transformer/config.json

{
  "_class_name": "MochiTransformer3DModel",
  "_diffusers_version": "0.32.0.dev0",
  "activation_fn": "swiglu",
  "attention_head_dim": 128,
  "in_channels": 12,
  "max_sequence_length": 256,
  "num_attention_heads": 24,
  "num_layers": 48,
  "out_channels": null,
  "patch_size": 2,
  "pooled_projection_dim": 1536,
  "qk_norm": "rms_norm",
  "text_embed_dim": 4096,
  "time_embed_dim": 256
}