王迎波

icon location 郑州

Results 5 comments of 王迎波

mmcv版本问题

请问能否提供一下你的conda环境的压缩包呢，我想进行克隆以避免不用的版本错误

can i run this as api mode?

> I've been using ` --api --listen` for api calling for a long time. it's working fine.我已经使用 ` --api --listen` 进行 api 调用很长时间了。工作正常。 I used --api 9999 to specify the...

I want to fine-tune a complete text encoder model, but it seems that the model trained by ft-B-train-OpenAI-CLIP-ViT-L-14.py is a visual encoder model.

For the "TE only" / text encoder only models on my HuggingFace, I fine-tuned the entire CLIP model (text + vision) and then simply "detached" the vision transformer (i.e. delete...

I want to fine-tune a complete text encoder model, but it seems that the model trained by ft-B-train-OpenAI-CLIP-ViT-L-14.py is a visual encoder model.

> I just committed `Convert-for-HuggingFace-Spaces-etc` - the folder contains all the scripts + documentation / how-to use. Please let me know if that works for you!我刚刚提交了 `Convert-for-HuggingFace-Spaces-etc` - 该文件夹包含所有脚本+文档/如何使用。请告诉我这是否适合您！ Thank...

I want to fine-tune a complete text encoder model, but it seems that the model trained by ft-B-train-OpenAI-CLIP-ViT-L-14.py is a visual encoder model.

> Can you open `/opt/conda/lib/python3.11/site-packages/clip/clip.py` and edit line 129? Where it says: > > `model = torch.jit.load(opened_file, map_location=device if jit else "cpu").eval() ` -> Change that to: > > ```...