snowykami/meilisearch
1
0
Fork 0
mirror of https://github.com/meilisearch/meilisearch.git synced 2025-09-07 05:06:31 +00:00
Code Issues Packages Projects Releases Wiki Activity

Remove lots of Arcs

Browse Source
This commit is contained in:
Mubelotix
2025-06-26 12:07:48 +02:00
parent ef007d547d
commit 29f6eeff8f
22 changed files with 112 additions and 118 deletions
Download Patch File Download Diff File Expand all files Collapse all files

23
crates/milli/src/vector/rest.rs
View File

@ -1,5 +1,4 @@
use std::collections::BTreeMap;
use std::sync::Arc;
use std::time::Instant;
use deserr::Deserr;
@ -170,7 +169,7 @@ impl Embedder {
&self,
texts: Vec<String>,
deadline: Option<Instant>,
embedder_stats: Option<Arc<EmbedderStats>>,
embedder_stats: Option<&EmbedderStats>,
) -> Result<Vec<Embedding>, EmbedError> {
embed(
&self.data,
@ -186,7 +185,7 @@ impl Embedder {
&self,
texts: &[S],
deadline: Option<Instant>,
embedder_stats: Option<Arc<EmbedderStats>>,
embedder_stats: Option<&EmbedderStats>,
) -> Result<Vec<Embedding>, EmbedError>
where
S: AsRef<str> + Serialize,
@ -208,21 +207,21 @@ impl Embedder {
&self,
text_chunks: Vec<Vec<String>>,
threads: &ThreadPoolNoAbort,
embedder_stats: Arc<EmbedderStats>,
embedder_stats: &EmbedderStats,
) -> Result<Vec<Vec<Embedding>>, EmbedError> {
// This condition helps reduce the number of active rayon jobs
// so that we avoid consuming all the LMDB rtxns and avoid stack overflows.
if threads.active_operations() >= REQUEST_PARALLELISM {
text_chunks
.into_iter()
.map(move |chunk| self.embed(chunk, None, Some(embedder_stats.clone())))
.map(move |chunk| self.embed(chunk, None, Some(embedder_stats)))
.collect()
} else {
threads
.install(move || {
text_chunks
.into_par_iter()
.map(move |chunk| self.embed(chunk, None, Some(embedder_stats.clone())))
.map(move |chunk| self.embed(chunk, None, Some(embedder_stats)))
.collect()
})
.map_err(|error| EmbedError {
@ -236,14 +235,14 @@ impl Embedder {
&self,
texts: &[&str],
threads: &ThreadPoolNoAbort,
embedder_stats: Arc<EmbedderStats>,
embedder_stats: &EmbedderStats,
) -> Result<Vec<Embedding>, EmbedError> {
// This condition helps reduce the number of active rayon jobs
// so that we avoid consuming all the LMDB rtxns and avoid stack overflows.
if threads.active_operations() >= REQUEST_PARALLELISM {
let embeddings: Result<Vec<Vec<Embedding>>, _> = texts
.chunks(self.prompt_count_in_chunk_hint())
.map(move |chunk| self.embed_ref(chunk, None, Some(embedder_stats.clone())))
.map(move |chunk| self.embed_ref(chunk, None, Some(embedder_stats)))
.collect();
let embeddings = embeddings?;
@ -253,7 +252,7 @@ impl Embedder {
.install(move || {
let embeddings: Result<Vec<Vec<Embedding>>, _> = texts
.par_chunks(self.prompt_count_in_chunk_hint())
.map(move |chunk| self.embed_ref(chunk, None, Some(embedder_stats.clone())))
.map(move |chunk| self.embed_ref(chunk, None, Some(embedder_stats)))
.collect();
let embeddings = embeddings?;
@ -303,7 +302,7 @@ fn embed<S>(
expected_count: usize,
expected_dimension: Option<usize>,
deadline: Option<Instant>,
embedder_stats: Option<Arc<EmbedderStats>>,
embedder_stats: Option<&EmbedderStats>,
) -> Result<Vec<Embedding>, EmbedError>
where
S: Serialize,
@ -323,7 +322,7 @@ where
for attempt in 0..10 {
if let Some(embedder_stats) = &embedder_stats {
embedder_stats.as_ref().total_count.fetch_add(1, std::sync::atomic::Ordering::Relaxed);
embedder_stats.total_count.fetch_add(1, std::sync::atomic::Ordering::Relaxed);
}
let response = request.clone().send_json(&body);
let result = check_response(response, data.configuration_source).and_then(|response| {
@ -367,7 +366,7 @@ where
}
if let Some(embedder_stats) = &embedder_stats {
embedder_stats.as_ref().total_count.fetch_add(1, std::sync::atomic::Ordering::Relaxed);
embedder_stats.total_count.fetch_add(1, std::sync::atomic::Ordering::Relaxed);
}
let response = request.send_json(&body);
let result = check_response(response, data.configuration_source).and_then(|response| {