README.md

Quick Inference Demo

1. Initialize the environment
   Run the environment setup script to install necessary dependencies:

   bash init_env.sh
   

2. Run the demo script
   Execute the demo script to test LLaDA-V on an example image:

   python generate_demo.py
   

Evaluation

We provide the evaluation code in this repository, following the lmms-eval library.

1. Initialize the environment
   Run the environment setup script to install necessary dependencies:

   bash init_env.sh
   

2. Run the demo script
   Execute the demo script to test RedVTP-LLaDA-V on an example image:

   cd eval && bash scripts/evaluate.sh
   

3. Modify the retaining ratio Change the ratio at line 1546 in RedVTP-LLaDA-V/train/llava/model/language_model/modeling_llada.py:

   keep_ratio = 0.50