scripts/dev/3_test_full.sh

#!/bin/bash
# Full test: 50 generations with eval service and WandB
#
# Prerequisites: Eval service should be running
#   bash scripts/dev/start_eval_server.sh

set -e

# ============================================================================
# Configuration
# ============================================================================
EXPERIMENT_NAME_PREFIX="mmv1_2"
NUM_GENERATIONS=200
MAX_PARALLEL_JOBS=4
META_INTERVAL=10


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"

USE_EVAL_SERVICE="--use-eval-service"
EVAL_SERVICE_URL="http://localhost:8765"
EVAL_TRIGGER_MODE="periodic"
EVAL_TRIGGER_INTERVAL=10

EXPERIMENT_NAME=${EXPERIMENT_NAME_PREFIX}_gen${NUM_GENERATIONS}_${EVAL_TRIGGER_MODE}${EVAL_TRIGGER_INTERVAL}


USE_WANDB="--use-wandb"
WANDB_PROJECT="ev2"
WANDB_TAGS="${TASK} eval-service full-experiment ${EVAL_TRIGGER_MODE}"

# ============================================================================
# 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 \
    $USE_EVAL_SERVICE \
    --eval-service-url "$EVAL_SERVICE_URL" \
    --eval-trigger-mode "$EVAL_TRIGGER_MODE" \
    --eval-trigger-interval "$EVAL_TRIGGER_INTERVAL" \
    $USE_WANDB \
    --wandb-project "$WANDB_PROJECT" \
    --wandb-tags $WANDB_TAGS \
    --verbose