transformer_ref/config.json
| 1 | { |
| 2 | "_class_name": "MiniMaxH3Transformer3DModel", |
| 3 | "_diffusers_version": "0.36.0.dev0", |
| 4 | "num_attention_heads": 56, |
| 5 | "attention_head_dim": 128, |
| 6 | "hidden_size": 5376, |
| 7 | "num_layers": 50, |
| 8 | "num_refiner_layers": 2, |
| 9 | "ffn_dim": 14336, |
| 10 | "in_channels": 24, |
| 11 | "audio_in_channels": 32, |
| 12 | "patch_size": [ |
| 13 | 1, |
| 14 | 2, |
| 15 | 2 |
| 16 | ], |
| 17 | "text_dim": 5120, |
| 18 | "freq_dim": 256, |
| 19 | "time_embed_hidden_dim": 5376, |
| 20 | "time_embed_dim": 2688, |
| 21 | "rope_freq_dim": 16, |
| 22 | "rope_theta": 10000.0, |
| 23 | "norm_eps": 1e-05, |
| 24 | "qk_norm_eps": 1e-05, |
| 25 | "final_norm_eps": 1e-05 |
| 26 | } |