Skip to content

Instantly share code, notes, and snippets.

@amirziai

amirziai/ch_part2_slices.py

Created September 25, 2022 22:35
Show Gist options

  • Select an option

    Learn more about clone URLs
  • Save amirziai/f3e33572a76b3a89c6ba25b7e2a3ef4c to your computer and use it in GitHub Desktop.

Select an option

Learn more about clone URLs
Save amirziai/f3e33572a76b3a89c6ba25b7e2a3ef4c to your computer and use it in GitHub Desktop.
Download ZIP

Revisions

  1. amirziai created this gist Sep 25, 2022.
    33 changes: 33 additions & 0 deletions ch_part2_slices.py
    View file
    Original file line number Diff line number Diff line change
    @@ -0,0 +1,33 @@
    from collections import defaultdict


    def compute_slices(
    n: int,
    k: int,
    seed: Optional[int] = None,
    ) -> List[float]:
    np.random.seed(seed)
    tokens = np.random.rand(n * k)
    servers = np.random.choice(range(n), n * k)
    idxs = np.argsort(tokens)
    fractions = defaultdict(int)
    for idx_prev, idx in zip(idxs, idxs[1:]):
    server_prev = servers[idx_prev]
    diff = tokens[idx] - tokens[idx_prev]
    fractions[server_prev] += diff

    last = 1 - tokens[idxs[-1]]
    last += tokens[idxs[0]]
    fractions[servers[-1]] += last
    assert sum(fractions.values()) == 1
    return fractions

    df_slices = pd.DataFrame(
    dict(
    k=k,
    exp_idx=exp_idx,
    server1=compute_slices(n=5, k=k, seed=exp_idx)[0],
    )
    for k in (1, 5, 10, 100, 1000, 10_000)
    for exp_idx in range(100)
    ).fillna(0)