Pocket-Sized Multimodal AI for content understanding and generation across multilingual texts, images, and 🔜 video, up to 5x faster than OpenAI CLIP and LLaVA 🖼️ & 🖋️
Currently, the models and preprocessors are downloaded to ~/.cache/huggingface/hub/ ... path and users do not have capability to change this. Most libs like sentence_transformers, fastembed provide something similar.
This feature request allows users to set a custom path where the model etc. can be downloaded.
This option can also prevent unnecessary downloads if the model already exists at the provided location.
Let me know if this request makes sense, i can drop a PR.
Currently, the models and preprocessors are downloaded to
~/.cache/huggingface/hub/ ...
path and users do not have capability to change this. Most libs like sentence_transformers, fastembed provide something similar.This feature request allows users to set a custom path where the model etc. can be downloaded.
This option can also prevent unnecessary downloads if the model already exists at the provided location.
Let me know if this request makes sense, i can drop a PR.