csbc/EVAL.md

csbc — evaluation

CNN + discrete-mouse BC baseline. Weights: steps_5000.pt (best eval), steps_15000.pt (final). Full guide + prerequisites: ../EVALUATION.md.

Run (from inside this folder; game server + client already up — see EVALUATION.md steps 1–2):

mkdir -p run/checkpoints && cp config.yaml dataset_statistics.json run/ && cp steps_5000.pt run/checkpoints/
python -m gamevla.envs.eval.run_eval \
  --env cs2/5e_mirage_prefire/connector_to_a_site --endpoint <host:port> \
  --policy examples.CS2.csbc.evaluation.policy:build_policy \
  --policy-kwargs '{"ckpt":"run/checkpoints/steps_5000.pt","port":10093,"sensitivity":1.0}' \
  --episodes 3 --bot-count 5

Defaults (policy.py): steps_per_bin=2, history_len=1. Mouse auto-derived from sensitivity; add "invert_dx":true / "invert_dy":true if a flick turns the wrong way.