{
  "conditioning_dim": 768,
  "max_condition_tokens": 256,
  "prefer_hidden_layer": -1,
  "use_native_embeddings": true,
  "norm_style": "rms_layer_adaptive",
  "enable_memory": true,
  "enable_steering": true,
  "memory_capacity": 128,
  "memory_top_k": 4,
  "memory_strength": 0.25,
  "use_high_fidelity_text_bridge": true,
  "bridge_dim": 768,
  "bridge_hidden_mult": 4,
  "bridge_gate_init": 0.0,
  "use_sdxl_conditioning_projector": true,
  "sdxl_token_dim": 2048,
  "sdxl_pooled_dim": 1280,
  "image_generator_class": "LightweightLatentImageGenerator",
  "image_generator_config": {
    "cond_dim": 768,
    "latent_channels": 4,
    "base_channels": 256,
    "diffusion_steps": 1000,
    "use_multiscale_refiner": true,
    "use_highfreq_head": true,
    "decoder_res_blocks": 0,
    "refiner_channels": 128,
    "use_attention_refiner": false,
    "generation_mode": "latent_diffusion",
    "vae_model_name_or_path": "models/Phillnet-2-SDXL-UNet-VAE",
    "vae_scale_factor": 0.13025,
    "decode_latents_on_generate": true,
    "latent_diffusion_channels": 256,
    "latent_diffusion_blocks": 2,
    "latent_diffusion_attention": false,
    "num_train_timesteps": 1000,
    "prediction_type": "epsilon",
    "default_inference_steps": 8,
    "denoiser_backbone": "multiscale_unet",
    "unet_base_channels": 192,
    "unet_res_blocks_per_stage": 2,
    "use_token_cross_attention": true,
    "cross_attention_heads": 8,
    "final_decode_mode": "unified",
    "final_rgb_blend": 0.35,
    "use_spatial_text_prior": true,
    "spatial_prior_hidden": 256,
    "spatial_prior_heads": 4,
    "spatial_prior_layers": 2,
    "spatial_prior_query_count": 256,
    "enable_quality_adapter": true,
    "quality_adapter_hidden": 64,
    "enable_visual_contract_adapter": true,
    "visual_contract_hidden": 64,
    "visual_contract_maps": 8,
    "enable_refiner_lora": true,
    "refiner_lora_rank": 16,
    "refiner_lora_hidden": 32,
    "enable_latent_refiner": true,
    "latent_refiner_hidden": 128,
    "enable_structure_prior": true,
    "structure_prior_hidden": 192,
    "structure_prior_seed_size": 16,
    "structure_prior_heads": 4,
    "use_pretrained_unet": true,
    "pretrained_unet_model_name_or_path": "models/Phillnet-2-SDXL-UNet-VAE"
  },
  "aligner_input_dims": [
    768,
    1024
  ],
  "use_qwen_text_refiner": true,
  "qwen_refiner_hidden": 1024,
  "qwen_refiner_intermediate": 3584,
  "qwen_refiner_layers": 16,
  "qwen_refiner_attention_indices": [
    3,
    7,
    11,
    15
  ],
  "qwen_refiner_weights": "models/qwen_aligned_refiner/deep_16.pt",
  "text_tokenizer_dir": "tokenizer",
  "use_vision_encoder": false,
  "vision_hidden_size": 768,
  "vision_target_dim": 1024,
  "image_processor_dir": null
}