scripts/dev/3b_ablation_no_eval_service.sh

#!/bin/bash
# Ablation: 50 generations WITHOUT eval service (baseline comparison)

set -e

# ============================================================================
# Configuration
# ============================================================================
EXPERIMENT_NAME_PREFIX="ablation"
NUM_GENERATIONS=50
MAX_PARALLEL_JOBS=4
META_INTERVAL=10

EXPERIMENT_NAME=${EXPERIMENT_NAME_PREFIX}_gen${NUM_GENERATIONS}

TASK="circle_packing"
LLM_MODELS="native-gemini-2.5-flash native-gemini-2.5-pro"
LLM_SELECTION="ucb1"
LLM_TEMPERATURES="0.5 0.7 1.0"

# NO eval service for this ablation
# USE_EVAL_SERVICE=""  # commented out

USE_WANDB="--use-wandb"
WANDB_PROJECT="ev2"
WANDB_TAGS="${TASK} ablation baseline no-eval-service"

# ============================================================================
# Run Experiment
# ============================================================================
python scripts/dev/run_experiment.py \
    --experiment-name "$EXPERIMENT_NAME" \
    --num-generations "$NUM_GENERATIONS" \
    --max-parallel-jobs "$MAX_PARALLEL_JOBS" \
    --meta-interval "$META_INTERVAL" \
    --task "$TASK" \
    --llm-models $LLM_MODELS \
    --llm-selection "$LLM_SELECTION" \
    --llm-temperatures $LLM_TEMPERATURES \
    $USE_WANDB \
    --wandb-project "$WANDB_PROJECT" \
    --wandb-tags $WANDB_TAGS \
    --verbose

echo ""
echo "Note: This is a baseline run WITHOUT eval service for comparison."