Skip to content

Latest commit

 

History

History
27 lines (20 loc) · 1.03 KB

README.md

File metadata and controls

27 lines (20 loc) · 1.03 KB

Coreco's LLaVA captioning fork

This is a total hack to use LLaVA to caption large image datasets To use it you must:

  • Follow the LLaVA installation instructions to set up this code base, acquire the llama model, and merge their delta.
  • Run the following command:
python3 -m llava.eval.run_llava --path /path/to/images --model_path /path/to/model

Full options:

usage: described [-h] --path PATH [--overwrite] [--model_name MODEL_NAME] [--prefix PREFIX] [--suffix SUFFIX] [--resize RESIZE]

options:
  -h, --help            show this help message and exit
  --path PATH           Path to images to be captioned
  --overwrite           Overwrite existing captions
  --model_name MODEL_NAME
                        path to the model
  --prefix PREFIX       a string applied at the beginning of each caption
  --suffix SUFFIX       a string applied at the end of each caption
  --resize RESIZE       additionally, resize and save the image where the longest side is the provided maximum