sd-scripts
sd-scripts copied to clipboard
kohya-ss
I tried to train a LoRa model on Kohya SS, but immediately afterwards it stopped, and this error appeared.I checked my GPU and system settings and they seems okay.I have...
**Description:** This PR introduces the functionality of token decay during the training process, allowing for a dynamic adjustment of token length as training progresses. Two new arguments, `--token_decay_min` and `--token_decay_step`,...
Adds [CAME: Confidence-guided Adaptive Memory Efficient Optimization](https://arxiv.org/abs/2307.02047). Requires you to `pip install pytorch-came`. Github: https://github.com/yangluo7/CAME
https://optimi.benjaminwarner.dev/ New optimizer: adamw、lion、ranger and stableadamw from Optimi New function: Low Precision Training with Kahan Summation(auto use when use above optimizers)  Gradient Release(same as fused background pass) Auto use...
Bumps [crate-ci/typos](https://github.com/crate-ci/typos) from 1.19.0 to 1.22.9. Release notes Sourced from crate-ci/typos's releases. v1.22.9 [1.22.9] - 2024-06-22 Fixes Stop correcting reoccurrence v1.22.8 [1.22.8] - 2024-06-21 Features (action) Add Arm, Mac support...
![dependabot[bot] avatar](https://avatars.githubusercontent.com/in/29110?v=4 "dependabot[bot] avatar"){kind=avatar}
I encountered the following error while training SD1.5 ControlNet using train_controlnet.py: “RuntimeError: Expected is_sm80 || is_sm90 to be true, but got false. (Could this error message be improved? If so,...
# When I was training model on 2x A100 80G machine, a few time later afrer start, there's an error occurred: ```steps: 0%| | 373/381280 [1:58:14
I saw someone post on reddit that they successfully added it to the training side, but did not say how or what the result was. And have not answered when...
Does caption dropout make sense at high noise levels? At low noise levels caption dropout might make a lot of sense because it should be possible to infer the correct...
``` import argparse import csv import os from pathlib import Path import cv2 import numpy as np import torch from huggingface_hub import hf_hub_download from PIL import Image from tqdm import...
