This is the first LoRA for SD XL Distilled by Segmind. A nice selection of images that will help SSD-1B achieve higher visual performance, and prompt adherence.
SD XL Distilled is fully compatible with deforum, comfy, and diffusers.
A modded Kohya sd-scripts was used for the training, the code for training should be available in a few days.
Description
FAQ
Comments (2)
Try my fork of kohya if you intend to train ssd-1b:
https://github.com/XmYx/kohya_ss/
Achref_ArtsDec 3, 2023
You have disabled the safety checker for <class 'diffusers.pipelines.stable_diffusion.pipeline_stable_diffusion.StableDiffusionPipeline'> by passing `safety_checker=None`. Ensure that you abide to the conditions of the Stable Diffusion license and do not expose unfiltered results in services or applications open to the public. Both the diffusers team and Hugging Face strongly recommend to keep the safety filter enabled in all public facing circumstances, disabling it only for use-cases that involve analyzing network behavior or auditing its results. For more information, please have a look at https://github.com/huggingface/diffusers/pull/254 .
UNet2DConditionModel: 128, [5, 10, 20], 2048, True, None
Traceback (most recent call last):
File "/content/kohya_ss/./train_network.py", line 1009, in <module>
trainer.train(args)
File "/content/kohya_ss/./train_network.py", line 224, in train
model_version, text_encoder, vae, unet = self.load_target_model(args, weight_dtype, accelerator)
File "/content/kohya_ss/./train_network.py", line 101, in load_target_model
text_encoder, vae, unet, _ = train_util.load_target_model(args, weight_dtype, accelerator)
File "/content/kohya_ss/library/train_util.py", line 3874, in load_target_model
text_encoder, vae, unet, load_stable_diffusion_format = _load_target_model(
File "/content/kohya_ss/library/train_util.py", line 3838, in _load_target_model
original_unet = UNet2DConditionModel(
File "/content/kohya_ss/library/original_unet.py", line 1365, in __init__
attn_num_head_channels=attention_head_dim[i],
IndexError: list index out of range
Traceback (most recent call last):
File "/usr/local/bin/accelerate", line 8, in <module>
sys.exit(main())
File "/usr/local/lib/python3.10/dist-packages/accelerate/commands/accelerate_cli.py", line 47, in main
args.func(args)
File "/usr/local/lib/python3.10/dist-packages/accelerate/commands/launch.py", line 986, in launch_command
simple_launcher(args)
File "/usr/local/lib/python3.10/dist-packages/accelerate/commands/launch.py", line 628, in simple_launcher
raise subprocess.CalledProcessError(returncode=process.returncode, cmd=cmd)
subprocess.CalledProcessError: Command '['/usr/bin/python3', './train_network.py', '--enable_bucket', '--min_bucket_reso=256', '--max_bucket_reso=2048', '--pretrained_model_name_or_path=segmind/SSD-1B', '--train_data_dir=/content/drive/MyDrive/data sdxl 1024p/img', '--resolution=512,512', '--output_dir=/content/drive/MyDrive/data sdxl 1024p/model', '--network_alpha=128', '--save_model_as=safetensors', '--network_module=networks.lora', '--text_encoder_lr=5e-05', '--unet_lr=0.0001', '--network_dim=128', '--output_name=ssd-1b-lora', '--lr_scheduler_num_cycles=1', '--no_half_vae', '--learning_rate=0.0001', '--lr_scheduler=adafactor', '--train_batch_size=1', '--max_train_steps=2000', '--save_every_n_epochs=1', '--mixed_precision=fp16', '--save_precision=fp16', '--cache_latents', '--optimizer_type=Adafactor', '--max_data_loader_n_workers=0', '--bucket_reso_steps=64', '--mem_eff_attn', '--gradient_checkpointing', '--full_fp16', '--sdpa', '--bucket_no_upscale', '--noise_offset=0.0', '--lowram']' returned non-zero exit status 1.