transformer/config.json

{
  "_class_name": "SanaTransformer2DModel",
  "_diffusers_version": "0.32.0.dev0",
  "activation_fn": [
    "silu",
    "silu",
    null
  ],
  "attention_bias": false,
  "attention_head_dim": 32,
  "attention_type": "default",
  "caption_channels": 2304,
  "caption_norm_scale_factor": 0.1,
  "cross_attention_dim": 2240,
  "cross_attention_head_dim": 112,
  "dropout": 0.0,
  "expand_ratio": 2.5,
  "ff_bias": [
    true,
    true,
    false
  ],
  "ff_norm": [
    null,
    null,
    null
  ],
  "in_channels": 32,
  "interpolation_scale": null,
  "norm_elementwise_affine": false,
  "norm_eps": 1e-06,
  "norm_num_groups": 32,
  "norm_type": "ada_norm_single",
  "num_attention_heads": 70,
  "num_cross_attention_heads": 20,
  "num_embeds_ada_norm": 1000,
  "num_layers": 20,
  "out_channels": 32,
  "patch_size": 1,
  "sample_size": 32,
  "upcast_attention": false,
  "use_additional_conditions": false,
  "use_caption_norm": true,
  "use_pe": false
}