初始化项目，由ModelHub XC社区提供模型

Model: pathcosmos/frankenstallm
Source: Original Platform

source/eval/setup_env.sh

@@ -0,0 +1,15 @@
+#!/bin/bash
+# lm-eval environment setup script
+# Sets up Python path and environment variables for lm-eval evaluation
+
+export PYTHONPATH="${HOME}/.local/lib/python3.12/site-packages:${PYTHONPATH}"
+export LM_EVAL_BIN="${HOME}/.local/bin/lm_eval"
+
+# Verify installation
+python -c "import lm_eval; print('✓ lm_eval', lm_eval.__version__)" || { echo "✗ lm_eval import failed"; exit 1; }
+python -c "import sacrebleu; print('✓ sacrebleu installed')" || { echo "✗ sacrebleu import failed"; exit 1; }
+python -c "import rouge_score; print('✓ rouge_score installed')" || { echo "✗ rouge_score import failed"; exit 1; }
+
+echo ""
+echo "Available Korean tasks:"
+$LM_EVAL_BIN ls tasks 2>&1 | grep -E "kobest|haerae|pawsx|global_mmlu_full_ko|mmmlu_ko"