vectordb / Exports / IvfPQIndexConfig
Interface: IvfPQIndexConfig
Table of contents
Properties
- column
- index_cache_size
- index_name
- max_iters
- max_opq_iters
- metric_type
- num_bits
- num_partitions
- num_sub_vectors
- replace
- type
- use_opq
Properties
column
• Optional column: string
The column to be indexed
Defined in
index_cache_size
• Optional index_cache_size: number
Cache size of the index
Defined in
index_name
• Optional index_name: string
A unique name for the index
Defined in
max_iters
• Optional max_iters: number
The max number of iterations for kmeans training.
Defined in
max_opq_iters
• Optional max_opq_iters: number
Max number of iterations to train OPQ, if use_opq is true.
Defined in
metric_type
• Optional metric_type: MetricType
Metric type, l2 or Cosine
Defined in
num_bits
• Optional num_bits: number
The number of bits to present one PQ centroid.
Defined in
num_partitions
• Optional num_partitions: number
The number of partitions this index
Defined in
num_sub_vectors
• Optional num_sub_vectors: number
Number of subvectors to build PQ code
Defined in
replace
• Optional replace: boolean
Replace an existing index with the same name if it exists.
Defined in
type
• type: "ivf_pq"
Defined in
use_opq
• Optional use_opq: boolean
Train as optimized product quantization.