Sherlock Evaluation Guide
May 28, 2025 ยท View on GitHub
This guide provides detailed instructions for the evaluation of Sherlock model.
Installation
cd inference/VLMEvalKit
pip install -e .
pip install transformers==4.45.2
Inference
Here is two way to conduct inference of our Sherlock model.
Quick Use
Runing the following command to quick use our Sherlock model and self-correction inference:
python ./inference/demo/inference.py
Benchmark Evaluation
Our evaluation using VLMEvalKit package.
-
Create
.envfile in./inference/VLMEvalKit/. -
Prepare your OPENAI_API_KEY in
.envfile. -
Using the following command to evaluate sherlock model (8 GPU evaluation):
bash ./inference/VLMEvalKit/scripts/eval_sherlock.sh