cpu_cache_capacity: 16 # how many memory are used as cache, unit: GB, range: 0 ~ less than total memory
cpu_cache_free_percent: 0.85 # old data will be erased from cache when cache is full, this value specify how much memory should be kept, range: greater than zero ~ 1.0
insert_cache_immediately: false # insert data will be load into cache immediately for hot query
gpu_cache_capacity: 5 # how many memory are used as cache in gpu, unit: GB, RANGE: 0 ~ less than total memory
gpu_cache_free_percent: 0.85 # old data will be erased from cache when cache is full, this value specify how much memory should be kept, range: greater than zero ~ 1.0
prometheus_config: # following are prometheus configure
port: 8080 # the port prometheus use to fetch metrics
push_gateway_ip_address: 127.0.0.1 # push method configure: push gateway ip address
push_gateway_port: 9091 # push method configure: push gateway port
cache_config:
cpu_cache_capacity: 16 # how many memory are used as cache, unit: GB, range: 0 ~ less than total memory
cpu_cache_free_percent: 0.85 # old data will be erased from cache when cache is full, this value specify how much memory should be kept, range: greater than zero ~ 1.0
insert_cache_immediately: false # insert data will be load into cache immediately for hot query
gpu_cache_capacity: 5 # how many memory are used as cache in gpu, unit: GB, RANGE: 0 ~ less than total memory
gpu_cache_free_percent: 0.85 # old data will be erased from cache when cache is full, this value specify how much memory should be kept, range: greater than zero ~ 1.0
gpu_ids: 0,1 # gpu id
engine_config:
nprobe: 10
nlist: 16384
use_blas_threshold: 20
metric_type: L2 # compare vectors by euclidean distance(L2) or inner product(IP), optional: L2 or IP
omp_thread_num: 0 # how many compute threads be used by engine, 0 means use all cpu core to compute
metric_type: L2 # compare vectors by euclidean distance(L2) or inner product(IP), optional: L2 or IP
omp_thread_num: 0 # how many cpu cores be used by engine, 0 means use all cpu cores used