⁉️ How to prompt to select objects (masking) in an image and replace with new items from the prompt
⁉️ How to prompt to select objects (masking) in an image and replace with new items from the prompt using the Comfy UI? Any existing workflow with which models to use? (Using SDXL)
17 Replies
i liked the one you created , love to know how you created for the dreambooth example
Updated my model today
Character - Yae Miko/八重神子 - Genshin Impact (SDXL0.9) - v2.0 | Stabl...
v2 is able to cosplay. I am stay in the SDXL0.9 because of this reason: https://civitai.com/articles/1444/the-humanoid-character-lora-compression-b...
i think your question is confused
what you want to do exactly
I didn't check out the SD community a while, is it SDXL1.0 is better then SDXL0.9 now?
i was watching your dreambooth video on how to create ckpt from the model , but you mentioned about created a yaml file too. I would love to know how you created it? I hope thats ok to share?
those yamls are automatically generated by the dreambooth training extension
ok, i guess watch latest videos then
@Dr. Furkan Gözükara while going over Stability AI documentation i observed VAE for SDXL-1.0 to be in saftensortensor form instead of ".pt" or .".bin". How to convert in the correct format before using refiners within SDXL?(reference: https://huggingface.co/stabilityai/stable-diffusion-xl-refiner-1.0/tree/main)
was looking into https://github.com/huggingface/diffusers/blob/main/scripts/convert_vae_pt_to_diffusers.py but was not sure
GitHub
diffusers/scripts/convert_vae_pt_to_diffusers.py at main · huggingf...
🤗 Diffusers: State-of-the-art diffusion models for image and audio generation in PyTorch - huggingface/diffusers
have you tried kohya converters?
yes, i am still new to it.. been stuck with HF.. will explore it
I setup a new pod and getting this error:
CalledProcessError: Command '['/workspace/kohya_ss/venv/bin/python', './sdxl_train_network.py', '--enable_bucket', '--min_bucket_reso=256', '--max_bucket_reso=2048',
'--pretrained_model_name_or_path=/workspace/sd_xl_base_1.0.safetensors', '--train_data_dir=/workspace/training/LoRA/LaraClassicXL/img', '--reg_data_dir=/workspace/training/LoRA/LaraClassicXL/reg', '--resolution=1024,1024',
'--output_dir=/workspace/training/LoRA/LaraClassicXL/model', '--logging_dir=/workspace/training/LoRA/LaraClassicXL/log', '--network_alpha=1', '--save_model_as=safetensors', '--network_module=networks.lora',
'--text_encoder_lr=5e-05', '--unet_lr=0.0001', '--network_dim=32', '--output_name=LaraClassicXL', '--lr_scheduler_num_cycles=1', '--no_half_vae', '--learning_rate=0.0001', '--lr_scheduler=cosine', '--lr_warmup_steps=244',
'--train_batch_size=1', '--max_train_steps=2440', '--save_every_n_epochs=1', '--mixed_precision=fp16', '--save_precision=fp16', '--cache_latents', '--optimizer_type=AdamW8bit', '--max_data_loader_n_workers=0',
'--bucket_reso_steps=64', '--xformers', '--bucket_no_upscale', '--noise_offset=0.0']' died with <Signals.SIGKILL: 9>.
let me connect your pc. please open anydesk and private message me
link?
install kohya