This video installs Agentless and tries it to out on SWE-Benchmark. It is an agentless approach to automatically solve software development problems.
Code:
conda create -n agentless python=3.11 -y && conda activate agentless
conda install jupyter -y
pip uninstall charset_normalizer -y
pip install charset_normalizer
jupyter notebook
pip install torch
pip install git+https://github.com/huggingface/transformers
pip install openai
git clone https://github.com/OpenAutoCoder/Agentless.git && cd Agentless
pip install -r requirements.txt
export PYTHONPATH=$PYTHONPATH:$(pwd)
export OPENAI_API_KEY=""
mkdir projectfiles
cd projectfiles
wget https://github.com/OpenAutoCoder/Agentless/releases/download/v0.1.0/swebench_lite_repo_structure.zip
unzip swebench_lite_repo_structure.zip
export PROJECT_FILE_LOC=/home/Ubuntu/projectfiles
mkdir results
python3 agentless/fl/localize.py --file_level --related_level --fine_grain_line_level --output_folder results/location --top_n 3 --compress --context_window=10 --temperature 0.8 --num_samples 4
python3 agentless/fl/localize.py --merge --output_folder results/location_merged --start_file results/location/loc_outputs.jsonl --num_samples 4