Describe the problem you're trying to solve
Most LLM repositories on HuggingFace bundle multiple quantization levels of the same model together (especially for GGUF files), leading to a large repo size often >100 GB, but one requires only one GGUF file (e.g just the 4-bit quantized version) for a ModelKit. Currently, import from hugging face will attempt to import the entire repo to Jozu Hub as a ModelKit.
Describe the solution you'd like
Include the option to exclude files you don't want from the repo, or only include files you want.
Describe alternatives you've considered
Manually downloading the file from HuggingFace, and packaging a ModelKit and uploading to Jozu Hub
Describe the problem you're trying to solve
Most LLM repositories on HuggingFace bundle multiple quantization levels of the same model together (especially for GGUF files), leading to a large repo size often >100 GB, but one requires only one GGUF file (e.g just the 4-bit quantized version) for a ModelKit. Currently,
import from hugging facewill attempt to import the entire repo to Jozu Hub as a ModelKit.Describe the solution you'd like
Include the option to exclude files you don't want from the repo, or only include files you want.
Describe alternatives you've considered
Manually downloading the file from HuggingFace, and packaging a ModelKit and uploading to Jozu Hub