AbstractPhil's picture
Config @ step 22000
00e81c8 verified
{
"use_illustrious_clips": true,
"illustrious_repo": "AbstractPhil/clips",
"clip_l_filename": "IllustriousV01_clip_l.safetensors",
"clip_g_filename": "IllustriousV01_clip_g.safetensors",
"auto_discover_clips": true,
"clip_skip": 2,
"modality_dims": {
"clip_l": 768,
"clip_g": 1280,
"t5_xl_l": 2048,
"t5_xl_g": 2048
},
"modality_seq_lens": {
"clip_l": 77,
"clip_g": 77,
"t5_xl_l": 512,
"t5_xl_g": 512
},
"binding_config": {
"clip_l": {
"t5_xl_l": 0.3
},
"clip_g": {
"t5_xl_g": 0.3
},
"t5_xl_l": {},
"t5_xl_g": {}
},
"latent_dim": 2048,
"seq_len": 77,
"encoder_layers": 3,
"decoder_layers": 3,
"hidden_dim": 2048,
"dropout": 0.1,
"fusion_strategy": "adaptive_cantor",
"fusion_heads": 8,
"fusion_dropout": 0.1,
"cantor_depth": 8,
"cantor_local_window": 3,
"alpha_init": 1.0,
"beta_init": 0.3,
"alpha_lr_scale": 0.1,
"beta_lr_scale": 1.0,
"beta_kl": 0.1,
"beta_reconstruction": 1.0,
"beta_cross_modal": 0.0,
"beta_alpha_regularization": 0.01,
"recon_type": "mse",
"modality_recon_weights": {
"clip_l": 1.0,
"clip_g": 1.0,
"t5_xl_l": 0.25,
"t5_xl_g": 0.25
},
"use_kl_annealing": true,
"kl_anneal_epochs": 10,
"kl_start_beta": 0.0,
"batch_size": 32,
"num_epochs": 100,
"learning_rate": 0.0001,
"weight_decay": 1e-05,
"gradient_clip": 1.0,
"use_scheduler": true,
"scheduler_type": "cosine",
"num_samples": 100000,
"prompt_source": "booru",
"booru_ratio": 0.7,
"synthetic_ratio": 0.15,
"laion_ratio": 0.15,
"danbooru_csv": "drive/MyDrive/booru_csvs/danbooru.csv",
"gelbooru_csv": null,
"e621_csv": null,
"rule34x_csv": null,
"use_coherent_gender": true,
"generate_t5_prefix": true,
"use_conduit": true,
"conduit_top_n": 1000,
"conduit_sample_k": 10,
"conduit_sample_k_min": 5,
"conduit_sample_k_max": 15,
"conduit_position": "prepend",
"conduit_exclude_categories": [
"artist",
"copyright",
"character",
"metadata"
],
"use_summarizer": true,
"summarizer_model": "qwen2.5-1.5b",
"summarizer_batch_size": 256,
"summary_separator": "\u00b6",
"shuffle_tags_before_summary": true,
"summarizer_max_new_tokens": 96,
"summarizer_temperature": 0.7,
"use_summarizer_int8": false,
"prompt_cache_dir": "./prompt_cache",
"prompt_cache_name": null,
"use_prompt_cache": true,
"save_prompt_cache": true,
"checkpoint_dir": "./checkpoints_lyra_illustrious_v2",
"save_every": 1000,
"keep_last_n": 3,
"model_name": "lyra_illustrious",
"resume_optimizer": true,
"hf_repo": "AbstractPhil/vae-lyra-xl-adaptive-cantor-illustrious",
"hub_checkpoint_file": "checkpoint_lyra_illustrious_21000.pt",
"push_to_hub": true,
"push_every": 1000,
"push_checkpoints": true,
"auto_load_from_hub": true,
"export_safetensors": true,
"weights_subdir": "weights",
"use_wandb": false,
"wandb_project": "vae-lyra-illustrious",
"wandb_entity": null,
"log_every": 5,
"device": "cuda",
"mixed_precision": true,
"seed": 42,
"num_workers": 0
}