| Wan2_2Transformer3DModel( |
| (patch_embedding): Conv3d(36, 5120, kernel_size=(1, 2, 2), stride=(1, 2, 2)) |
| (text_embedding): Sequential( |
| (0): Linear(in_features=4096, out_features=5120, bias=True) |
| (1): GELU(approximate='tanh') |
| (2): Linear(in_features=5120, out_features=5120, bias=True) |
| ) |
| (time_embedding): Sequential( |
| (0): Linear(in_features=256, out_features=5120, bias=True) |
| (1): SiLU() |
| (2): Linear(in_features=5120, out_features=5120, bias=True) |
| ) |
| (time_projection): Sequential( |
| (0): SiLU() |
| (1): Linear(in_features=5120, out_features=30720, bias=True) |
| ) |
| (blocks): ModuleList( |
| (0-39): 40 x WanAttentionBlock( |
| (norm1): WanLayerNorm((5120,), eps=1e-06, elementwise_affine=False) |
| (self_attn): WanSelfAttention( |
| (q): Linear(in_features=5120, out_features=5120, bias=True) |
| (k): Linear(in_features=5120, out_features=5120, bias=True) |
| (v): Linear(in_features=5120, out_features=5120, bias=True) |
| (o): Linear(in_features=5120, out_features=5120, bias=True) |
| (norm_q): WanRMSNorm() |
| (norm_k): WanRMSNorm() |
| ) |
| (norm3): WanLayerNorm((5120,), eps=1e-06, elementwise_affine=True) |
| (cross_attn): WanCrossAttention( |
| (q): Linear(in_features=5120, out_features=5120, bias=True) |
| (k): Linear(in_features=5120, out_features=5120, bias=True) |
| (v): Linear(in_features=5120, out_features=5120, bias=True) |
| (o): Linear(in_features=5120, out_features=5120, bias=True) |
| (norm_q): WanRMSNorm() |
| (norm_k): WanRMSNorm() |
| ) |
| (norm2): WanLayerNorm((5120,), eps=1e-06, elementwise_affine=False) |
| (ffn): Sequential( |
| (0): Linear(in_features=5120, out_features=13824, bias=True) |
| (1): GELU(approximate='tanh') |
| (2): Linear(in_features=13824, out_features=5120, bias=True) |
| ) |
| ) |
| ) |
| (head): Head( |
| (norm): WanLayerNorm((5120,), eps=1e-06, elementwise_affine=False) |
| (head): Linear(in_features=5120, out_features=64, bias=True) |
| ) |
| ) |
