Spaces:
Paused
Paused
File size: 3,507 Bytes
1c72248 |
1 2 3 4 5 6 7 8 9 10 11 12 13 14 15 16 17 18 19 20 21 22 23 24 25 26 27 28 29 30 31 32 33 34 35 36 37 38 39 40 41 42 43 44 45 46 47 48 49 50 51 52 53 54 55 56 57 58 59 60 61 62 63 64 65 66 67 68 69 70 71 72 73 74 75 76 77 78 79 80 81 82 83 84 85 |
import torch
import torch.nn.functional as F
from typing import Optional
from diffusers.models.attention_processor import Attention
# modified to set the image embedder size
class WanAttnProcessor2_0:
def __init__(self, num_img_tokens: int = 257):
self.num_img_tokens = num_img_tokens
if not hasattr(F, "scaled_dot_product_attention"):
raise ImportError(
"WanAttnProcessor2_0 requires PyTorch 2.0. To use it, please upgrade PyTorch to 2.0.")
def __call__(
self,
attn: Attention,
hidden_states: torch.Tensor,
encoder_hidden_states: Optional[torch.Tensor] = None,
attention_mask: Optional[torch.Tensor] = None,
rotary_emb: Optional[torch.Tensor] = None,
) -> torch.Tensor:
encoder_hidden_states_img = None
if attn.add_k_proj is not None:
encoder_hidden_states_img = encoder_hidden_states[:,
:self.num_img_tokens]
encoder_hidden_states = encoder_hidden_states[:,
self.num_img_tokens:]
if encoder_hidden_states is None:
encoder_hidden_states = hidden_states
query = attn.to_q(hidden_states)
key = attn.to_k(encoder_hidden_states)
value = attn.to_v(encoder_hidden_states)
if attn.norm_q is not None:
query = attn.norm_q(query)
if attn.norm_k is not None:
key = attn.norm_k(key)
query = query.unflatten(2, (attn.heads, -1)).transpose(1, 2)
key = key.unflatten(2, (attn.heads, -1)).transpose(1, 2)
value = value.unflatten(2, (attn.heads, -1)).transpose(1, 2)
if rotary_emb is not None:
def apply_rotary_emb(hidden_states: torch.Tensor, freqs: torch.Tensor):
x_rotated = torch.view_as_complex(
hidden_states.to(torch.float64).unflatten(3, (-1, 2)))
x_out = torch.view_as_real(x_rotated * freqs).flatten(3, 4)
return x_out.type_as(hidden_states)
query = apply_rotary_emb(query, rotary_emb)
key = apply_rotary_emb(key, rotary_emb)
# I2V task
hidden_states_img = None
if encoder_hidden_states_img is not None:
key_img = attn.add_k_proj(encoder_hidden_states_img)
key_img = attn.norm_added_k(key_img)
value_img = attn.add_v_proj(encoder_hidden_states_img)
key_img = key_img.unflatten(2, (attn.heads, -1)).transpose(1, 2)
value_img = value_img.unflatten(
2, (attn.heads, -1)).transpose(1, 2)
hidden_states_img = F.scaled_dot_product_attention(
query, key_img, value_img, attn_mask=None, dropout_p=0.0, is_causal=False
)
hidden_states_img = hidden_states_img.transpose(1, 2).flatten(2, 3)
hidden_states_img = hidden_states_img.type_as(query)
hidden_states = F.scaled_dot_product_attention(
query, key, value, attn_mask=attention_mask, dropout_p=0.0, is_causal=False
)
hidden_states = hidden_states.transpose(1, 2).flatten(2, 3)
hidden_states = hidden_states.type_as(query)
if hidden_states_img is not None:
hidden_states = hidden_states + hidden_states_img
hidden_states = attn.to_out[0](hidden_states)
hidden_states = attn.to_out[1](hidden_states)
return hidden_states
|