Skip to content

The most powerful and modular diffusion model GUI, api and backend with a graph/nodes interface. The fastest local inference engine in the world.

Fix clip vision lowvram mode not working.

comfyanonymousc782144433e4
diff --git a/comfy/clip_model.py b/comfy/clip_model.py
index 850b5fdbe..7397b7a26 100644
--- a/comfy/clip_model.py
+++ b/comfy/clip_model.py
@@ -151,7 +151,7 @@ class CLIPVisionEmbeddings(torch.nn.Module):
def forward(self, pixel_values):
embeds = self.patch_embedding(pixel_values).flatten(2).transpose(1, 2)
- return torch.cat([self.class_embedding.expand(pixel_values.shape[0], 1, -1), embeds], dim=1) + self.position_embedding.weight
+ return torch.cat([self.class_embedding.to(embeds.device).expand(pixel_values.shape[0], 1, -1), embeds], dim=1) + self.position_embedding.weight.to(embeds.device)
class CLIPVision(torch.nn.Module):