umar-sharif821's picture
feat: reproducible eval + shared policy module; fix smart_agent wasted-capacity bug
ddf831c
name: CI
on:
push:
branches: [master, main]
pull_request:
branches: [master, main]
jobs:
smoke:
name: Syntax + Import Smoke Test
runs-on: ubuntu-latest
steps:
- name: Checkout
uses: actions/checkout@v4
- name: Set up Python
uses: actions/setup-python@v5
with:
python-version: "3.11"
- name: Install runtime dependencies
run: |
python -m pip install --upgrade pip
pip install -r requirements.txt
- name: Python syntax check (all tracked .py files)
run: |
python -m compileall -q env api server app.py colab_submission_script.py
- name: Import environment modules
run: |
python -c "from env.cache import CDNCacheEnv, TASK_CONFIGS; print('env.cache OK, tasks:', list(TASK_CONFIGS.keys()))"
python -c "from env.models import Action, Observation, FileEntry, Reward, StepResult, TaskConfig; print('env.models OK')"
- name: One-episode baseline smoke run
run: |
python - <<'PY'
from env.cache import CDNCacheEnv
from env.models import Action
env = CDNCacheEnv(task_id="task_easy", seed=0)
env.reset()
total = 0.0
for _ in range(20):
result = env.step(Action(evict_file_id=None))
total += result.reward.total
if result.done:
break
print(f"Smoke OK - 20 steps, reward sum={total:.3f}")
PY
- name: Reproducible evaluator (quick)
run: python scripts/eval.py --quick