transformer/config.json
| 1 | { |
| 2 | "_class_name": "WanTransformer3DModel", |
| 3 | "_diffusers_version": "0.33.0.dev0", |
| 4 | "added_kv_proj_dim": null, |
| 5 | "attention_head_dim": 128, |
| 6 | "cross_attn_norm": true, |
| 7 | "eps": 1e-06, |
| 8 | "ffn_dim": 8960, |
| 9 | "freq_dim": 256, |
| 10 | "image_dim": null, |
| 11 | "in_channels": 16, |
| 12 | "num_attention_heads": 12, |
| 13 | "num_layers": 30, |
| 14 | "out_channels": 16, |
| 15 | "patch_size": [ |
| 16 | 1, |
| 17 | 2, |
| 18 | 2 |
| 19 | ], |
| 20 | "qk_norm": "rms_norm_across_heads", |
| 21 | "rope_max_seq_len": 1024, |
| 22 | "text_dim": 4096 |
| 23 | } |
| 24 | |