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
2%|▏ | 1641/100000 [00:00<00:05, 16402.12it/s]
3%|▎ | 3282/100000 [00:00<00:06, 15292.22it/s]
5%|▍ | 4847/100000 [00:00<00:06, 15447.04it/s]
7%|▋ | 6504/100000 [00:00<00:05, 15878.01it/s]
8%|▊ | 8095/100000 [00:00<00:05, 15617.16it/s]
10%|▉ | 9750/100000 [00:00<00:05, 15924.37it/s]10000 / 100000 = 0.1
11%|█▏ | 11392/100000 [00:00<00:05, 16083.04it/s]
13%|█▎ | 13094/100000 [00:00<00:05, 16376.34it/s]
15%|█▍ | 14835/100000 [00:00<00:05, 16696.69it/s]
17%|█▋ | 16506/100000 [00:01<00:05, 16466.69it/s]
18%|█▊ | 18155/100000 [00:01<00:05, 16236.85it/s]
20%|█▉ | 19781/100000 [00:01<00:04, 16188.85it/s]20000 / 100000 = 0.2
21%|██▏ | 21418/100000 [00:01<00:04, 16240.77it/s]
23%|██▎ | 23162/100000 [00:01<00:04, 16596.77it/s]
25%|██▍ | 24875/100000 [00:01<00:04, 16756.05it/s]
27%|██▋ | 26561/100000 [00:01<00:04, 16786.59it/s]
28%|██▊ | 28241/100000 [00:01<00:04, 16017.21it/s]
30%|██▉ | 29851/100000 [00:01<00:04, 15847.76it/s]30000 / 100000 = 0.3
31%|███▏ | 31441/100000 [00:01<00:04, 15776.25it/s]
33%|███▎ | 33023/100000 [00:02<00:04, 15318.54it/s]
35%|███▍ | 34560/100000 [00:02<00:04, 15295.29it/s]
36%|███▋ | 36255/100000 [00:02<00:04, 15774.62it/s]
38%|███▊ | 37997/100000 [00:02<00:03, 16255.89it/s]
40%|███▉ | 39742/100000 [00:02<00:03, 16608.24it/s]40000 / 100000 = 0.4
41%|████▏ | 41488/100000 [00:02<00:03, 16859.20it/s]
43%|████▎ | 43233/100000 [00:02<00:03, 17032.68it/s]
45%|████▍ | 44977/100000 [00:02<00:03, 17151.56it/s]
47%|████▋ | 46694/100000 [00:02<00:03, 17052.75it/s]
48%|████▊ | 48439/100000 [00:02<00:03, 17169.66it/s]50000 / 100000 = 0.5
50%|█████ | 50183/100000 [00:03<00:02, 17248.53it/s]
52%|█████▏ | 51928/100000 [00:03<00:02, 17307.93it/s]
54%|█████▎ | 53660/100000 [00:03<00:02, 17260.73it/s]
55%|█████▌ | 55399/100000 [00:03<00:02, 17296.83it/s]
57%|█████▋ | 57146/100000 [00:03<00:02, 17347.55it/s]
59%|█████▉ | 58881/100000 [00:03<00:02, 17128.56it/s]60000 / 100000 = 0.6
61%|██████ | 60623/100000 [00:03<00:02, 17212.79it/s]
62%|██████▏ | 62363/100000 [00:03<00:02, 17266.18it/s]
64%|██████▍ | 64092/100000 [00:03<00:02, 17270.31it/s]
66%|██████▌ | 65835/100000 [00:03<00:01, 17315.77it/s]
68%|██████▊ | 67570/100000 [00:04<00:01, 17324.45it/s]
69%|██████▉ | 69303/100000 [00:04<00:01, 17044.21it/s]70000 / 100000 = 0.7
71%|███████ | 71009/100000 [00:04<00:01, 15844.08it/s]
73%|███████▎ | 72739/100000 [00:04<00:01, 16254.15it/s]
74%|███████▍ | 74479/100000 [00:04<00:01, 16583.10it/s]
76%|███████▌ | 76149/100000 [00:04<00:01, 16486.74it/s]
78%|███████▊ | 77886/100000 [00:04<00:01, 16741.83it/s]
80%|███████▉ | 79615/100000 [00:04<00:01, 16900.45it/s]80000 / 100000 = 0.8
81%|████████▏ | 81335/100000 [00:04<00:01, 16987.95it/s]
83%|████████▎ | 83057/100000 [00:05<00:00, 17054.10it/s]
85%|████████▍ | 84765/100000 [00:05<00:00, 17060.04it/s]
86%|████████▋ | 86487/100000 [00:05<00:00, 17105.65it/s]
88%|████████▊ | 88207/100000 [00:05<00:00, 17130.92it/s]
90%|████████▉ | 89923/100000 [00:05<00:00, 17139.02it/s]90000 / 100000 = 0.9
92%|█████████▏| 91646/100000 [00:05<00:00, 17164.57it/s]
93%|█████████▎| 93363/100000 [00:05<00:00, 17101.86it/s]
95%|█████████▌| 95074/100000 [00:05<00:00, 15699.81it/s]
97%|█████████▋| 96667/100000 [00:05<00:00, 14776.96it/s]
98%|█████████▊| 98168/100000 [00:05<00:00, 14751.69it/s]
100%|█████████▉| 99660/100000 [00:06<00:00, 13678.61it/s]
100%|██████████| 100000/100000 [00:06<00:00, 16360.23it/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]
60%|██████ | 6/10 [00:00<00:00, 52.64it/s]
100%|██████████| 10/10 [00:00<00:00, 54.21it/s]
limit: 10 precision: 10.00% avg time: 0.000153s
limit: 100 precision: 16.00% avg time: 0.000131s
limit: 1000 precision: 33.00% avg time: 0.000393s
limit: 10000 precision: 86.00% avg time: 0.002349s
Total running time of the script: (2 minutes 14.487 seconds)
Related examples
