{
  "_class_name": "OpenSoraT2V",
  "_diffusers_version": "0.28.0",
  "activation_fn": "gelu-approximate",
  "attention_bias": true,
  "attention_head_dim": 96,
  "attention_mode": "xformers",
  "attention_type": "default",
  "caption_channels": 4096,
  "cross_attention_dim": 2304,
  "decay": 0.9999,
  "double_self_attention": false,
  "downsampler": null,
  "dropout": 0.0,
  "in_channels": 4,
  "interpolation_scale_h": 1.0,
  "interpolation_scale_t": 1.0,
  "interpolation_scale_w": 1.0,
  "inv_gamma": 1.0,
  "min_decay": 0.0,
  "norm_elementwise_affine": false,
  "norm_eps": 1e-06,
  "norm_num_groups": 32,
  "norm_type": "ada_norm_single",
  "num_attention_heads": 24,
  "num_embeds_ada_norm": 1000,
  "num_layers": 32,
  "num_vector_embeds": null,
  "only_cross_attention": false,
  "optimization_step": 3000,
  "out_channels": 4,
  "patch_size": 2,
  "patch_size_t": 1,
  "power": 0.6666666666666666,
  "sample_size": [
    60,
    80
  ],
  "sample_size_t": 24,
  "upcast_attention": false,
  "update_after_step": 0,
  "use_additional_conditions": null,
  "use_ema_warmup": false,
  "use_linear_projection": false,
  "use_rope": true,
  "use_stable_fp32": false
}