Thomas Germer

> Could you please merge it? I can not merge this PR because I do not have write access to this repository. You can go through recently merged PRs to...

> Please use “torch_dtype=torch.bfloat16“ to load model, try again~ I added `torch_dtype=torch.bfloat16` to the `Qwen3VLForConditionalGeneration.from_pretrained`, but still get the infinite repeat bug (code below). I also get the warning `torch_dtype`...

I think the problem lies in the way that the Qwen models were trained. When the same token has been generated a few times, the model is stuck in a...

@scouzi1966 It is nice that it passes for you on a larger model with llama.cpp backend, but it would be nice if it worked all the time on all images,...

Which tasks? And what do you mean with "open"?

What makes you think so?

The line from your first message initializes the member `block.attn.out_proj.weight` of `self.transformer.resblocks` * https://github.com/openai/CLIP/blob/dcba3cb2e2827b402d2701e7e1c7d9fed8a20ef1/clip/model.py#L321 * https://github.com/openai/CLIP/blob/dcba3cb2e2827b402d2701e7e1c7d9fed8a20ef1/clip/model.py#L319 * https://github.com/openai/CLIP/blob/dcba3cb2e2827b402d2701e7e1c7d9fed8a20ef1/clip/model.py#L282 * https://github.com/openai/CLIP/blob/dcba3cb2e2827b402d2701e7e1c7d9fed8a20ef1/clip/model.py#L200 * https://github.com/openai/CLIP/blob/dcba3cb2e2827b402d2701e7e1c7d9fed8a20ef1/clip/model.py#L175 * https://github.com/pytorch/pytorch/blob/3f77002b968c093b2b668fed24d85f6a365d6b3c/torch/nn/modules/activation.py#L1102 But in your last message, you...

This is a common mistake. Here you can track the progress of [VS Code](https://github.com/microsoft/vscode/issues/3884), [Claude Code](https://github.com/anthropics/claude-code/issues/1455) and [Aider](https://github.com/Aider-AI/aider/issues/216) as they fix the same issue. > `XDG_CONFIG_HOME` on Mac (I know...

I also extend my thanks! Feel free to close this issue at your convenience (or keep it open if you want to discuss more). My question is thoroughly answered.

Do you have uncompressed full resolution images and the trimap you used?