Cache RoPE freqs on device to avoid repeated CPU-GPU copy in QwenImage (#13406)

* Cache RoPE freqs on device to avoid repeated CPU-GPU copy in QwenImage

* Apply style fixes

* use lru_cache_unless_export

---------

Co-authored-by: github-actions[bot] <github-actions[bot]@users.noreply.github.com>
Co-authored-by: dg845 <58458699+dg845@users.noreply.github.com>
Co-authored-by: Sayak Paul <spsayakpaul@gmail.com>

Author: 4 people

src/diffusers/models/transformers/transformer_qwenimage.py

@@ -233,6 +233,11 @@ def rope_params(self, index, dim, theta=10000):
         freqs = torch.polar(torch.ones_like(freqs), freqs)
         return freqs
 
+    @lru_cache_unless_export(maxsize=None)
+    def _get_device_freqs(self, device: torch.device) -> tuple[torch.Tensor, torch.Tensor]:
+        """Return pos_freqs and neg_freqs on the given device."""
+        return self.pos_freqs.to(device), self.neg_freqs.to(device)
+
     def forward(
         self,
         video_fhw: tuple[int, int, int, list[tuple[int, int, int]]],
@@ -300,8 +305,9 @@ def forward(
                 max_vid_index = max(height, width, max_vid_index)
 
         max_txt_seq_len_int = int(max_txt_seq_len)
-        # Create device-specific copy for text freqs without modifying self.pos_freqs
-        txt_freqs = self.pos_freqs.to(device)[max_vid_index : max_vid_index + max_txt_seq_len_int, ...]
+        # Use cached device-transferred freqs to avoid CPU→GPU sync every forward call
+        pos_freqs_device, _ = self._get_device_freqs(device)
+        txt_freqs = pos_freqs_device[max_vid_index : max_vid_index + max_txt_seq_len_int, ...]
         vid_freqs = torch.cat(vid_freqs, dim=0)
 
         return vid_freqs, txt_freqs
@@ -311,8 +317,9 @@ def _compute_video_freqs(
         self, frame: int, height: int, width: int, idx: int = 0, device: torch.device = None
     ) -> torch.Tensor:
         seq_lens = frame * height * width
-        pos_freqs = self.pos_freqs.to(device) if device is not None else self.pos_freqs
-        neg_freqs = self.neg_freqs.to(device) if device is not None else self.neg_freqs
+        pos_freqs, neg_freqs = (
+            self._get_device_freqs(device) if device is not None else (self.pos_freqs, self.neg_freqs)
+        )
 
         freqs_pos = pos_freqs.split([x // 2 for x in self.axes_dim], dim=1)
         freqs_neg = neg_freqs.split([x // 2 for x in self.axes_dim], dim=1)
@@ -367,6 +374,11 @@ def rope_params(self, index, dim, theta=10000):
         freqs = torch.polar(torch.ones_like(freqs), freqs)
         return freqs
 
+    @lru_cache_unless_export(maxsize=None)
+    def _get_device_freqs(self, device: torch.device) -> tuple[torch.Tensor, torch.Tensor]:
+        """Return pos_freqs and neg_freqs on the given device."""
+        return self.pos_freqs.to(device), self.neg_freqs.to(device)
+
     def forward(
         self,
         video_fhw: tuple[int, int, int, list[tuple[int, int, int]]],
@@ -421,17 +433,19 @@ def forward(
 
         max_vid_index = max(max_vid_index, layer_num)
         max_txt_seq_len_int = int(max_txt_seq_len)
-        # Create device-specific copy for text freqs without modifying self.pos_freqs
-        txt_freqs = self.pos_freqs.to(device)[max_vid_index : max_vid_index + max_txt_seq_len_int, ...]
+        # Use cached device-transferred freqs to avoid CPU→GPU sync every forward call
+        pos_freqs_device, _ = self._get_device_freqs(device)
+        txt_freqs = pos_freqs_device[max_vid_index : max_vid_index + max_txt_seq_len_int, ...]
         vid_freqs = torch.cat(vid_freqs, dim=0)
 
         return vid_freqs, txt_freqs
 
     @lru_cache_unless_export(maxsize=None)
     def _compute_video_freqs(self, frame, height, width, idx=0, device: torch.device = None):
         seq_lens = frame * height * width
-        pos_freqs = self.pos_freqs.to(device) if device is not None else self.pos_freqs
-        neg_freqs = self.neg_freqs.to(device) if device is not None else self.neg_freqs
+        pos_freqs, neg_freqs = (
+            self._get_device_freqs(device) if device is not None else (self.pos_freqs, self.neg_freqs)
+        )
 
         freqs_pos = pos_freqs.split([x // 2 for x in self.axes_dim], dim=1)
         freqs_neg = neg_freqs.split([x // 2 for x in self.axes_dim], dim=1)
@@ -452,8 +466,9 @@ def _compute_video_freqs(self, frame, height, width, idx=0, device: torch.device
     @lru_cache_unless_export(maxsize=None)
     def _compute_condition_freqs(self, frame, height, width, device: torch.device = None):
         seq_lens = frame * height * width
-        pos_freqs = self.pos_freqs.to(device) if device is not None else self.pos_freqs
-        neg_freqs = self.neg_freqs.to(device) if device is not None else self.neg_freqs
+        pos_freqs, neg_freqs = (
+            self._get_device_freqs(device) if device is not None else (self.pos_freqs, self.neg_freqs)
+        )
 
         freqs_pos = pos_freqs.split([x // 2 for x in self.axes_dim], dim=1)
         freqs_neg = neg_freqs.split([x // 2 for x in self.axes_dim], dim=1)