dinov3

dinov3 copied to clipboard

Update import statement for torchvision transforms

2

Fixed the transforms import in the readme for the segmenting usage.

Mapika

How to efficiently load the model after training

3

I am trying to train DINOv3 again on my custom dataset from the begining (3 stages). I can be able to train the first stage (8 H100 GPUs on one...

quannguyenminh103

vicdxxx

Reducing size and increasing inference speed

1

We have seen some promising preliminary results on segmentation using the DinoV3 ViTs16 backbone with the ViTAdapter+Mask2Former head. However, this setup (~89M params with hidden_dim=384) is larger and slower than...

Kodemannen

How to use gram anchoring loss in my custom dataset?

8

Hello, I have successfully run the fast setup using ViT-L/16 in my own histopathology dataset, and I am now trying to conduct the gram anchoring step. I notice that the...

Extord1108

Anyone tested quantization with success? my results are bad

1

Anyone tested quantization with success? my results are bad compared to the float16

ilan-design

Patch embeddings are all the same with float16 on mps

1

the following code works with - float16 or float32 on cuda - float32 on mps When I execute it with float16 on mps, all the patch embeddings are equal. ```...

oilst

> def _make_dinov3_detector( *, backbone_name: str, pretrained: bool = True, detector_weights: str | DetectionWeights, backbone_weights: str | BackboneWeights, check_hash: bool = False, **kwargs, ): detection_kwargs = dict( with_box_refine=True, two_stage=True, mixed_selection=True,...

YifengDeng1

How to Fine-Tune DINOv3 on Custom Dataset with Self-Supervised Learning Locally

26

Hi team, I'm new to DINO and want to fine-tune DINOv3 using self-supervised learning on my custom dataset. Hardware: 8 NVIDIA 4090 GPU. Goal: Continue training from pre-trained checkpoints. Use...

MaybeRichard

CPU memory leak

4

When training DINOv3 on a server with 500 GB of CPU memory using my own dataset, I noticed an unreasonable linear growth in memory usage, which eventually caused the server...

Extord1108