| { |
| "_class_name": "LTX2VideoTransformer3DModel", |
| "_diffusers_version": "0.37.0.dev0", |
| "activation_fn": "gelu-approximate", |
| "attention_bias": true, |
| "attention_head_dim": 128, |
| "attention_out_bias": true, |
| "audio_attention_head_dim": 64, |
| "audio_cross_attention_dim": 2048, |
| "audio_dim": 2048, |
| "audio_freq_embed_dim": 256, |
| "audio_in_channels": 128, |
| "audio_num_attention_heads": 32, |
| "audio_patch_size": 1, |
| "audio_patch_size_t": 1, |
| "caption_channels": 4096, |
| "caption_projection_dim": 4096, |
| "cross_attention_dim": 4096, |
| "in_channels": 128, |
| "norm_elementwise_affine": false, |
| "norm_eps": 1e-06, |
| "num_attention_heads": 32, |
| "num_layers": 48, |
| "out_channels": 128, |
| "patch_size": 1, |
| "patch_size_t": 1, |
| "qk_norm": "rms_norm_across_heads", |
| "rope_type": "interleaved", |
| "video_gated_attn": true, |
| "video_cross_attn_adaln": true, |
| "audio_gated_attn": true, |
| "audio_cross_attn_adaln": true |
| } |