{ "attention_head_dim": 128, "attention_type": "default", "caption_channels": 3840, "cross_attention_dim": 4096, "in_channels": 128, "norm_eps": 1e-06, "num_attention_heads": 32, "num_layers": 48, "out_channels": 128, "positional_embedding_theta": 10000.0, "positional_embedding_max_pos": [ 20, 2048, 2048 ], "timestep_scale_multiplier": 1000, "av_ca_timestep_scale_multiplier": 1000.0, "audio_num_attention_heads": 32, "audio_attention_head_dim": 64, "audio_out_channels": 128, "audio_cross_attention_dim": 2048, "audio_positional_embedding_max_pos": [ 20 ], "connector_attention_head_dim": 128, "connector_num_attention_heads": 32, "connector_num_layers": 8, "use_middle_indices_grid": true, "apply_gated_attention": true, "caption_projection_first_linear": false, "caption_projection_second_linear": false, "caption_proj_input_norm": false, "caption_proj_before_connector": true, "audio_connector_attention_head_dim": 64, "audio_connector_num_attention_heads": 32, "cross_attention_adaln": true, "rope_type": "split", "double_precision_rope": true, "_class_name": "LTX2Transformer3DModel" }