{ "_class_name": "MiniMaxH3Transformer3DModel", "_diffusers_version": "0.40.0.dev0", "attention_head_dim": 16, "audio_in_channels": 8, "ffn_dim": 32, "final_norm_eps": 1e-05, "freq_dim": 8, "hidden_size": 24, "in_channels": 4, "norm_eps": 1e-05, "num_attention_heads": 2, "num_layers": 2, "num_refiner_layers": 2, "patch_size": [ 1, 2, 2 ], "qk_norm_eps": 1e-05, "rope_freq_dim": 2, "rope_theta": 10000.0, "text_dim": 16, "time_embed_dim": 16, "time_embed_hidden_dim": 24 }