bghira / SimpleTuner

A general fine-tuning kit geared toward Stable Diffusion 2.1, Stable Diffusion 3, DeepFloyd, and SDXL.

bghira/SimpleTuner Issues

2 Multi GPU related bugs with fix
Closed 22 days ago2
Crash due to random_dataloader_iterator not being a generator and possible fix
Closed 22 days ago2
error validating
Closed 23 days ago9
validation dtype error, cpubfloat16 vs gpu bfloat16
Closed 23 days ago8
how do I maintain the input image sizes I've manually sized?
Closed 24 days ago3
Cached latents are bigger when batch size > 1
Closed a month ago7
Unable to detect images in the dataset
Closed a month ago19
No images were discovered by the bucket manager in the dataset ( continue from #545 )
Closed a month ago1
Feature request: Convert single sdxl safetensors to diffusers format using Kohya's SD scripts.
Closed a month ago3
Install dependency triton-library: wanted 1.0.0rc3, have only <= 1.0.0rc2
Closed a month ago2
TypeError: new(): invalid data type 'str'
Closed a month ago4
Problem with `gradient_checkpointing`
Closed a month ago1
cannot import name 'random_dataloader_iterator_with_prefetch' from 'helpers.data_backend.factory'
Closed a month ago1
SD3.0 inference
Closed a month ago1
excluding specifc images ?
Closed a month ago13
UnboundLocalError: cannot access local variable 'minimum_image_size'
Closed a month ago1
SD3 LoRA Inference..?
Closed a month ago1
`(PromptHandler) Could not locate image`
Closed a month ago11
What does the JSONL dataloader look like
Closed a month ago1
Training controlnet sdxl using finetuned sdxl model.
Closed a month ago
ValueError: Your dataloader config must contain at least one image dataset AND at least one text_embed dataset.
Closed a month ago1
deep zero3
Closed a month ago2
No ability to use multiple captions per image
Closed a month ago2
Can not use vae_cache_preprocess for Training
Closed a month ago13
Have a complete end-to-end test that is extensible for testing new features
Updated a month ago
EMA: Support Karras style EMA with post-hoc weight generation with variable decay
Updated a month ago
[Feature Request] Support Hunyuan-DiT
Updated a month ago5
EMA: support side-by-side validations with base model network
Updated a month ago
SD3 cannot be trained
Closed a month ago1
Training SD3 Lora without training at least Lora on top of the Two CLIP is a big issue !
Closed a month ago2
Images missing, repeats ignored, VAECache warnings - SD3 full finetuning
Closed a month ago17
tensorboard support
Closed a month ago6
multi nodes multi gpu is support ？
Closed a month ago13
TypeError: layer_norm(): argument 'input' (position 1) must be Tensor, not tuple
Closed a month ago18
Calculation of epochs vs steps and total progress is wrong
Closed a month ago18
Training SD3 Lora underfit discuss
Closed a month ago3
parquet metadata backend has trouble with maintaining aspect ratio when using square crops
Closed a month ago2
dataset config preparation ?
Closed a month ago10
DummyOptim is not an Optimizer
Closed a month ago11
Should use vae.config.shift_factor in sd3 finetuning?
Closed a month ago1
Validation.__init__() missing 1 required positional argument: 'transformer'
Closed a month ago2
'NoneType' object has no attribute 'gradient_checkpointing_enable'
Closed a month ago3
(question)SD3 Training Memory Usage Issue
Closed a month ago6
Sigma Issue
Closed a month ago1
SD3 full finetuning LR question
Closed a month ago3
`AttributeError: 'Validation' object has no attribute 'validation_negative_prompt_mask'`
Closed a month ago29
when cropping is disabled, images may be processed incorrectly
Closed a month ago1
A lot of VAECache warnings and OOM exception when testing SD3 full finetuning with resolution 512px
Closed a month ago3
[pixart] any batch size greater than 1 leads to an error
Closed a month ago
SD3 full finetuning checkpoint fails to load
Closed a month ago5