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%|▏ | 1264/100000 [00:00<00:07, 12635.46it/s]
3%|▎ | 2528/100000 [00:00<00:07, 12599.45it/s]
4%|▍ | 3788/100000 [00:00<00:07, 12570.54it/s]
5%|▌ | 5052/100000 [00:00<00:07, 12596.63it/s]
6%|▋ | 6321/100000 [00:00<00:07, 12629.39it/s]
8%|▊ | 7589/100000 [00:00<00:07, 12644.91it/s]
9%|▉ | 8854/100000 [00:00<00:07, 12630.54it/s]10000 / 100000 = 0.1
10%|█ | 10118/100000 [00:00<00:07, 12631.17it/s]
11%|█▏ | 11382/100000 [00:00<00:07, 12633.48it/s]
13%|█▎ | 12646/100000 [00:01<00:06, 12575.68it/s]
14%|█▍ | 13916/100000 [00:01<00:06, 12611.23it/s]
15%|█▌ | 15178/100000 [00:01<00:06, 12553.32it/s]
16%|█▋ | 16434/100000 [00:01<00:06, 12548.65it/s]
18%|█▊ | 17689/100000 [00:01<00:06, 12548.82it/s]
19%|█▉ | 18944/100000 [00:01<00:06, 12521.22it/s]20000 / 100000 = 0.2
20%|██ | 20199/100000 [00:01<00:06, 12526.94it/s]
21%|██▏ | 21452/100000 [00:01<00:06, 12490.71it/s]
23%|██▎ | 22715/100000 [00:01<00:06, 12530.89it/s]
24%|██▍ | 23969/100000 [00:01<00:06, 12523.65it/s]
25%|██▌ | 25230/100000 [00:02<00:05, 12546.92it/s]
26%|██▋ | 26496/100000 [00:02<00:05, 12578.58it/s]
28%|██▊ | 27754/100000 [00:02<00:05, 12558.27it/s]
29%|██▉ | 29010/100000 [00:02<00:05, 12547.70it/s]30000 / 100000 = 0.3
30%|███ | 30265/100000 [00:02<00:05, 12544.74it/s]
32%|███▏ | 31520/100000 [00:02<00:05, 12537.45it/s]
33%|███▎ | 32777/100000 [00:02<00:05, 12545.93it/s]
34%|███▍ | 34032/100000 [00:02<00:05, 12311.56it/s]
35%|███▌ | 35265/100000 [00:02<00:05, 12286.43it/s]
37%|███▋ | 36523/100000 [00:02<00:05, 12370.78it/s]
38%|███▊ | 37778/100000 [00:03<00:05, 12423.09it/s]
39%|███▉ | 39028/100000 [00:03<00:04, 12443.78it/s]40000 / 100000 = 0.4
40%|████ | 40273/100000 [00:03<00:04, 12362.88it/s]
42%|████▏ | 41510/100000 [00:03<00:04, 12098.02it/s]
43%|████▎ | 42722/100000 [00:03<00:04, 11972.39it/s]
44%|████▍ | 43949/100000 [00:03<00:04, 12056.94it/s]
45%|████▌ | 45188/100000 [00:03<00:04, 12155.16it/s]
46%|████▋ | 46438/100000 [00:03<00:04, 12257.14it/s]
48%|████▊ | 47686/100000 [00:03<00:04, 12321.31it/s]
49%|████▉ | 48938/100000 [00:03<00:04, 12379.11it/s]50000 / 100000 = 0.5
50%|█████ | 50199/100000 [00:04<00:04, 12445.44it/s]
51%|█████▏ | 51456/100000 [00:04<00:03, 12481.21it/s]
53%|█████▎ | 52717/100000 [00:04<00:03, 12517.39it/s]
54%|█████▍ | 53976/100000 [00:04<00:03, 12538.57it/s]
55%|█████▌ | 55232/100000 [00:04<00:03, 12544.03it/s]
56%|█████▋ | 56487/100000 [00:04<00:03, 12506.15it/s]
58%|█████▊ | 57738/100000 [00:04<00:03, 12498.70it/s]
59%|█████▉ | 58988/100000 [00:04<00:03, 12481.01it/s]60000 / 100000 = 0.6
60%|██████ | 60243/100000 [00:04<00:03, 12498.70it/s]
62%|██████▏ | 61507/100000 [00:04<00:03, 12540.27it/s]
63%|██████▎ | 62772/100000 [00:05<00:02, 12572.58it/s]
64%|██████▍ | 64046/100000 [00:05<00:02, 12621.51it/s]
65%|██████▌ | 65310/100000 [00:05<00:02, 12626.13it/s]
67%|██████▋ | 66573/100000 [00:05<00:02, 12610.00it/s]
68%|██████▊ | 67835/100000 [00:05<00:02, 12599.81it/s]
69%|██████▉ | 69096/100000 [00:05<00:02, 12600.20it/s]70000 / 100000 = 0.7
70%|███████ | 70357/100000 [00:05<00:02, 12592.45it/s]
72%|███████▏ | 71617/100000 [00:05<00:02, 12591.31it/s]
73%|███████▎ | 72877/100000 [00:05<00:02, 12583.14it/s]
74%|███████▍ | 74139/100000 [00:05<00:02, 12591.98it/s]
75%|███████▌ | 75404/100000 [00:06<00:01, 12606.47it/s]
77%|███████▋ | 76665/100000 [00:06<00:01, 12546.62it/s]
78%|███████▊ | 77920/100000 [00:06<00:01, 12527.79it/s]
79%|███████▉ | 79176/100000 [00:06<00:01, 12537.07it/s]80000 / 100000 = 0.8
80%|████████ | 80434/100000 [00:06<00:01, 12547.85it/s]
82%|████████▏ | 81692/100000 [00:06<00:01, 12555.07it/s]
83%|████████▎ | 82955/100000 [00:06<00:01, 12577.37it/s]
84%|████████▍ | 84213/100000 [00:06<00:01, 12565.26it/s]
85%|████████▌ | 85470/100000 [00:06<00:01, 12553.37it/s]
87%|████████▋ | 86727/100000 [00:06<00:01, 12558.25it/s]
88%|████████▊ | 87987/100000 [00:07<00:00, 12568.98it/s]
89%|████████▉ | 89247/100000 [00:07<00:00, 12576.74it/s]90000 / 100000 = 0.9
91%|█████████ | 90505/100000 [00:07<00:00, 12548.82it/s]
92%|█████████▏| 91766/100000 [00:07<00:00, 12565.69it/s]
93%|█████████▎| 93029/100000 [00:07<00:00, 12582.53it/s]
94%|█████████▍| 94288/100000 [00:07<00:00, 12581.97it/s]
96%|█████████▌| 95554/100000 [00:07<00:00, 12603.45it/s]
97%|█████████▋| 96815/100000 [00:07<00:00, 12593.35it/s]
98%|█████████▊| 98077/100000 [00:07<00:00, 12599.31it/s]
99%|█████████▉| 99337/100000 [00:07<00:00, 12533.11it/s]
100%|██████████| 100000/100000 [00:07<00:00, 12511.90it/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
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:110: UserWarning: seed=0 resets to Annoy's default seed
q.set_seed(0)
0%| | 0/10 [00:00<?, ?it/s]limit: 10 precision: 10.00% avg time: 0.000111s
limit: 100 precision: 10.00% avg time: 0.000136s
limit: 1000 precision: 20.00% avg time: 0.000322s
limit: 10000 precision: 80.00% avg time: 0.002099s
limit: 10 precision: 10.00% avg time: 0.000116s
limit: 100 precision: 10.00% avg time: 0.000158s
limit: 1000 precision: 20.00% avg time: 0.000362s
limit: 10000 precision: 85.00% avg time: 0.002066s
limit: 10 precision: 10.00% avg time: 0.000123s
limit: 100 precision: 10.00% avg time: 0.000143s
limit: 1000 precision: 26.67% avg time: 0.000367s
limit: 10000 precision: 90.00% avg time: 0.002067s
limit: 10 precision: 10.00% avg time: 0.000136s
limit: 100 precision: 10.00% avg time: 0.000157s
limit: 1000 precision: 27.50% avg time: 0.000367s
limit: 10000 precision: 80.00% avg time: 0.002050s
limit: 10 precision: 10.00% avg time: 0.000127s
limit: 100 precision: 10.00% avg time: 0.000150s
limit: 1000 precision: 26.00% avg time: 0.000361s
limit: 10000 precision: 84.00% avg time: 0.002033s
limit: 10 precision: 10.00% avg time: 0.000125s
limit: 100 precision: 11.67% avg time: 0.000144s
limit: 1000 precision: 26.67% avg time: 0.000354s
limit: 10000 precision: 81.67% avg time: 0.002012s
60%|██████ | 6/10 [00:00<00:00, 56.23it/s]limit: 10 precision: 10.00% avg time: 0.000132s
limit: 100 precision: 17.14% avg time: 0.000153s
limit: 1000 precision: 34.29% avg time: 0.000380s
limit: 10000 precision: 84.29% avg time: 0.002044s
limit: 10 precision: 10.00% avg time: 0.000133s
limit: 100 precision: 17.50% avg time: 0.000153s
limit: 1000 precision: 33.75% avg time: 0.000377s
limit: 10000 precision: 85.00% avg time: 0.002042s
limit: 10 precision: 10.00% avg time: 0.000134s
limit: 100 precision: 16.67% avg time: 0.000151s
limit: 1000 precision: 34.44% avg time: 0.000370s
limit: 10000 precision: 85.56% avg time: 0.002024s
limit: 10 precision: 10.00% avg time: 0.000136s
limit: 100 precision: 16.00% avg time: 0.000152s
limit: 1000 precision: 33.00% avg time: 0.000367s
limit: 10000 precision: 86.00% avg time: 0.002027s
100%|██████████| 10/10 [00:00<00:00, 56.53it/s]
Total running time of the script: (2 minutes 9.321 seconds)
Related examples
