base/transformer/config.json

{
  "_class_name": "BooguImageTransformer2DModel",
  "_diffusers_version": "0.35.2",
  "axes_dim_rope": [
    40,
    40,
    40
  ],
  "axes_lens": [
    2048,
    1664,
    1664
  ],
  "ffn_dim_multiplier": null,
  "hidden_size": 3360,
  "in_channels": 16,
  "instruction_feature_configs": {
    "instruction_feat_dim": 4096,
    "num_instruction_feature_layers": 1,
    "reduce_type": "mean"
  },
  "multiple_of": 256,
  "norm_eps": 0.00001,
  "num_attention_heads": 28,
  "num_double_stream_layers": 8,
  "num_kv_heads": 7,
  "num_layers": 40,
  "num_refiner_layers": 2,
  "out_channels": null,
  "patch_size": 2,
  "prompt_tuning_configs": {
    "ffn_dim_multiplier": null,
    "hidden_size": 4096,
    "multiple_of": 256,
    "norm_eps": 0.00001,
    "num_attention_heads": 32,
    "num_kv_heads": 8,
    "num_layers": 0,
    "num_trainable_prompt_tokens": 0,
    "use_causal_mask": true,
    "use_prompt_tuning": false
  },
  "quantization": {
    "bits": 8,
    "group_size": 32
  },
  "timestep_scale": 1000.0
}