xyzhang626

@oahzxl Could you further add the FID you obtain and corresponding training steps with this repo into the README? I think that will be more attractive for people want to...

Hey @Eurus-Holmes awesome script! just what i'm searching for. This repo shows a lot first stage eval results but does not provide the eval script. BTW have you compared the...

hey @MaelSeed I met the same problem and have a similar thought with you. It's related to the multiprocess IO issue but can not figure out where to solve it....

hey @BrianPulfer have you made some progress after this post? It seems DINOv2 is the only pre-trained model evaluated with linear prob in segmentation. No other refs could be found....

yes I can confirm the multi-gpu training will cause the logger IO issue some now and then ``` File "/opt/conda/lib/python3.10/site-packages/wandb/sdk/internal/sender.py", line 1174, in _update_summary with open(summary_path, "w") as f: FileNotFoundError:...

Hey @skeskinen and guys, I've made another fork to support multilingual tokenizer, real batch inference and the bge series model. Check this out if you're interested. [embeddings.cpp](https://github.com/xyzhang626/embeddings.cpp.git)

+1 that will benefit the community a lot.

Also curious about the hypeparam settings. It's important to reproduce and understand the performance. It will benefit a lot for new work based on QwenVL-2.5. Could you kindly further answer...

hey @zgzxy001 Could you share the script or any reference you used to reproduce the result? That will be super helpful to me. Thanks!