faiss/benchs/bench_all_ivf/bench_all_ivf.py

567 lines
19 KiB
Python

# Copyright (c) Facebook, Inc. and its affiliates.
#
# This source code is licensed under the MIT license found in the
# LICENSE file in the root directory of this source tree.
import argparse
import os
import sys
import time
import json
import faiss
import numpy as np
try:
import datasets_fb as datasets
except ModuleNotFoundError:
import datasets_oss as datasets
sanitize = datasets.sanitize
def unwind_index_ivf(index):
if isinstance(index, faiss.IndexPreTransform):
assert index.chain.size() == 1
vt = index.chain.at(0)
index_ivf, vt2 = unwind_index_ivf(faiss.downcast_index(index.index))
assert vt2 is None
if vt is None:
vt = lambda x: x
else:
vt = faiss.downcast_VectorTransform(vt)
return index_ivf, vt
if hasattr(faiss, "IndexRefine") and isinstance(index, faiss.IndexRefine):
return unwind_index_ivf(faiss.downcast_index(index.base_index))
if isinstance(index, faiss.IndexIVF):
return index, None
else:
return None, None
def apply_AQ_options(index, args):
# if not(
# isinstance(index, faiss.IndexAdditiveQuantize) or
# isinstance(index, faiss.IndexIVFAdditiveQuantizer)):
# return
if args.RQ_train_default:
print("set default training for RQ")
index.rq.train_type
index.rq.train_type = faiss.ResidualQuantizer.Train_default
if args.RQ_beam_size != -1:
print("set RQ beam size to", args.RQ_beam_size)
index.rq.max_beam_size
index.rq.max_beam_size = args.RQ_beam_size
if args.LSQ_encode_ils_iters != -1:
print("set LSQ ils iterations to", args.LSQ_encode_ils_iters)
index.lsq.encode_ils_iters
index.lsq.encode_ils_iters = args.LSQ_encode_ils_iters
if args.RQ_use_beam_LUT != -1:
print("set RQ beam LUT to", args.RQ_use_beam_LUT)
index.rq.use_beam_LUT
index.rq.use_beam_LUT = args.RQ_use_beam_LUT
def eval_setting(index, xq, gt, k, inter, min_time):
""" evaluate searching in terms of precision vs. speed """
nq = xq.shape[0]
ivf_stats = faiss.cvar.indexIVF_stats
ivf_stats.reset()
nrun = 0
t0 = time.time()
while True:
D, I = index.search(xq, k)
nrun += 1
t1 = time.time()
if t1 - t0 > min_time:
break
ms_per_query = ((t1 - t0) * 1000.0 / nq / nrun)
res = {
"ms_per_query": ms_per_query,
"nrun": nrun
}
res["n"] = ms_per_query
if inter:
rank = k
inter_measure = faiss.eval_intersection(gt[:, :rank], I[:, :rank]) / (nq * rank)
print("%.4f" % inter_measure, end=' ')
res["inter_measure"] = inter_measure
else:
res["recalls"] = {}
for rank in 1, 10, 100:
recall = (I[:, :rank] == gt[:, :1]).sum() / float(nq)
print("%.4f" % recall, end=' ')
res["recalls"][rank] = recall
print(" %9.5f " % ms_per_query, end=' ')
print("%12d " % (ivf_stats.ndis / nrun), end=' ')
print(nrun)
res["ndis"] = ivf_stats.ndis / nrun
return res
######################################################
# Training
######################################################
def run_train(args, ds, res):
nq, d = ds.nq, ds.d
nb, d = ds.nq, ds.d
print("build index, key=", args.indexkey)
index = faiss.index_factory(
d, args.indexkey, faiss.METRIC_L2 if ds.metric == "L2" else
faiss.METRIC_INNER_PRODUCT
)
index_ivf, vec_transform = unwind_index_ivf(index)
if args.by_residual != -1:
by_residual = args.by_residual == 1
print("setting by_residual = ", by_residual)
index_ivf.by_residual # check if field exists
index_ivf.by_residual = by_residual
if index_ivf:
print("Update add-time parameters")
# adjust default parameters used at add time for quantizers
# because otherwise the assignment is inaccurate
quantizer = faiss.downcast_index(index_ivf.quantizer)
if isinstance(quantizer, faiss.IndexRefine):
print(" update quantizer k_factor=", quantizer.k_factor, end=" -> ")
quantizer.k_factor = 32 if index_ivf.nlist < 1e6 else 64
print(quantizer.k_factor)
base_index = faiss.downcast_index(quantizer.base_index)
if isinstance(base_index, faiss.IndexIVF):
print(" update quantizer nprobe=", base_index.nprobe, end=" -> ")
base_index.nprobe = (
16 if base_index.nlist < 1e5 else
32 if base_index.nlist < 4e6 else
64)
print(base_index.nprobe)
elif isinstance(quantizer, faiss.IndexHNSW):
hnsw = quantizer.hnsw
print(
f" update HNSW quantizer options, before: "
f"{hnsw.efSearch=:} {hnsw.efConstruction=:}"
)
hnsw.efSearch = 40 if index_ivf.nlist < 4e6 else 64
hnsw.efConstruction = 200
print(f" after: {hnsw.efSearch=:} {hnsw.efConstruction=:}")
apply_AQ_options(index_ivf or index, args)
if index_ivf:
index_ivf.verbose = True
index_ivf.quantizer.verbose = True
index_ivf.cp.verbose = True
else:
index.verbose = True
maxtrain = args.maxtrain
if maxtrain == 0:
if 'IMI' in args.indexkey:
maxtrain = int(256 * 2 ** (np.log2(index_ivf.nlist) / 2))
elif index_ivf:
maxtrain = 50 * index_ivf.nlist
else:
# just guess...
maxtrain = 256 * 100
maxtrain = max(maxtrain, 256 * 100)
print("setting maxtrain to %d" % maxtrain)
try:
xt2 = ds.get_train(maxtrain=maxtrain)
except NotImplementedError:
print("No training set: training on database")
xt2 = ds.get_database()[:maxtrain]
print("train, size", xt2.shape)
assert np.all(np.isfinite(xt2))
if (isinstance(vec_transform, faiss.OPQMatrix) and
isinstance(index_ivf, faiss.IndexIVFPQFastScan)):
print(" Forcing OPQ training PQ to PQ4")
ref_pq = index_ivf.pq
training_pq = faiss.ProductQuantizer(
ref_pq.d, ref_pq.M, ref_pq.nbits
)
vec_transform.pq
vec_transform.pq = training_pq
if args.get_centroids_from == '':
if args.clustering_niter >= 0:
print(("setting nb of clustering iterations to %d" %
args.clustering_niter))
index_ivf.cp.niter = args.clustering_niter
if args.train_on_gpu:
print("add a training index on GPU")
train_index = faiss.index_cpu_to_all_gpus(
faiss.IndexFlatL2(index_ivf.d))
index_ivf.clustering_index = train_index
else:
print("Getting centroids from", args.get_centroids_from)
src_index = faiss.read_index(args.get_centroids_from)
src_quant = faiss.downcast_index(src_index.quantizer)
centroids = src_quant.reconstruct_n()
print(" centroid table shape", centroids.shape)
if isinstance(vec_transform, faiss.VectorTransform):
print(" training vector transform")
vec_transform.train(xt2)
print(" transform centroids")
centroids = vec_transform.apply_py(centroids)
if not index_ivf.quantizer.is_trained:
print(" training quantizer")
index_ivf.quantizer.train(centroids)
print(" add centroids to quantizer")
index_ivf.quantizer.add(centroids)
del src_index
t0 = time.time()
index.train(xt2)
res.train_time = time.time() - t0
print(" train in %.3f s" % res.train_time)
return index
######################################################
# Populating index
######################################################
def run_add(args, ds, index, res):
print("adding")
t0 = time.time()
if args.add_bs == -1:
assert args.split == [1, 0], "split not supported with full batch add"
index.add(sanitize(ds.get_database()))
else:
totn = ds.nb // args.split[0] # approximate
i0 = 0
print(f"Adding in block sizes {args.add_bs} with split {args.split}")
for xblock in ds.database_iterator(bs=args.add_bs, split=args.split):
i1 = i0 + len(xblock)
print(" adding %d:%d / %d [%.3f s, RSS %d kiB] " % (
i0, i1, totn, time.time() - t0,
faiss.get_mem_usage_kb()))
index.add(xblock)
i0 = i1
res.t_add = time.time() - t0
print(f" add in {res.t_add:.3f} s index size {index.ntotal}")
######################################################
# Search
######################################################
def run_search(args, ds, index, res):
index_ivf, vec_transform = unwind_index_ivf(index)
if args.no_precomputed_tables:
if isinstance(index_ivf, faiss.IndexIVFPQ):
print("disabling precomputed table")
index_ivf.use_precomputed_table = -1
index_ivf.precomputed_table.clear()
if args.indexfile:
print("index size on disk: ", os.stat(args.indexfile).st_size)
if hasattr(index, "code_size"):
print("vector code_size", index.code_size)
if hasattr(index_ivf, "code_size"):
print("vector code_size (IVF)", index_ivf.code_size)
print("current RSS:", faiss.get_mem_usage_kb() * 1024)
precomputed_table_size = 0
if hasattr(index_ivf, 'precomputed_table'):
precomputed_table_size = index_ivf.precomputed_table.size() * 4
print("precomputed tables size:", precomputed_table_size)
# Index is ready
xq = sanitize(ds.get_queries())
nq, d = xq.shape
gt = ds.get_groundtruth(k=args.k)
if not args.accept_short_gt: # Deep1B has only a single NN per query
assert gt.shape[1] == args.k
if args.searchthreads != -1:
print("Setting nb of threads to", args.searchthreads)
faiss.omp_set_num_threads(args.searchthreads)
else:
print("nb search threads: ", faiss.omp_get_max_threads())
ps = faiss.ParameterSpace()
ps.initialize(index)
parametersets = args.searchparams
if args.inter:
header = (
'%-40s inter@%3d time(ms/q) nb distances #runs' %
("parameters", args.k)
)
else:
header = (
'%-40s R@1 R@10 R@100 time(ms/q) nb distances #runs' %
"parameters"
)
res.search_results = {}
if parametersets == ['autotune']:
ps.n_experiments = args.n_autotune
ps.min_test_duration = args.min_test_duration
for kv in args.autotune_max:
k, vmax = kv.split(':')
vmax = float(vmax)
print("limiting %s to %g" % (k, vmax))
pr = ps.add_range(k)
values = faiss.vector_to_array(pr.values)
values = np.array([v for v in values if v < vmax])
faiss.copy_array_to_vector(values, pr.values)
for kv in args.autotune_range:
k, vals = kv.split(':')
vals = np.fromstring(vals, sep=',')
print("setting %s to %s" % (k, vals))
pr = ps.add_range(k)
faiss.copy_array_to_vector(vals, pr.values)
# setup the Criterion object
if args.inter:
print("Optimize for intersection @ ", args.k)
crit = faiss.IntersectionCriterion(nq, args.k)
else:
print("Optimize for 1-recall @ 1")
crit = faiss.OneRecallAtRCriterion(nq, 1)
# by default, the criterion will request only 1 NN
crit.nnn = args.k
crit.set_groundtruth(None, gt.astype('int64'))
# then we let Faiss find the optimal parameters by itself
print("exploring operating points, %d threads" % faiss.omp_get_max_threads());
ps.display()
t0 = time.time()
op = ps.explore(index, xq, crit)
res.t_explore = time.time() - t0
print("Done in %.3f s, available OPs:" % res.t_explore)
op.display()
print("Re-running evaluation on selected OPs")
print(header)
opv = op.optimal_pts
maxw = max(max(len(opv.at(i).key) for i in range(opv.size())), 40)
for i in range(opv.size()):
opt = opv.at(i)
ps.set_index_parameters(index, opt.key)
print(opt.key.ljust(maxw), end=' ')
sys.stdout.flush()
res_i = eval_setting(index, xq, gt, args.k, args.inter, args.min_test_duration)
res.search_results[opt.key] = res_i
else:
print(header)
for param in parametersets:
print("%-40s " % param, end=' ')
sys.stdout.flush()
ps.set_index_parameters(index, param)
res_i = eval_setting(index, xq, gt, args.k, args.inter, args.min_test_duration)
res.search_results[param] = res_i
######################################################
# Driver function
######################################################
def main():
parser = argparse.ArgumentParser()
def aa(*args, **kwargs):
group.add_argument(*args, **kwargs)
group = parser.add_argument_group('general options')
aa('--nthreads', default=-1, type=int,
help='nb of threads to use at train and add time')
aa('--json', default=False, action="store_true",
help="output stats in JSON format at the end")
aa('--todo', default=["check_files"],
choices=["train", "add", "search", "check_files"],
nargs="+", help='what to do (check_files means decide depending on which index files exist)')
group = parser.add_argument_group('dataset options')
aa('--db', default='deep1M', help='dataset')
aa('--compute_gt', default=False, action='store_true',
help='compute and store the groundtruth')
aa('--force_IP', default=False, action="store_true",
help='force IP search instead of L2')
aa('--accept_short_gt', default=False, action='store_true',
help='work around a problem with Deep1B GT')
group = parser.add_argument_group('index construction')
aa('--indexkey', default='HNSW32', help='index_factory type')
aa('--trained_indexfile', default='',
help='file to read or write a trained index from')
aa('--maxtrain', default=256 * 256, type=int,
help='maximum number of training points (0 to set automatically)')
aa('--indexfile', default='', help='file to read or write index from')
aa('--split', default=[1, 0], type=int, nargs=2, help="database split")
aa('--add_bs', default=-1, type=int,
help='add elements index by batches of this size')
group = parser.add_argument_group('IVF options')
aa('--by_residual', default=-1, type=int,
help="set if index should use residuals (default=unchanged)")
aa('--no_precomputed_tables', action='store_true', default=False,
help='disable precomputed tables (uses less memory)')
aa('--get_centroids_from', default='',
help='get the centroids from this index (to speed up training)')
aa('--clustering_niter', default=-1, type=int,
help='number of clustering iterations (-1 = leave default)')
aa('--train_on_gpu', default=False, action='store_true',
help='do training on GPU')
group = parser.add_argument_group('index-specific options')
aa('--M0', default=-1, type=int, help='size of base level for HNSW')
aa('--RQ_train_default', default=False, action="store_true",
help='disable progressive dim training for RQ')
aa('--RQ_beam_size', default=-1, type=int,
help='set beam size at add time')
aa('--LSQ_encode_ils_iters', default=-1, type=int,
help='ILS iterations for LSQ')
aa('--RQ_use_beam_LUT', default=-1, type=int,
help='use beam LUT at add time')
group = parser.add_argument_group('searching')
aa('--k', default=100, type=int, help='nb of nearest neighbors')
aa('--inter', default=False, action='store_true',
help='use intersection measure instead of 1-recall as metric')
aa('--searchthreads', default=-1, type=int,
help='nb of threads to use at search time')
aa('--searchparams', nargs='+', default=['autotune'],
help="search parameters to use (can be autotune or a list of params)")
aa('--n_autotune', default=500, type=int,
help="max nb of autotune experiments")
aa('--autotune_max', default=[], nargs='*',
help='set max value for autotune variables format "var:val" (exclusive)')
aa('--autotune_range', default=[], nargs='*',
help='set complete autotune range, format "var:val1,val2,..."')
aa('--min_test_duration', default=3.0, type=float,
help='run test at least for so long to avoid jitter')
aa('--indexes_to_merge', default=[], nargs="*",
help="load these indexes to search and merge them before searching")
args = parser.parse_args()
if args.todo == ["check_files"]:
if os.path.exists(args.indexfile):
args.todo = ["search"]
elif os.path.exists(args.trained_indexfile):
args.todo = ["add", "search"]
else:
args.todo = ["train", "add", "search"]
print("setting todo to", args.todo)
print("args:", args)
os.system('echo -n "nb processors "; '
'cat /proc/cpuinfo | grep ^processor | wc -l; '
'cat /proc/cpuinfo | grep ^"model name" | tail -1')
# object to collect results
res = argparse.Namespace()
res.args = args.__dict__
res.cpu_model = [
l for l in open("/proc/cpuinfo", "r")
if "model name" in l][0]
print("Load dataset")
ds = datasets.load_dataset(
dataset=args.db, compute_gt=args.compute_gt)
if args.force_IP:
ds.metric = "IP"
print(ds)
if args.nthreads != -1:
print("Set nb of threads to", args.nthreads)
faiss.omp_set_num_threads(args.nthreads)
else:
print("nb threads: ", faiss.omp_get_max_threads())
index = None
if "train" in args.todo:
print("================== Training index")
index = run_train(args, ds, res)
if args.trained_indexfile:
print("storing trained index", args.trained_indexfile)
faiss.write_index(index, args.trained_indexfile)
if "add" in args.todo:
if not index:
assert args.trained_indexfile
print("reading trained index", args.trained_indexfile)
index = faiss.read_index(args.trained_indexfile)
print("================== Adding vectors to index")
run_add(args, ds, index, res)
if args.indexfile:
print("storing", args.indexfile)
faiss.write_index(index, args.indexfile)
if "search" in args.todo:
if not index:
if args.indexfile:
print("reading index", args.indexfile)
index = faiss.read_index(args.indexfile)
elif args.indexes_to_merge:
print(f"Merging {len(args.indexes_to_merge)} indexes")
sz = 0
for fname in args.indexes_to_merge:
print(f" reading {fname} (current size {sz})")
index_i = faiss.read_index(fname)
if index is None:
index = index_i
else:
index.merge_from(index_i, index.ntotal)
sz = index.ntotal
else:
assert False, "provide --indexfile"
print("================== Searching")
run_search(args, ds, index, res)
if args.json:
print("JSON results:", json.dumps(res.__dict__))
if __name__ == "__main__":
main()