config.json

{
  "model": "JiT-DDT",
  "architecture": {
    "dim": 2944,
    "num_heads": 23,
    "num_layers": 11,
    "ffn_dim": 7936,
    "refiner_blocks": 2,
    "refiner_ffn_dim": 4096,
    "text_dim": 7680,
    "text_len": 256,
    "in_channels": 3,
    "out_channels": 3,
    "freq_dim": 256,
    "eps": 1e-06,
    "bottleneck_dim": 256,
    "encoder_patch": [
      1,
      64,
      64
    ],
    "encoder_output_patch": [
      1,
      8,
      8
    ],
    "decoder_patch": [
      1,
      32,
      32
    ],
    "rope_max_positions": 1024,
    "text_layers": [
      7,
      15,
      27
    ]
  },
  "text_encoder": {
    "model": "Qwen3.5-4B",
    "extraction_layers": [
      7,
      15,
      27
    ],
    "max_length": 256
  },
  "sampler_defaults": {
    "height": 512,
    "width": 512,
    "sampling_steps": 50,
    "guidance_scale": 15.0,
    "apg_momentum": -0.75,
    "apg_eta": 0.0,
    "apg_rescale": 10.0,
    "noise_scale": 2.0,
    "negative_prompt": "watermark, signature, logo, copyright, url",
    "num_timesteps": 1000
  },
  "source": {
    "checkpoint": "ep2-step105450-step.pt",
    "global_step": 105450,
    "weights": "ema",
    "ema_half_life_steps": 6594,
    "ema_start_step": 7321,
    "num_parameters": 2487506290
  }
}