config.json

{
  "motion_input_dim": 272,
  "motion_hidden_dim": 768,
  "embed_dim": 512,
  "motion_num_heads": 12,
  "motion_num_layers": 8,
  "motion_max_seq_len": 784,
  "motion_dropout": 0.1,
  "text_encoder_name": "openai/clip-vit-base-patch32"
}