Asunatan comments

Results 8 comments of


                                            Asunatan

Some questions about the details

The second question is a good one and deserves further research.

LLaVA-OneVision 的支持

Missing mm_projector in latest LLaVa

如何保证function_call每次调用都返回两条massages，分别是content以及function_call?

+1，prompt engineering在较小的Qwen2-VL 2B上并不好使，偶尔会出现这个问题

swift微调Qwen2.5-VL，合并模型后，分别使用swift和transform推理结果不一致

我遇到同样的问题，请问是否解决了？

Questions on Bounding Box Coordinate Format and Image Resizing for Qwen3-VL Fine-tuning

Thanks a lot for the clear explanation! Your example made the coordinate system totally intuitive, and the conversion formula will save me a ton of time. Really appreciate your quick...

Questions on Bounding Box Coordinate Format and Image Resizing for Qwen3-VL Fine-tuning

@Rakshith12-pixel Hi! You can directly use the official Qwen3-VL source—it's extremely helpful. For a quick start, just call `apply_chat_template `on your conversations, then grab everything between `assistant\n` and `\n` to...

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

Hello. In Section 5.2 “Model Distillation,” the document clearly states that the smaller models (ViT-Small, ViT-Base, ViT-Large) do not use Gram loss (i.e., the Gram-anchoring technique) during distillation. Specifically, in...