transformer/config.json

{
  "LFE_depth": 10,
  "LFE_dim_head": 64,
  "LFE_ff_mult": 4,
  "LFE_id_dim": 1280,
  "LFE_num_heads": 16,
  "LFE_num_id_token": 5,
  "LFE_num_querie": 32,
  "LFE_num_scale": 5,
  "LFE_output_dim": 2048,
  "LFE_vit_dim": 1024,
  "_class_name": "ConsisIDTransformer3DModel",
  "_diffusers_version": "0.33.0.dev0",
  "activation_fn": "gelu-approximate",
  "attention_bias": true,
  "attention_head_dim": 64,
  "audio_attn_interval": 1,
  "cross_attn_dim_head": 128,
  "cross_attn_interval": 2,
  "cross_attn_num_heads": 16,
  "dropout": 0.0,
  "flip_sin_to_cos": true,
  "freq_shift": 0,
  "in_channels": 48,
  "is_kps": false,
  "is_train_audio": true,
  "is_train_face": true,
  "local_face_scale": 1.0,
  "max_text_seq_length": 226,
  "norm_elementwise_affine": true,
  "norm_eps": 1e-05,
  "num_attention_heads": 48,
  "num_layers": 42,
  "ofs_embed_dim": null,
  "out_channels": 16,
  "patch_bias": true,
  "patch_size": 2,
  "patch_size_t": null,
  "sample_frames": 49,
  "sample_height": 60,
  "sample_width": 90,
  "spatial_interpolation_scale": 1.875,
  "temporal_compression_ratio": 4,
  "temporal_interpolation_scale": 1.0,
  "text_embed_dim": 4096,
  "time_embed_dim": 512,
  "timestep_activation_fn": "silu",
  "use_learned_positional_embeddings": true,
  "use_rotary_positional_embeddings": true
}