Use RL to predict perturbations to language model's outputs.
Use the below to install the F_GPT environment.
# Download the latest Linux installer via wget
wget https://repo.anaconda.com/miniconda/Miniconda3-latest-Linux-x86_64.sh -O Miniconda3-latest-Linux-x86_64.sh
# Execute .sh
bash Miniconda3-latest-Linux-x86_64.sh
# After installation, open a new terminal or run `source ~/.bashrc` to activate the conda env.
# Change the install path to /home/ziangcao2022/workspace/miniconda3# Open a new Terminal, make sure you are in the (base) conda env
conda env list
# Create F_GPT conda env
conda create -n F_GPT python==3.9 -y
Then either run src/fast_detect_gpt/gpt_batch_a1.ipynb, or run python3 private_support_code/fast-detect-gpt/GPT_TRUE_BATCH_A1.py or python3 private_support_code/fast-detect-gpt/GPT_A1.py.