Precision annoy.AnnoyIndex with examples#
An example showing the AnnoyIndex class.
from __future__ import print_function
import random; random.seed(0)
import time
# from annoy import AnnoyIndex
# from scikitplot.annoy import AnnoyIndex
from scikitplot.annoy import Index as AnnoyIndex
try:
from tqdm.auto import tqdm, trange
except ImportError:
# Fallback: dummy versions that ignore all args/kwargs
tqdm = lambda iterable, *args, **kwargs: iterable
trange = lambda n, *args, **kwargs: range(n)
n, f = 1_000_000, 100 # 100~2.5GB
n, f = 100_000, 100 # 100~0.25GB 256~0.6GB
idx = AnnoyIndex(
f=f,
metric='angular',
)
idx.set_seed(0)
for i in trange(n):
if(i % (n//10) == 0): print(f"{i} / {n} = {1.0 * i / n}")
# v = []
# for z in range(f):
# v.append(random.gauss(0, 1))
v = [random.gauss(0, 1) for _ in range(f)]
idx.add_item(i, v)
idx.build(2 * f)
idx.save('test.annoy')
idx.info()
/home/circleci/repo/galleries/examples/annoy/plot_precision_script.py:37: UserWarning: seed=0 resets to Annoy's default seed
idx.set_seed(0)
0%| | 0/100000 [00:00<?, ?it/s]0 / 100000 = 0.0
1%|▏ | 1454/100000 [00:00<00:06, 14537.24it/s]
3%|▎ | 2908/100000 [00:00<00:06, 14401.84it/s]
4%|▍ | 4349/100000 [00:00<00:06, 14038.23it/s]
6%|▌ | 5817/100000 [00:00<00:06, 14283.31it/s]
7%|▋ | 7312/100000 [00:00<00:06, 14517.85it/s]
9%|▉ | 8780/100000 [00:00<00:06, 14570.57it/s]10000 / 100000 = 0.1
10%|█ | 10238/100000 [00:00<00:06, 14107.17it/s]
12%|█▏ | 11761/100000 [00:00<00:06, 14453.34it/s]
13%|█▎ | 13295/100000 [00:00<00:05, 14722.44it/s]
15%|█▍ | 14940/100000 [00:01<00:05, 15248.19it/s]
16%|█▋ | 16468/100000 [00:01<00:05, 14736.74it/s]
18%|█▊ | 18029/100000 [00:01<00:05, 14989.44it/s]
20%|█▉ | 19533/100000 [00:01<00:05, 14577.66it/s]20000 / 100000 = 0.2
21%|██ | 21187/100000 [00:01<00:05, 15146.49it/s]
23%|██▎ | 22707/100000 [00:01<00:05, 14533.83it/s]
24%|██▍ | 24169/100000 [00:01<00:05, 14229.45it/s]
26%|██▌ | 25598/100000 [00:01<00:05, 13937.14it/s]
27%|██▋ | 27043/100000 [00:01<00:05, 14082.40it/s]
29%|██▊ | 28639/100000 [00:01<00:04, 14625.38it/s]30000 / 100000 = 0.3
30%|███ | 30171/100000 [00:02<00:04, 14828.42it/s]
32%|███▏ | 31768/100000 [00:02<00:04, 15161.74it/s]
33%|███▎ | 33365/100000 [00:02<00:04, 15400.87it/s]
35%|███▍ | 34908/100000 [00:02<00:04, 15404.12it/s]
36%|███▋ | 36489/100000 [00:02<00:04, 15524.65it/s]
38%|███▊ | 38072/100000 [00:02<00:03, 15614.16it/s]
40%|███▉ | 39674/100000 [00:02<00:03, 15733.91it/s]40000 / 100000 = 0.4
41%|████▏ | 41258/100000 [00:02<00:03, 15764.04it/s]
43%|████▎ | 42842/100000 [00:02<00:03, 15786.57it/s]
44%|████▍ | 44430/100000 [00:02<00:03, 15811.76it/s]
46%|████▌ | 46012/100000 [00:03<00:03, 15634.42it/s]
48%|████▊ | 47607/100000 [00:03<00:03, 15727.87it/s]
49%|████▉ | 49204/100000 [00:03<00:03, 15799.11it/s]50000 / 100000 = 0.5
51%|█████ | 50798/100000 [00:03<00:03, 15838.70it/s]
52%|█████▏ | 52383/100000 [00:03<00:03, 15690.44it/s]
54%|█████▍ | 53977/100000 [00:03<00:02, 15763.92it/s]
56%|█████▌ | 55554/100000 [00:03<00:02, 15726.73it/s]
57%|█████▋ | 57167/100000 [00:03<00:02, 15843.99it/s]
59%|█████▉ | 58752/100000 [00:03<00:02, 15653.75it/s]60000 / 100000 = 0.6
60%|██████ | 60348/100000 [00:03<00:02, 15742.55it/s]
62%|██████▏ | 61923/100000 [00:04<00:02, 15644.37it/s]
64%|██████▎ | 63522/100000 [00:04<00:02, 15746.77it/s]
65%|██████▌ | 65184/100000 [00:04<00:02, 16005.91it/s]
67%|██████▋ | 66940/100000 [00:04<00:02, 16469.28it/s]
69%|██████▊ | 68692/100000 [00:04<00:01, 16782.31it/s]70000 / 100000 = 0.7
70%|███████ | 70442/100000 [00:04<00:01, 16995.85it/s]
72%|███████▏ | 72185/100000 [00:04<00:01, 17124.65it/s]
74%|███████▍ | 73926/100000 [00:04<00:01, 17208.72it/s]
76%|███████▌ | 75665/100000 [00:04<00:01, 17260.76it/s]
77%|███████▋ | 77392/100000 [00:04<00:01, 17028.41it/s]
79%|███████▉ | 79134/100000 [00:05<00:01, 17143.25it/s]80000 / 100000 = 0.8
81%|████████ | 80873/100000 [00:05<00:01, 17216.16it/s]
83%|████████▎ | 82615/100000 [00:05<00:01, 17276.23it/s]
84%|████████▍ | 84355/100000 [00:05<00:00, 17312.72it/s]
86%|████████▌ | 86092/100000 [00:05<00:00, 17328.76it/s]
88%|████████▊ | 87826/100000 [00:05<00:00, 17297.42it/s]
90%|████████▉ | 89556/100000 [00:05<00:00, 17280.74it/s]90000 / 100000 = 0.9
91%|█████████▏| 91285/100000 [00:05<00:00, 17270.47it/s]
93%|█████████▎| 93028/100000 [00:05<00:00, 17315.07it/s]
95%|█████████▍| 94766/100000 [00:05<00:00, 17332.50it/s]
97%|█████████▋| 96511/100000 [00:06<00:00, 17365.38it/s]
98%|█████████▊| 98249/100000 [00:06<00:00, 17368.62it/s]
100%|█████████▉| 99986/100000 [00:06<00:00, 17053.16it/s]
100%|██████████| 100000/100000 [00:06<00:00, 15876.87it/s]
{'f': 100, 'metric': 'angular', 'n_neighbors': 5, 'on_disk_path': 'test.annoy', 'prefault': False, 'seed': None, 'verbose': None, 'schema_version': 0, 'n_items': 100000, 'n_trees': 200, 'memory_usage_byte': 499984128, 'memory_usage_mib': 476.822021484375}
def plot(idx, y=None, **kwargs):
import numpy as np
import matplotlib.pyplot as plt
import scikitplot.cexternals._annoy._plotting as utils
single = np.zeros(idx.get_n_items(), dtype=int)
if y is None:
double = np.random.uniform(0, 1, idx.get_n_items()).round()
# single vs double
fig, ax = plt.subplots(ncols=2, figsize=(12, 5))
alpha = kwargs.pop("alpha", 0.8)
y2 = utils.plot_annoy_index(
idx,
dims = list(range(idx.f)),
plot_kwargs={"draw_legend": False},
ax=ax[0],
)[0]
utils.plot_annoy_knn_edges(
idx,
y2,
k=1,
line_kwargs={"alpha": alpha},
ax=ax[1],
)
# idx.unbuild()
# idx.build(10)
plot(idx)
def precision(q):
limits = [10, 100, 1_000, 10_000]
k = 10
prec_n = 10
prec_sum = {}
time_sum = {}
for i in trange(prec_n):
j = random.randrange(0, n)
closest = set(q.get_nns_by_item(j, k, n))
for limit in limits:
t0 = time.time()
toplist = q.get_nns_by_item(j, k, limit)
T = time.time() - t0
found = len(closest.intersection(toplist))
hitrate = 1.0 * found / k
prec_sum[limit] = prec_sum.get(limit, 0.0) + hitrate
time_sum[limit] = time_sum.get(limit, 0.0) + T
for limit in limits:
print('limit: %-9d precision: %6.2f%% avg time: %.6fs'
% (limit, 100.0 * prec_sum[limit] / (i + 1), time_sum[limit] / (i + 1)))
q = AnnoyIndex(f, 'angular')
q.set_seed(0)
q.load('test.annoy')
precision(q)
/home/circleci/repo/galleries/examples/annoy/plot_precision_script.py:111: UserWarning: seed=0 resets to Annoy's default seed
q.set_seed(0)
0%| | 0/10 [00:00<?, ?it/s]
70%|███████ | 7/10 [00:00<00:00, 61.81it/s]
100%|██████████| 10/10 [00:00<00:00, 63.04it/s]
limit: 10 precision: 10.00% avg time: 0.000119s
limit: 100 precision: 16.00% avg time: 0.000110s
limit: 1000 precision: 33.00% avg time: 0.000324s
limit: 10000 precision: 86.00% avg time: 0.001898s
Total running time of the script: (2 minutes 1.794 seconds)
Related examples
