Embeddings don't work with stable-diffusion-webui with sdxl right now. hopefully fixed soon.
This embedding works with ComfyUI put into ComfyUI/models/embeddings/
May or may not need the trigger word depending on the version of ComfyUI your using. Latest version no longer needs the trigger word for me.
With trigger word, old version of comfyui
"embedding:hannahferguson hannahferguson a woman and so on ...."
without trigger word
"embedding:hannahferguson a woman and so on ...."
Description
FAQ
Comments (14)
Heads up, in your examples you typo'd "embedding" to "embeddibg", I am positive it messed with your resulting preview images.
thank you for catching that. the prompt is wrong. The been fooling around with it some more. I have to re-post the images. Need to use the trigger word. "embedding:hannahferguson hannahferguson a woman and so on ...." embedding:filename triggerword prompt ...." . I am going to delete and repost the model. I am new to comfyui and still figuring out the sdxl embeddings.
@summer4love263 I'm new to ComfyUI as well, didn't realize you had to repeat the trigger word, thanks for the info!
@BilboTaggins I tried with embeddings trained with stable-diffusion-webui with 1.5 and you don't have to repeat the trigger. Its the way I trained it with kohya_ss. I have to see if I can get rid of that, because it is annoying. Normally I train TI with stable-diffusion-webui. Also saves them as a safetensor file instead of a pt file, so could be a different format and you need the trigger like with a lora. Not totally sure.
@BilboTaggins Updated my ComyUI and no longer need to put in the trigger word to make it work. Not sure whats going on, looks like a bug, because it didn't work for either of us and now it works for me. Update when you get a chance.
@summer4love263 Ah, thanks for the heads up!
Good to see SDXL get some Embeddings! :D
what code did you use to train this?
https://github.com/bmaltais/kohya_ss
I can share my setting if you need them.
@summer4love263 that would be great, I tried to train with the same code but I had an init word error
@nicolai256
uploaded the config file to a github page.
Your going to have to change some setting to make sure its pointing to the right places.
Can you share settings used to train this?
These are the setting from the json file. Not for this model but the last one I did. You should be able to cut and paste into a new json file and load it and change what you need.
{
"adaptive_noise_scale": 0,
"additional_parameters": "",
"bucket_no_upscale": false,
"bucket_reso_steps": 32,
"cache_latents": true,
"cache_latents_to_disk": true,
"caption_dropout_every_n_epochs": 0.0,
"caption_dropout_rate": 0,
"caption_extension": ".txt",
"clip_skip": "1",
"color_aug": false,
"enable_bucket": true,
"epoch": 60,
"flip_aug": false,
"full_fp16": false,
"gradient_accumulation_steps": 1,
"gradient_checkpointing": false,
"init_word": "woman 1girl",
"keep_tokens": "0",
"learning_rate": 0.005,
"logging_dir": "/home/johnb/kohya_ss/lora/logs",
"lr_scheduler": "constant",
"lr_scheduler_num_cycles": "",
"lr_scheduler_power": "",
"lr_warmup": 1,
"max_bucket_reso": 4096,
"max_data_loader_n_workers": "0",
"max_resolution": "1024,1024",
"max_timestep": 1000,
"max_token_length": "75",
"max_train_epochs": "",
"max_train_steps": "",
"mem_eff_attn": false,
"min_bucket_reso": 256,
"min_snr_gamma": 0,
"min_timestep": 0,
"mixed_precision": "fp16",
"model_list": "custom",
"multires_noise_discount": 0,
"multires_noise_iterations": 0,
"no_token_padding": false,
"noise_offset": 0,
"noise_offset_type": "Original",
"num_cpu_threads_per_process": 2,
"num_vectors_per_token": 6,
"optimizer": "Adafactor",
"optimizer_args": "",
"output_dir": "/home/johnb/stable-diffusion-webui/embeddings",
"output_name": "GiseleB\u00fcndchen",
"persistent_data_loader_workers": false,
"pretrained_model_name_or_path": "/home/johnb/sd_xl_base_1.0.safetensors",
"prior_loss_weight": 1.0,
"random_crop": false,
"reg_data_dir": "",
"resume": "",
"sample_every_n_epochs": 0,
"sample_every_n_steps": 0,
"sample_prompts": "",
"sample_sampler": "euler_a",
"save_every_n_epochs": 2,
"save_every_n_steps": 0,
"save_last_n_steps": 0,
"save_last_n_steps_state": 0,
"save_model_as": "safetensors",
"save_precision": "fp16",
"save_state": false,
"scale_v_pred_loss_like_noise_pred": false,
"sdxl": true,
"sdxl_no_half_vae": true,
"seed": "24356754",
"shuffle_caption": false,
"stop_text_encoder_training": 0,
"template": "caption",
"token_string": "GiseleB\u00fcndchen",
"train_batch_size": 1,
"train_data_dir": "/home/johnb/kohya_ss/lora/img",
"use_wandb": false,
"v2": false,
"v_parameterization": false,
"vae": "",
"vae_batch_size": 0,
"wandb_api_key": "",
"weights": "",
"xformers": "xformers"
}
@summer4love263 thank you, I will try
Looks like we don't have an active mirror for this file right now.
CivArchive is a community-maintained index — we catalog mirrors that volunteers upload to HuggingFace, torrents, and other public hosts. Looks like no one has uploaded a copy of this file yet.
Some files do get recovered over time through contributions. If you're looking for this one, feel free to ask in Discord, or help preserve it if you have a copy.