mirror of
https://github.com/qdrant/qdrant.git
synced 2026-09-29 09:27:53 -05:00
Ivan PleshkovandClaude Fable 5
4aa8066b75
* TurboQuantizer::score_precomputed_batch: score a contiguous run of vectors Batch counterpart of `score_precomputed` for vectors stored back to back at `quantized_size()`: the width's kernel scores the whole run of codes in one `dotprod_batch` call, then a second pass applies each vector's extras. L1 dequantizes per vector and stays a plain loop. Tested against per-vector `score_precomputed` for every width, distance, and mode over run lengths that leave every group remainder. Co-Authored-By: Claude Fable 5 <noreply@anthropic.com> * EncodedStorage::for_each_run: serve consecutive offsets as contiguous runs `for_each_run(offsets, callback(first, count, bytes))` splits the offsets into maximal runs of consecutive ids the storage can serve from one contiguous slice, so a sequential scan resolves chunk lookups and reads once per run instead of once per vector. The default serves every vector as its own run; `for_each_consecutive_run` is the shared run detection for storages that override it, with a per-run cap for chunk boundaries. The test storage overrides it (its data is one flat buffer). Co-Authored-By: Claude Fable 5 <noreply@anthropic.com> * EncodedVectors::score_points: batched scoring entry point, run-batched for TQ `score_points(query, offsets, scores)` scores a batch of points. The default keeps the per-vector loop the scorers run today, so SQ/PQ/BQ are unchanged. TurboQuant overrides it: on RAM/mmap storages it walks `for_each_run` and scores each contiguous run with one `score_precomputed_batch` call, hoisting the score inversion out of the loop; backends with async reads keep the pipelined per-vector path. Non-consecutive offsets degrade to single-vector runs, so scattered access keeps its previous cost. Integration test: `score_points` vs `score_point` for every bit width and mode, Dot and inverted L2, over sequential, scattered and descending id orders. Co-Authored-By: Claude Fable 5 <noreply@anthropic.com> * Quantized storages: for_each_run over their contiguous regions The RAM storage and both chunked mmap storages cap runs at their chunk boundary and serve each run with one `get_many`; the single-file mmap storage serves any run as one sequential read. Unit test on the RAM storage: runs cover every offset once, in order, with bytes identical to per-point reads, across the internal chunk boundary. Co-Authored-By: Claude Fable 5 <noreply@anthropic.com> * QuantizedQueryScorer: score batches through EncodedVectors::score_points Routes `score_stored_batch` through the batched entry point, so TurboQuant-as-quantization scans score contiguous runs with one kernel call per run; SQ/PQ/BQ keep the per-vector loop via the default. Co-Authored-By: Claude Fable 5 <noreply@anthropic.com> * TurboScoring::score_query_batch: run-batched scoring for Turbo4 storages Adds the batch counterpart of `score_query_bytes` to the trait, with one shared implementation over the storage's `EncodedStorage`: consecutive ids are coalesced into contiguous runs, each run scored by a single `score_precomputed_batch` call, and the metric sign applied once over the batch. Backends with async reads keep the pipelined per-vector path. `TurboQueryScorer::score_stored_batch` now calls it. The batch-vs-single storage test grows to 8192 vectors so a full ascending scan crosses a chunk boundary of the chunked backend, and runs that scan on the chunked, mmap and io_uring backends. Co-Authored-By: Claude Fable 5 <noreply@anthropic.com> * score_precomputed_batch: keep the extras pass in L1 The kernel pass and the extras pass now alternate over sub-runs of 64 vectors instead of each covering the whole run: for a run of several hundred vectors the second pass otherwise refetched every vector's extras from L2. Measured with 512-vector runs from the full-scan driver at dim 512: the regression against 64-vector runs went from +11 % to +2 %. Co-Authored-By: Claude Fable 5 <noreply@anthropic.com> * Bench: exhaustive search over Turbo4 storages through the plain-index driver `turbo4_full_scan` runs `BatchFilteredSearcher::peek_top_visible` — the exact path of a non-indexed search — over 200k normalized random vectors for Turbo4 as datatype (appendable chunked, in RAM) and Turbo4 as quantization (over a RAM dense storage), at dims 64 to 1024, so the fixed per-point cost of the scan driver is measured next to the kernel. `TURBO_SCAN_DIMS=64,128` narrows the dims while iterating. Co-Authored-By: Claude Fable 5 <noreply@anthropic.com> --------- Co-authored-by: Claude Fable 5 <noreply@anthropic.com>
236 lines
5.3 KiB
TOML
236 lines
5.3 KiB
TOML
[package]
|
|
name = "segment"
|
|
version = "0.6.0"
|
|
authors = [
|
|
"Andrey Vasnetsov <vasnetsov93@gmail.com>",
|
|
"Qdrant Team <info@qdrant.tech>",
|
|
]
|
|
license = "Apache-2.0"
|
|
edition = "2024"
|
|
|
|
[lints]
|
|
workspace = true
|
|
|
|
[features]
|
|
default = []
|
|
testing = ["common/testing", "sparse/testing", "gpu/testing", "quantization/testing"]
|
|
gpu = ["gpu/gpu"]
|
|
|
|
[build-dependencies]
|
|
cc = { workspace = true }
|
|
|
|
[dev-dependencies]
|
|
clap = { workspace = true }
|
|
criterion = { workspace = true }
|
|
dataset = { path = "../common/dataset" }
|
|
env_logger = { workspace = true }
|
|
fs-err = { workspace = true, features = ["debug"] }
|
|
humantime = { workspace = true }
|
|
indicatif = { workspace = true }
|
|
ndarray = "0.17.2"
|
|
ndarray-npy = { version = "0.10.0", default-features = false }
|
|
rmp-serde = { workspace = true }
|
|
rand_distr = { workspace = true }
|
|
walkdir = { workspace = true }
|
|
rstest = { workspace = true }
|
|
segment = { path = ".", default-features = false, features = ["testing"] }
|
|
proptest = { workspace = true }
|
|
anyhow = { workspace = true }
|
|
object_store = { workspace = true }
|
|
bytes = { workspace = true }
|
|
tokio = { workspace = true }
|
|
tap = { workspace = true }
|
|
|
|
[target.'cfg(not(target_os = "windows"))'.dev-dependencies]
|
|
pprof = { workspace = true }
|
|
|
|
[dependencies]
|
|
io_bridge_object_store = { path = "../common/io_bridge_object_store" }
|
|
blink-alloc = { workspace = true }
|
|
futures = { workspace = true }
|
|
bytemuck = { workspace = true }
|
|
data-encoding = { workspace = true }
|
|
fs-err = { workspace = true }
|
|
integer-encoding = { workspace = true }
|
|
tempfile = { workspace = true }
|
|
parking_lot = { workspace = true }
|
|
rayon = { workspace = true }
|
|
itertools = { workspace = true }
|
|
uuid = { workspace = true }
|
|
bincode = { workspace = true }
|
|
serde = { workspace = true }
|
|
serde_json = { workspace = true }
|
|
serde_cbor = { workspace = true }
|
|
serde-value = "0.7"
|
|
serde_variant = { workspace = true }
|
|
serde-untagged = "0.1.9"
|
|
ordered-float = { workspace = true }
|
|
thiserror = { workspace = true }
|
|
atomic_refcell = { workspace = true }
|
|
atomicwrites = { workspace = true }
|
|
memmap2 = { workspace = true }
|
|
schemars = { workspace = true }
|
|
log = { workspace = true }
|
|
geo = { version = "0.33.1", default-features = false }
|
|
geohash = "0.13.2"
|
|
num-traits = { workspace = true }
|
|
num-derive = "0.5.1"
|
|
num-cmp = "0.1.0"
|
|
once_cell = "1.21"
|
|
rand = { workspace = true }
|
|
bitvec = { workspace = true }
|
|
fs_extra = { workspace = true }
|
|
tinyvec = { workspace = true }
|
|
validator = { workspace = true }
|
|
chrono = { workspace = true }
|
|
ecow = { workspace = true }
|
|
fnv = { workspace = true }
|
|
indexmap = { workspace = true }
|
|
ahash = { workspace = true }
|
|
self_cell.workspace = true
|
|
sha2 = { workspace = true }
|
|
siphasher = "1.0.3"
|
|
smallvec = { workspace = true }
|
|
strum = { workspace = true }
|
|
tokio = { workspace = true }
|
|
byteorder = { workspace = true }
|
|
stumpalo = { workspace = true }
|
|
zerocopy = { workspace = true }
|
|
vaporetto = { version = "0.6.5" }
|
|
rust-stemmers = { package = "qdrant-rust-stemmers", version = "1.2.2" }
|
|
# Memory only; disable `user` so macOS does not pull objc2 via open-directory
|
|
# (objc2::Retained's IntoIterator blanket impl overflows trait resolution).
|
|
sysinfo = { workspace = true }
|
|
charabia = { version = "0.10.0", default-features = false, features = [
|
|
"greek",
|
|
"hebrew",
|
|
"thai",
|
|
"chinese-segmentation",
|
|
"chinese-normalization",
|
|
] }
|
|
|
|
blobstore = { path = "../blobstore" }
|
|
|
|
common = { path = "../common/common" }
|
|
macros = { path = "../macros" }
|
|
posting_list = { path = "../posting_list" }
|
|
quantization = { path = "../quantization" }
|
|
sparse = { path = "../sparse" }
|
|
gpu = { path = "../gpu" }
|
|
|
|
tracing = { workspace = true, optional = true }
|
|
macro_rules_attribute = "0.2.2"
|
|
nom = "8.0.0"
|
|
half = { workspace = true }
|
|
roaring = { workspace = true }
|
|
duplicate = { workspace = true }
|
|
|
|
[target.'cfg(target_os = "linux")'.dependencies]
|
|
procfs = { version = "0.18", default-features = false }
|
|
|
|
[[bench]]
|
|
name = "vector_search"
|
|
harness = false
|
|
|
|
[[bench]]
|
|
name = "turbo_vector_search"
|
|
harness = false
|
|
|
|
[[bench]]
|
|
name = "turbo_full_scan"
|
|
harness = false
|
|
|
|
[[bench]]
|
|
name = "read_vectors"
|
|
harness = false
|
|
|
|
[[bench]]
|
|
name = "read_payloads"
|
|
harness = false
|
|
|
|
[[bench]]
|
|
name = "hnsw_build_graph"
|
|
harness = false
|
|
|
|
[[bench]]
|
|
name = "hnsw_search_graph"
|
|
harness = false
|
|
|
|
[[bench]]
|
|
name = "conditional_search"
|
|
harness = false
|
|
|
|
[[bench]]
|
|
name = "dynamic_mmap_flags"
|
|
harness = false
|
|
|
|
[[bench]]
|
|
name = "hnsw_build_asymptotic"
|
|
harness = false
|
|
|
|
[[bench]]
|
|
name = "serde_formats"
|
|
harness = false
|
|
|
|
[[bench]]
|
|
name = "id_type_benchmark"
|
|
harness = false
|
|
|
|
[[bench]]
|
|
name = "map_benchmark"
|
|
harness = false
|
|
|
|
[[bench]]
|
|
name = "boolean_filtering"
|
|
harness = false
|
|
|
|
[[bench]]
|
|
name = "numeric_index_check_values"
|
|
harness = false
|
|
|
|
[[bench]]
|
|
name = "sparse_index_search"
|
|
harness = false
|
|
|
|
[[bench]]
|
|
name = "sparse_index_build"
|
|
harness = false
|
|
bench = false # TODO(rocksdb): fix and re-enable (or just remove)
|
|
|
|
[[bench]]
|
|
name = "sparse_vector_storage"
|
|
harness = false
|
|
bench = false # TODO(rocksdb): fix and re-enable (or just remove)
|
|
|
|
[[bench]]
|
|
name = "multi_vector_search"
|
|
harness = false
|
|
|
|
[[bench]]
|
|
name = "metrics"
|
|
harness = false
|
|
|
|
[[bench]]
|
|
name = "range_filtering"
|
|
harness = false
|
|
|
|
[[bench]]
|
|
name = "in_memory_id_tracker"
|
|
harness = false
|
|
|
|
[[bench]]
|
|
name = "buffered_update_bitslice"
|
|
harness = false
|
|
|
|
[[bench]]
|
|
name = "segment_info"
|
|
harness = false
|
|
|
|
[[bench]]
|
|
name = "hnsw_incremental_build"
|
|
harness = false
|
|
|
|
[[bench]]
|
|
name = "facets"
|
|
harness = false
|