Some cast/dtype fixes for the birefnet and dino3 models.#14217
Conversation
|
No actionable comments were generated in the recent review. 🎉 ℹ️ Recent review info⚙️ Run configurationConfiguration used: Path: .coderabbit.yaml Review profile: CHILL Plan: Pro Run ID: 📒 Files selected for processing (1)
💤 Files with no reviewable changes (1)
📝 WalkthroughWalkthroughThis PR removes explicit dtype casts for DINOv3 image inputs: DINOv3ViTEmbeddings and DINOv3ViTModel no longer coerce pixel_values to the patch conv weight dtype, and ClipVisionModel.init no longer overrides dtype to bfloat16 for DINOv3. Separately, WindowAttention.forward in BiRefNet now casts the permuted relative positional bias with comfy.ops.cast_to_input before adding it to attention logits. 🚥 Pre-merge checks | ✅ 3 | ❌ 2❌ Failed checks (1 warning, 1 inconclusive)
✅ Passed checks (3 passed)
✏️ Tip: You can configure your own custom pre-merge checks in the settings. Thanks for using CodeRabbit! It's free for OSS, and your support helps us grow. If you like it, consider giving us a shout-out. Comment |
No description provided.