sd-scripts

sd-scripts copied to clipboard

step_offset for dpmsolver++ sampler now crashes

2

There has been this deprecation error for stable diffusion 1.5 sampling scheduler. But in the latest versions in requirements has made this crash instead. ``` /mnt/900/builds/sd-scripts/.venv/lib/python3.10/site-packages/diffusers/pipelines/stable_diffusion/pipeline_stable_diffusion.py:182: FutureWarning: The configuration file...

rockerBOO

Help to fix the code while running in the colab

1

2024-04-25 10:32:01.418766: W tensorflow/compiler/tf2tensorrt/utils/py_utils.cc:38] TF-TRT Warning: Could not find TensorRT ╭─────────────────────────────── Traceback (most recent call last) ────────────────────────────────╮ │ /content/kohya-trainer/finetune/make_captions.py:14 in │ │ │ │ 11 from torchvision import transforms │...

ImamaAI

I opened a similar issue with bmaltais on his repo, but for whatever reason I can't get the training to start any longer. If enabled, caching latents gets stuck at...

machineminded

`fix_noise_scheduler_betas_for_zero_terminal_snr` should come before `prepare_scheduler_for_custom_training`

13

I think the correct implementation should be like this. ```python noise_scheduler = DDPMScheduler( beta_start=0.00085, beta_end=0.012, beta_schedule="scaled_linear", num_train_timesteps=1000, clip_sample=False ) if args.zero_terminal_snr: custom_train_functions.fix_noise_scheduler_betas_for_zero_terminal_snr(noise_scheduler) prepare_scheduler_for_custom_training(noise_scheduler, accelerator.device) ``` The current implementation calculates `noise_scheduler.all_snr`...

jihnenglin

How to use more GPUs for training under Windows? When clicking enable multi-GPU and setting up 2 GPU processes, it always shows nccl error. But when I searched for nccl,...

littleyeson

DeepSpeed Accelerator can NOT save optimizer state

3

Same as title. ~I'm in investigating.~ With following modification, scripts can save optimizer state **only for small dataset**, _IDK why it does not work big dataset._ Error messages will be...

BootsofLagrangian

Stable cascade stage b training

9

When will stage b training code be released? Thanks!!

riolys

Training Ultra Slow with more than 1 GPU - very likely affecting all users with more than 1 GPU

1

Hi there, i am running on 4x RTX4090 and as soon as i use more than 1 GPU the training gery super slow with the newer scripts starting from 22.x.x...

SteVoit

Some RGBA images cause broken backgrounds due to incorrect RGB conversion

4

Currently, the `load_image` function in `train_utils` uses the PIL `image.convert("RGB")` method for converting the loaded 4 RGBA channel images to the expected 3 channel RGB format. This is incorrect for...

city96

How multi-gpu latent caching works?

4

Currently I am trainning on Kaggle with 15 train images and 100 repeat and thus 1500 reg images Actually ideally I would like each GPU to get different set of...

FurkanGozukara