Refactor rotary embedding calculations in SkyReelsV2 to separate cosine and sine frequencies

Author: tolgacangoz

src/diffusers/models/transformers/transformer_skyreels_v2.py

@@ -108,13 +108,21 @@ def __call__(
 
         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.float32).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)
+            def apply_rotary_emb(
+                hidden_states: torch.Tensor,
+                freqs_cos: torch.Tensor,
+                freqs_sin: torch.Tensor,
+            ):
+                x1, x2 = hidden_states.unflatten(-1, (-1, 2)).unbind(-1)
+                cos = freqs_cos[..., 0::2]
+                sin = freqs_sin[..., 1::2]
+                out = torch.empty_like(hidden_states)
+                out[..., 0::2] = x1 * cos - x2 * sin
+                out[..., 1::2] = x1 * sin + x2 * cos
+                return 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
@@ -358,7 +366,11 @@ def forward(
 
 class SkyReelsV2RotaryPosEmbed(nn.Module):
     def __init__(
-        self, attention_head_dim: int, patch_size: Tuple[int, int, int], max_seq_len: int, theta: float = 10000.0
+        self,
+        attention_head_dim: int,
+        patch_size: Tuple[int, int, int],
+        max_seq_len: int,
+        theta: float = 10000.0,
     ):
         super().__init__()
 
@@ -368,35 +380,52 @@ def __init__(
 
         h_dim = w_dim = 2 * (attention_head_dim // 6)
         t_dim = attention_head_dim - h_dim - w_dim
+        freqs_dtype = torch.float32 if torch.backends.mps.is_available() else torch.float64
+
+        freqs_cos = []
+        freqs_sin = []
 
-        freqs = []
         for dim in [t_dim, h_dim, w_dim]:
-            freq = get_1d_rotary_pos_embed(
-                dim, max_seq_len, theta, use_real=False, repeat_interleave_real=False, freqs_dtype=torch.float32
+            freq_cos, freq_sin = get_1d_rotary_pos_embed(
+                dim,
+                max_seq_len,
+                theta,
+                use_real=True,
+                repeat_interleave_real=True,
+                freqs_dtype=freqs_dtype,
             )
-            freqs.append(freq)
-        self.freqs = torch.cat(freqs, dim=1)
+            freqs_cos.append(freq_cos)
+            freqs_sin.append(freq_sin)
+
+        self.register_buffer("freqs_cos", torch.cat(freqs_cos, dim=1), persistent=False)
+        self.register_buffer("freqs_sin", torch.cat(freqs_sin, dim=1), persistent=False)
 
     def forward(self, hidden_states: torch.Tensor) -> torch.Tensor:
         batch_size, num_channels, num_frames, height, width = hidden_states.shape
         p_t, p_h, p_w = self.patch_size
         ppf, pph, ppw = num_frames // p_t, height // p_h, width // p_w
 
-        freqs = self.freqs.to(hidden_states.device)
-        freqs = freqs.split_with_sizes(
-            [
-                self.attention_head_dim // 2 - 2 * (self.attention_head_dim // 6),
-                self.attention_head_dim // 6,
-                self.attention_head_dim // 6,
-            ],
-            dim=1,
-        )
+        split_sizes = [
+            self.attention_head_dim - 2 * (self.attention_head_dim // 3),
+            self.attention_head_dim // 3,
+            self.attention_head_dim // 3,
+        ]
+
+        freqs_cos = self.freqs_cos.split(split_sizes, dim=1)
+        freqs_sin = self.freqs_sin.split(split_sizes, dim=1)
+
+        freqs_cos_f = freqs_cos[0][:ppf].view(ppf, 1, 1, -1).expand(ppf, pph, ppw, -1)
+        freqs_cos_h = freqs_cos[1][:pph].view(1, pph, 1, -1).expand(ppf, pph, ppw, -1)
+        freqs_cos_w = freqs_cos[2][:ppw].view(1, 1, ppw, -1).expand(ppf, pph, ppw, -1)
+
+        freqs_sin_f = freqs_sin[0][:ppf].view(ppf, 1, 1, -1).expand(ppf, pph, ppw, -1)
+        freqs_sin_h = freqs_sin[1][:pph].view(1, pph, 1, -1).expand(ppf, pph, ppw, -1)
+        freqs_sin_w = freqs_sin[2][:ppw].view(1, 1, ppw, -1).expand(ppf, pph, ppw, -1)
+
+        freqs_cos = torch.cat([freqs_cos_f, freqs_cos_h, freqs_cos_w], dim=-1).reshape(1, ppf * pph * ppw, 1, -1)
+        freqs_sin = torch.cat([freqs_sin_f, freqs_sin_h, freqs_sin_w], dim=-1).reshape(1, ppf * pph * ppw, 1, -1)
 
-        freqs_f = freqs[0][:ppf].view(ppf, 1, 1, -1).expand(ppf, pph, ppw, -1)
-        freqs_h = freqs[1][:pph].view(1, pph, 1, -1).expand(ppf, pph, ppw, -1)
-        freqs_w = freqs[2][:ppw].view(1, 1, ppw, -1).expand(ppf, pph, ppw, -1)
-        freqs = torch.cat([freqs_f, freqs_h, freqs_w], dim=-1).reshape(1, 1, ppf * pph * ppw, -1)
-        return freqs
+        return freqs_cos, freqs_sin
 
 
 @maybe_allow_in_graph