Great Tool. one question.

[JacobGoldenArt]

Hey, I've got all of your examples running and it's working well. It seems when I add a model to the model string, it is downloaded, but what is the process of using a local directory of mlx models that I've downloaded already? Also, I know that once I've run the example the first time which takes longer because it is downloading the model. Are you then loading the model? I ask because, If I want to have quick access to 3 or 4 models, is there a way of 'preloading them?' like you might with LLM Studio or Ollama? or is that not necessary with your tool?

Thanks!

[madroidmaq]

This project adopts the same method as the mlx-examles project, both of which load and manage models through huggingface. Generally speaking, you can find the downloaded model files in the directory /Users/user_name/.cache/huggingface/hub/.

Additionally, you can also use the huggingface-cli command-line tool to pre-download, such as huggingface-cli download mlx-community/phi-4-4bit.