snowykami/meilisearch
1
0
Fork 0
mirror of https://github.com/meilisearch/meilisearch.git synced 2024-11-22 01:57:41 +08:00
Code Issues Actions 9 Packages Projects Releases Wiki Activity

Reintroduce vector errors

Browse Source
This commit is contained in:
Louis Dureuil 2024-10-30 10:55:57 +01:00
parent 0f6a1dbce7
commit df5bc3c9fd
No known key found for this signature in database
3 changed files with 117 additions and 32 deletions
Show Stats Download Patch File Download Diff File Expand all files Collapse all files

108
milli/src/update/new/extract/vectors/mod.rs
View File

@ -11,13 +11,16 @@ use crate::update::new::channel::EmbeddingSender;
use crate::update::new::indexer::document_changes::{Extractor, FullySend};
use crate::update::new::vector_document::VectorDocument;
use crate::update::new::DocumentChange;
use crate::vector::error::EmbedErrorKind;
use crate::vector::error::{
EmbedErrorKind, PossibleEmbeddingMistakes, UnusedVectorsDistributionBump,
};
use crate::vector::{Embedder, Embedding, EmbeddingConfigs};
use crate::{DocumentId, InternalError, Result, ThreadPoolNoAbort, UserError};
use crate::{DocumentId, FieldDistribution, InternalError, Result, ThreadPoolNoAbort, UserError};
pub struct EmbeddingExtractor<'a> {
embedders: &'a EmbeddingConfigs,
sender: &'a EmbeddingSender<'a>,
possible_embedding_mistakes: PossibleEmbeddingMistakes,
threads: &'a ThreadPoolNoAbort,
}
@ -25,9 +28,11 @@ impl<'a> EmbeddingExtractor<'a> {
pub fn new(
embedders: &'a EmbeddingConfigs,
sender: &'a EmbeddingSender<'a>,
field_distribution: &'a FieldDistribution,
threads: &'a ThreadPoolNoAbort,
) -> Self {
Self { embedders, sender, threads }
let possible_embedding_mistakes = PossibleEmbeddingMistakes::new(field_distribution);
Self { embedders, sender, threads, possible_embedding_mistakes }
}
}
@ -50,6 +55,8 @@ impl<'a, 'extractor> Extractor<'extractor> for EmbeddingExtractor<'a> {
>,
) -> crate::Result<()> {
let embedders = self.embedders.inner_as_ref();
let mut unused_vectors_distribution =
UnusedVectorsDistributionBump::new_in(&context.doc_alloc);
let mut all_chunks = BVec::with_capacity_in(embedders.len(), &context.doc_alloc);
for (embedder_name, (embedder, prompt, _is_quantized)) in embedders {
@ -66,6 +73,7 @@ impl<'a, 'extractor> Extractor<'extractor> for EmbeddingExtractor<'a> {
embedder_name,
prompt,
&context.data.0,
&self.possible_embedding_mistakes,
self.threads,
self.sender,
&context.doc_alloc,
@ -87,6 +95,10 @@ impl<'a, 'extractor> Extractor<'extractor> for EmbeddingExtractor<'a> {
)?;
let new_vectors = update.updated_vectors(&context.doc_alloc)?;
if let Some(new_vectors) = &new_vectors {
unused_vectors_distribution.append(new_vectors);
}
for chunks in &mut all_chunks {
let embedder_name = chunks.embedder_name();
let prompt = chunks.prompt();
@ -128,7 +140,11 @@ impl<'a, 'extractor> Extractor<'extractor> for EmbeddingExtractor<'a> {
&context.doc_alloc,
)?;
if new_rendered != old_rendered {
chunks.set_autogenerated(update.docid(), new_rendered)?;
chunks.set_autogenerated(
update.docid(),
new_rendered,
&unused_vectors_distribution,
)?;
}
}
} else if old_vectors.regenerate {
@ -151,17 +167,25 @@ impl<'a, 'extractor> Extractor<'extractor> for EmbeddingExtractor<'a> {
&context.doc_alloc,
)?;
if new_rendered != old_rendered {
chunks.set_autogenerated(update.docid(), new_rendered)?;
chunks.set_autogenerated(
update.docid(),
new_rendered,
&unused_vectors_distribution,
)?;
}
}
}
}
DocumentChange::Insertion(insertion) => {
let new_vectors = insertion.inserted_vectors(&context.doc_alloc)?;
if let Some(new_vectors) = &new_vectors {
unused_vectors_distribution.append(new_vectors);
}
for chunks in &mut all_chunks {
let embedder_name = chunks.embedder_name();
let prompt = chunks.prompt();
// if no inserted vectors, then regenerate: true + no embeddings => autogenerate
let new_vectors = insertion.inserted_vectors(&context.doc_alloc)?;
if let Some(new_vectors) = new_vectors.as_ref().and_then(|new_vectors| {
new_vectors.vectors_for_key(embedder_name).transpose()
}) {
@ -178,7 +202,11 @@ impl<'a, 'extractor> Extractor<'extractor> for EmbeddingExtractor<'a> {
context.new_fields_ids_map,
&context.doc_alloc,
)?;
chunks.set_autogenerated(insertion.docid(), rendered)?;
chunks.set_autogenerated(
insertion.docid(),
rendered,
&unused_vectors_distribution,
)?;
}
} else {
let rendered = prompt.render_document(
@ -186,7 +214,11 @@ impl<'a, 'extractor> Extractor<'extractor> for EmbeddingExtractor<'a> {
context.new_fields_ids_map,
&context.doc_alloc,
)?;
chunks.set_autogenerated(insertion.docid(), rendered)?;
chunks.set_autogenerated(
insertion.docid(),
rendered,
&unused_vectors_distribution,
)?;
}
}
}
@ -194,7 +226,7 @@ impl<'a, 'extractor> Extractor<'extractor> for EmbeddingExtractor<'a> {
}
for chunk in all_chunks {
chunk.drain()?;
chunk.drain(&unused_vectors_distribution)?;
}
Ok(())
}
@ -215,7 +247,7 @@ struct Chunks<'a> {
embedder_id: u8,
embedder_name: &'a str,
prompt: &'a Prompt,
possible_embedding_mistakes: &'a PossibleEmbeddingMistakes,
user_provided: &'a RefCell<HashMap<String, DelAddRoaringBitmap>>,
threads: &'a ThreadPoolNoAbort,
sender: &'a EmbeddingSender<'a>,
@ -229,6 +261,7 @@ impl<'a> Chunks<'a> {
embedder_name: &'a str,
prompt: &'a Prompt,
user_provided: &'a RefCell<HashMap<String, DelAddRoaringBitmap>>,
possible_embedding_mistakes: &'a PossibleEmbeddingMistakes,
threads: &'a ThreadPoolNoAbort,
sender: &'a EmbeddingSender<'a>,
doc_alloc: &'a Bump,
@ -241,6 +274,7 @@ impl<'a> Chunks<'a> {
ids,
embedder,
prompt,
possible_embedding_mistakes,
threads,
sender,
embedder_id,
@ -249,7 +283,12 @@ impl<'a> Chunks<'a> {
}
}
pub fn set_autogenerated(&mut self, docid: DocumentId, rendered: &'a str) -> Result<()> {
pub fn set_autogenerated(
&mut self,
docid: DocumentId,
rendered: &'a str,
unused_vectors_distribution: &UnusedVectorsDistributionBump,
) -> Result<()> {
if self.texts.len() < self.texts.capacity() {
self.texts.push(rendered);
self.ids.push(docid);
@ -262,18 +301,25 @@ impl<'a> Chunks<'a> {
self.embedder,
self.embedder_id,
self.embedder_name,
self.possible_embedding_mistakes,
unused_vectors_distribution,
self.threads,
self.sender,
)
}
pub fn drain(mut self) -> Result<()> {
pub fn drain(
mut self,
unused_vectors_distribution: &UnusedVectorsDistributionBump,
) -> Result<()> {
let res = Self::embed_chunks(
&mut self.texts,
&mut self.ids,
self.embedder,
self.embedder_id,
self.embedder_name,
self.possible_embedding_mistakes,
unused_vectors_distribution,
self.threads,
self.sender,
);
@ -285,11 +331,13 @@ impl<'a> Chunks<'a> {
pub fn embed_chunks(
texts: &mut BVec<'a, &'a str>,
ids: &mut BVec<'a, DocumentId>,
embedder: &'a Embedder,
embedder: &Embedder,
embedder_id: u8,
embedder_name: &str,
threads: &'a ThreadPoolNoAbort,
sender: &'a EmbeddingSender<'a>,
possible_embedding_mistakes: &PossibleEmbeddingMistakes,
unused_vectors_distribution: &UnusedVectorsDistributionBump,
threads: &ThreadPoolNoAbort,
sender: &EmbeddingSender<'a>,
) -> Result<()> {
let res = match embedder.embed_chunks_ref(texts.as_slice(), threads) {
Ok(embeddings) => {
@ -312,25 +360,23 @@ impl<'a> Chunks<'a> {
msg += &format!("\n- Note: `{embedder_name}` has `source: userProvided`, so documents must provide embeddings as an array in `_vectors.{embedder_name}`.");
}
/// FIXME: reintroduce possible_embedding_mistakes and possible_embedding_mistakes
let mut hint_count = 0;
/*
for (vector_misspelling, count) in
possible_embedding_mistakes.vector_mistakes().take(2)
{
msg += &format!("\n- Hint: try replacing `{vector_misspelling}` by `_vectors` in {count} document(s).");
hint_count += 1;
}
for (vector_misspelling, count) in
possible_embedding_mistakes.vector_mistakes().take(2)
{
msg += &format!("\n- Hint: try replacing `{vector_misspelling}` by `_vectors` in {count} document(s).");
hint_count += 1;
}
for (embedder_misspelling, count) in possible_embedding_mistakes
.embedder_mistakes_bump(embedder_name, unused_vectors_distribution)
.take(2)
{
msg += &format!("\n- Hint: try replacing `_vectors.{embedder_misspelling}` by `_vectors.{embedder_name}` in {count} document(s).");
hint_count += 1;
}
for (embedder_misspelling, count) in possible_embedding_mistakes
.embedder_mistakes(embedder_name, unused_vectors_distribution)
.take(2)
{
msg += &format!("\n- Hint: try replacing `_vectors.{embedder_misspelling}` by `_vectors.{embedder_name}` in {count} document(s).");
hint_count += 1;
}
*/
if hint_count == 0 {
if let EmbedErrorKind::ManualEmbed(_) = &error.kind {
msg += &format!(

2
milli/src/update/new/indexer/mod.rs
View File

@ -296,7 +296,7 @@ where
}
/// FIXME: need access to `merger_sender`
let embedding_sender = todo!();
let extractor = EmbeddingExtractor::new(embedders, &embedding_sender, request_threads());
let extractor = EmbeddingExtractor::new(embedders, &embedding_sender, &field_distribution, request_threads());
let datastore = ThreadLocal::with_capacity(pool.current_num_threads());
for_each_document_change(document_changes, &extractor, indexing_context, &mut extractor_allocs, &datastore)?;

39
milli/src/vector/error.rs
View File

@ -1,11 +1,13 @@
use std::collections::BTreeMap;
use std::path::PathBuf;
use bumpalo::Bump;
use hf_hub::api::sync::ApiError;
use super::parsed_vectors::ParsedVectorsDiff;
use super::rest::ConfigurationSource;
use crate::error::FaultSource;
use crate::update::new::vector_document::VectorDocument;
use crate::{FieldDistribution, PanicCatched};
#[derive(Debug, thiserror::Error)]
@ -417,6 +419,23 @@ impl PossibleEmbeddingMistakes {
}
})
}
pub fn embedder_mistakes_bump<'a, 'doc: 'a>(
&'a self,
embedder_name: &'a str,
unused_vectors_distribution: &'a UnusedVectorsDistributionBump<'doc>,
) -> impl Iterator<Item = (&'a str, u64)> + 'a {
let builder = levenshtein_automata::LevenshteinAutomatonBuilder::new(2, true);
let automata = builder.build_dfa(embedder_name);
unused_vectors_distribution.0.iter().filter_map(move |(field, count)| {
match automata.eval(field) {
levenshtein_automata::Distance::Exact(0) => None,
levenshtein_automata::Distance::Exact(_) => Some((*field, *count)),
levenshtein_automata::Distance::AtLeast(_) => None,
}
})
}
}
#[derive(Default)]
@ -433,3 +452,23 @@ impl UnusedVectorsDistribution {
}
}
}
pub struct UnusedVectorsDistributionBump<'doc>(
hashbrown::HashMap<&'doc str, u64, hashbrown::hash_map::DefaultHashBuilder, &'doc Bump>,
);
impl<'doc> UnusedVectorsDistributionBump<'doc> {
pub fn new_in(doc_alloc: &'doc Bump) -> Self {
Self(hashbrown::HashMap::new_in(doc_alloc))
}
pub fn append(&mut self, vectors: &impl VectorDocument<'doc>) -> Result<(), crate::Error> {
for res in vectors.iter_vectors() {
let (embedder_name, entry) = res?;
if !entry.has_configured_embedder {
*self.0.entry(embedder_name).or_default() += 1;
}
}
Ok(())
}
}