run_trick_sweep.sh

timesteps=10000000
env_id="Breakout-v5 MsPacman-v5 Assault-v5 Zaxxon-v5 Tennis-v5"
seeds=3
workers=1
gittag=$(git describe --tags)

for (( startseed=1; startseed<=$seeds; startseed++ ))
do

    poetry run python -m cleanrl_utils.benchmark \
        --env-ids $env_id \
        --command "poetry run python ppo_v3/ppo_envpool_tricks.py --exp-name ppo_envpool_tricks_none --total-timesteps $timesteps --track" \
        --start-seed $startseed \
        --num-seeds 1 \
        --workers $workers

    poetry run python -m cleanrl_utils.benchmark \
        --env-ids $env_id \
        --command "poetry run python ppo_v3/ppo_envpool_tricks.py --exp-name ppo_envpool_tricks_all --symlog True --two-hot True --percentile-scale True --critic-ema --unimix 0.01 --critic-zero-init True --total-timesteps $timesteps --track" \
        --start-seed $startseed \
        --num-seeds 1 \
        --workers $workers

    poetry run python -m cleanrl_utils.benchmark \
        --env-ids $env_id \
        --command "poetry run python ppo_v3/ppo_envpool_tricks.py --exp-name ppo_envpool_tricks_symlog --symlog True --total-timesteps $timesteps --track" \
        --start-seed $startseed \
        --num-seeds 1 \
        --workers $workers

    poetry run python -m cleanrl_utils.benchmark \
        --env-ids $env_id \
        --command "poetry run python ppo_v3/ppo_envpool_tricks.py --exp-name ppo_envpool_tricks_twohot --two-hot True --total-timesteps $timesteps --track" \
        --start-seed $startseed \
        --num-seeds 1 \
        --workers $workers

    poetry run python -m cleanrl_utils.benchmark \
        --env-ids $env_id \
        --command "poetry run python ppo_v3/ppo_envpool_tricks.py --exp-name ppo_envpool_tricks_percentile --percentile-scale True --total-timesteps $timesteps --track" \
        --start-seed $startseed \
        --num-seeds 1 \
        --workers $workers

    poetry run python -m cleanrl_utils.benchmark \
        --env-ids $env_id \
        --command "poetry run python ppo_v3/ppo_envpool_tricks.py --exp-name ppo_envpool_tricks_twohot_criticema --two-hot True --critic-ema True --total-timesteps $timesteps --track" \
        --start-seed $startseed \
        --num-seeds 1 \
        --workers $workers

    poetry run python -m cleanrl_utils.benchmark \
        --env-ids $env_id \
        --command "poetry run python ppo_v3/ppo_envpool_tricks.py --exp-name ppo_envpool_tricks_unimix --unimix 0.01 --total-timesteps $timesteps --track" \
        --start-seed $startseed \
        --num-seeds 1 \
        --workers $workers

    poetry run python -m cleanrl_utils.benchmark \
        --env-ids $env_id \
        --command "poetry run python ppo_v3/ppo_envpool_tricks.py --exp-name ppo_envpool_tricks_zeroinit --critic-zero-init True --total-timesteps $timesteps --track" \
        --start-seed $startseed \
        --num-seeds 1 \
        --workers $workers
done

poetry run python -m openrlbenchmark.rlops \
    --filters '?we=ryan-colab&wpn=PPO-v3&ceik=env_id&cen=exp_name&metric=charts/human_normalized_score' \
    "ppo_envpool_tricks_none?tag=${gittag}" \
    "ppo_envpool_tricks_symlog?tag=${gittag}" \
    "ppo_envpool_tricks_twohot?tag=${gittag}" \
    "ppo_envpool_tricks_percentile?tag=${gittag}" \
    "ppo_envpool_tricks_twohot_criticema?tag=${gittag}" \
    "ppo_envpool_tricks_unimix?tag=${gittag}" \
    "ppo_envpool_tricks_zeroinit?tag=${gittag}" \
    "ppo_envpool_tricks_all?tag=${gittag}" \
    --env-ids $env_id \
    --ncols 2 \
    --ncols-legend 2 \
    --output-filename static/cleanrl_vs_baselines \
    --scan-history \
    --report