ilvina-12rep-16d-16a-12ep-0005lr-4bs-adamw8bit/training_config.json

{
    "pretrained_model_name_or_path": "/workspace/models/flux1-dev.safetensors",
    "ae": "/workspace/models/ae.safetensors",
    "t5xxl": "/workspace/models/t5xxl_fp16.safetensors",
    "clip_l": "/workspace/models/clip_l.safetensors",
    "network_dim": 16,
    "network_alpha": 16,
    "train_batch_size": 4,
    "optimizer_type": "Adamw8bit",
    "unet_lr": 0.0005,
    "epoch": 12,
    "max_train_steps": 1250,
    "apply_t5_attn_mask": false,
    "cache_latents": true,
    "cache_latents_to_disk": true,
    "cache_text_encoder_outputs": true,
    "cache_text_encoder_outputs_to_disk": true,
    "clip_skip": 1,
    "discrete_flow_shift": 3.1582,
    "full_bf16": true,
    "mixed_precision": "bf16",
    "gradient_accumulation_steps": 1,
    "gradient_checkpointing": true,
    "guidance_scale": 1.0,
    "highvram": true,
    "huber_c": 0.1,
    "huber_schedule": "snr",
    "loss_type": "l2",
    "lr_scheduler": "cosine_with_restarts",
    "lr_scheduler_args": [],
    "lr_scheduler_num_cycles": 3,
    "lr_scheduler_power": 1,
    "max_data_loader_n_workers": 0,
    "max_grad_norm": 1,
    "max_timestep": 1000,
    "min_snr_gamma": 5,
    "model_prediction_type": "raw",
    "network_args": [
        "train_double_block_indices=all",
        "train_single_block_indices=all"
    ],
    "network_module": "networks.lora_flux",
    "network_train_unet_only": true,
    "noise_offset": 0.1,
    "noise_offset_type": "Original",
    "optimizer_args": [],
    "prior_loss_weight": 1,
    "sample_sampler": "euler",
    "sdpa": true,
    "seed": 42,
    "t5xxl_max_token_length": 512,
    "text_encoder_lr": [],
    "timestep_sampling": "sigmoid"
}