Trigger not needed but
illustration_by_omone_hokoma_agm
was added to every caption and does seem to make the effect stronger but tends to add agm signature a lot more so id recommend if you use the trigger word adding signature,watermark,artist_name in the negative seems to get rid of it, ill update this though after im done editing all the dataset's images by manually removing each signature using photoshop and then train on that dataset so no extra negative words will be needed
Here's The Parameters
Training info▼
Most frequent tags in captions
illustration_by_omone_hokoma_agm
100%
Dataset folder structure
Name
40_agm
Image Count
595
Repeats
40
Total Images
23800
(Total)
595
Image Count
595
Repeats
40
Total Images
23800
Training Parameters
{
"ss_adaptive_noise_scale": "None",
"ss_caption_dropout_rate": "0.0",
"ss_steps": "8500",
"ss_noise_offset": "None",
"ss_sd_scripts_commit_hash": "15dd0a638af86f89dd0c457428e165598d4884a2",
"ss_num_batches_per_epoch": "23800",
"ss_color_aug": "False",
"ss_epoch": "0",
"ss_total_batch_size": "1",
"ss_network_alpha": "64.0",
"ss_ip_noise_gamma": "None",
"ss_num_epochs": "1",
"ss_session_id": "3173715531",
"ss_network_dim": "64",
"ss_keep_tokens": "0",
"ss_learning_rate": "1.0",
"ss_new_sd_model_hash": "e6bb9ea85bbf7bf6478a7c6d18b71246f22e95d41bcdd80ed40aa212c33cfeff",
"ss_lr_warmup_steps": "0",
"ss_optimizer": "prodigyopt.prodigy.Prodigy",
"ss_caption_dropout_every_n_epochs": "0",
"ss_network_module": "lycoris.kohya",
"ss_reg_dataset_dirs": "{}",
"ss_sd_model_hash": "be9edd61",
"ss_gradient_accumulation_steps": "1",
"ss_bucket_no_upscale": "False",
"ss_bucket_info": "null",
"ss_full_fp16": "False",
"ss_mixed_precision": "fp16",
"ss_network_dropout": "0.0",
"ss_gradient_checkpointing": "True",
"ss_random_crop": "False",
"ss_prior_loss_weight": "1.0",
"ss_max_grad_norm": "1.0",
"ss_max_bucket_reso": "None",
"ss_training_comment": "None",
"ss_num_reg_images": "0",
"ss_max_train_steps": "10000",
"ss_min_snr_gamma": "10.0",
"ss_num_train_images": "23800",
"ss_network_args": "{\"conv_dim\": \"64\", \"conv_alpha\": \"64\", \"factor\": \"-1\", \"use_cp\": \"True\", \"algo\": \"lokr\", \"dropout\": 0.0}",
"ss_shuffle_caption": "False",
"ss_unet_lr": "1.0",
"ss_resolution": "(1024, 1024)",
"ss_batch_size_per_device": "1",
"ss_multires_noise_discount": "0.2",
"ss_flip_aug": "False",
"ss_text_encoder_lr": "1.0",
"ss_lr_scheduler": "constant",
"ss_min_bucket_reso": "None",
"ss_zero_terminal_snr": "False",
"ss_lowram": "False",
"ss_seed": "12345",
"ss_multires_noise_iterations": "6",
"ss_base_model_version": "sdxl_base_v1-0",
"ss_enable_bucket": "False",
"ss_training_started_at": "1708377854.7039077",
"ss_clip_skip": "None",
"ss_v2": "False",
"ss_caption_tag_dropout_rate": "0.0",
"ss_max_token_length": "225",
"ss_output_name": "AGMXL",
"ss_scale_weight_norms": "None",
"ss_training_finished_at": "1708441002.8566256",
"ss_sd_model_name": "sdXL_v10VAEFix.safetensors",
"ss_cache_latents": "True",
"ss_face_crop_aug_range": "None"
}
This is my 2nd Lora, first was a character Lora so I figured I'd try a style Lora and haven't seen any SDXL Lora's of AGM's styles like there are for SD1.5..Like my first Lora im uploading the exact training Data I used so if those with more experience than me want to make it better having the dataset should make it easier. I used a caption for each image but in my testing it doesn't seem to make much of a difference so no trigger word is needed. Please leave a rating and comment if this works for you, feedback is very important to me, thanks!