{
  "hidden_size": 4096,
  "num_attention_heads": 32,
  "num_layers": 36,
  "ffn_hidden_size": 12288,
  "in_channels": 128,
  "out_channels": 128,
  "patch_size": 1,
  "text_in_dim": 3072,
  "rope_theta": 256,
  "rope_axes_dim": [
    32,
    48,
    48
  ],
  "eps": 1e-06
}