text_cond_config.yaml

d_latent: 1048576
inner:
  cond_drop_prob: 0.1
  heads: 16
  init_scale: 0.25
  layers: 24
  name: CLIPImagePointDiffusionTransformer
  pos_emb_init_scale: 0.05
  time_token_cond: true
  token_cond: true
  use_pos_emb: true
  width: 1024
latent_ctx: 1024
name: SplitVectorDiffusion