / README.txt
Created July 3, 2023 17:55
DPM++ 2M SDE Adaptive Sampler | A modified 2M SDE sampler complete with cosine similarity matching with a 2S step, and adaptive second order sampling.
# How to setup in A1111 Stable Diffusion and its various forks
1) You can install the sampler by adding it to the bottom of `repositories/k-diffusion/k_diffusion/sampling.py`
2) Then within `modules/sd_samplers_kdiffusion.py`, add the following to `samplers_k_diffusion`
('DPM++ 2M SDE Adaptive', 'sample_dpmpp_2m_sde_adaptive', ['c_dpmpp_2m_sde_ad'], {"brownian_noise": True, 'scheduler': 'karras', 'discard_next_to_last_sigma': True}),
Notes: This has been tested on Vlad's A1111 fork. Step 2 may not be exactly the same on native A1111, but you oughta be able to work around any errors with some contextual clues.
@Clybius
Clybius / perlin.py
Created May 17, 2023 17:08 — forked from vadimkantorov/perlin.py
Perlin noise in PyTorch
# ported from https://github.com/pvigier/perlin-numpy/blob/master/perlin2d.py
import torch
import math
def rand_perlin_2d(shape, res, fade = lambda t: 6*t**5 - 15*t**4 + 10*t**3):
delta = (res[0] / shape[0], res[1] / shape[1])
d = (shape[0] // res[0], shape[1] // res[1])
grid = torch.stack(torch.meshgrid(torch.arange(0, res[0], delta[0]), torch.arange(0, res[1], delta[1])), dim = -1) % 1
@Clybius
Clybius / sampling.py
Created April 26, 2023 16:07
DPM++ 2M LSA (Low Sigma Ancestral) Sampler (K-Diffusion)
@torch.no_grad()
def sample_dpmpp_2m_lsa(model, x, sigmas, extra_args=None, callback=None, disable=None, eta=1., s_noise=1., noise_sampler=None):
"""DPM-Solver++(2M) with ancestral noise at sigmas below 1.1."""
extra_args = {} if extra_args is None else extra_args
noise_sampler = default_noise_sampler(x) if noise_sampler is None else noise_sampler
s_in = x.new_ones([x.shape[0]])
sigma_fn = lambda t: t.neg().exp()
t_fn = lambda sigma: sigma.log().neg()
old_denoised = None
@Clybius
Clybius / index.html
Last active May 9, 2025 22:02
Embed AV1/No File Size Capped Videos in Discord
<head>
<meta property="og:image" content="GifToEmbedURL"> # Change the content to the link of a gif of your choice, which will be shown as the embed.
<meta property="og:type" content="video.other">
<meta property="og:video:url" content="VideoToEmbedURL"> # Change the content to the link of a video of your choice. Will work with videos over 50 MB, and even unsupported codecs such as AV1!
<meta property="og:video:width" content="1920"> # Set this to the video's width and height, not required, but will show the video as intended if the aspect ratio and size is correct.
<meta property="og:video:height" content="1080">
</head>