neural-mesh-v2 / evaluation /eval_setting.sh
hjkim00's picture
Restore all essential files - code, configs, and MBPP/HumanEval data
24c2665 verified
#!/bin/bash
# 1. κ°€μƒν™˜κ²½ 생성 및 ν™œμ„±ν™”
# conda ν™˜κ²½ 생성
conda create -n azr_eval python=3.10.16 -y
# ν™˜κ²½ ν™œμ„±ν™”
conda activate azr_eval
# 2. latex2sympy μ„€μΉ˜
# AZR λ””λ ‰ν† λ¦¬λ‘œ 이동
cd /home/ubuntu/RLVR/Absolute-Zero-Reasoner/evaluation/math_eval
# latex2sympy μ••μΆ• ν•΄μ œ (eval λ””λ ‰ν† λ¦¬μ—μ„œ)
cd eval
tar -xzvf latex2sympy.tar.gz
# latex2sympy μ„€μΉ˜
cd latex2sympy
pip install -e .
# 3. 평가 μ˜μ‘΄μ„± μ„€μΉ˜
# math_eval λ””λ ‰ν† λ¦¬λ‘œ λŒμ•„κ°€μ„œ
cd /home/ubuntu/RLVR/Absolute-Zero-Reasoner/evaluation/math_eval
# μ˜μ‘΄μ„± μ„€μΉ˜
pip install -r requirements.txt
# flash-attention μ„€μΉ˜
pip install flash_attn==2.7.4.post1
# 4. 평가 슀크립트 μ‹€ν–‰
# μ™„μ „ν•œ 평가 슀크립트 μ‹€ν–‰
cd /home/ubuntu/RLVR/Absolute-Zero-Reasoner
bash evaluation/run_complete_evaluation.sh