[inventory_q.md] Update np.random → Generator API

[Chihiro2000GitHub]

Summary

This PR migrates legacy NumPy random API usage in inventory_q.md as part of QuantEcon/meta#299.

All of the random draws happen inside @numba.jit(nopython=True) functions, so the two of them now take a rng argument and the generator is built by the calling Python code.

Related PRs and issues

I checked for open PRs and issues related to this lecture. No open PR modifies inventory_q.md. Issues #881 and #882 concern other aspects of this lecture (the demand grid and the Q-learning loop), and this PR does not attempt to address them.

Details

• sim_inventories takes rng in place of seed, and draws with rng.geometric(p).
• q_learning_kernel takes rng in place of seed, and draws with rng.geometric(p), rng.integers(...) and rng.random().
• The q_learning wrapper is ordinary Python, so it keeps its seed=1234 argument and builds the generator itself. Its call site is unchanged.
• The existing seeds (0, 1234, 5678) are all kept, and no new seed was introduced.
• The comparison figure builds a fresh np.random.default_rng(sim_seed) for each call, so the optimal policy and the Q-learning snapshots still face the same demand sequence, as they did when each call re-seeded.
• Numba supports geometric, integers and random on a Generator passed in as an argument, but does not support constructing one inside a jitted function, which is why the generator is built by the caller. There is no prange or parallel=True in this lecture, so a single generator is safe here. On a 5-million-iteration benchmark the Generator version ran at the same speed as the legacy calls.
• No prose changes were needed.
• A full local build completed successfully and inventory_q.md executed without errors.

Note for reviewers

Both jitted functions take rng in place of seed, which changes their signatures. q_learning keeps its seed argument so the call in the lecture is unchanged, but sim_inventories is called directly and now receives a generator at each call site.

Hi @mmcky and @HumphreyYang, I'd be grateful if you could take a look when you have time.

[github-actions]

📖 Netlify Preview Ready!

Preview URL: https://pr-1012--sunny-cactus-210e3e.netlify.app

Commit: c98c2fb

📚 Changed Lectures

• inventory_q

---

Build Info

• Workflow: Build Project [using jupyter-book]