[Core] Ensure memory is pinned prior to async h2d copy (#45424)

Signed-off-by: Nick Hill <nickhill123@gmail.com>
This commit is contained in:
Nick Hill
2026-06-20 20:02:24 -07:00
committed by GitHub
parent 8dd1b702f2
commit 7df3d7dada
49 changed files with 254 additions and 264 deletions
+2 -3
View File
@@ -49,12 +49,11 @@ class EncoderRunner:
@torch.inference_mode()
def execute_mm_encoder(
self,
mm_kwargs: list[tuple[str, MultiModalKwargsItem]],
self, mm_kwargs: list[tuple[str, MultiModalKwargsItem]]
) -> list[torch.Tensor]:
encoder_outputs: list[torch.Tensor] = []
for modality, num_items, mm_kwargs_batch in group_and_batch_mm_kwargs(
mm_kwargs, device=self.device, pin_memory=False
mm_kwargs, device=self.device, pin_memory=True
):
batch_outputs = self.model.embed_multimodal(**mm_kwargs_batch)
sanity_check_mm_encoder_outputs(batch_outputs, expected_num_items=num_items)