-
Notifications
You must be signed in to change notification settings - Fork 53
New issue
Have a question about this project? Sign up for a free GitHub account to open an issue and contact its maintainers and the community.
By clicking “Sign up for GitHub”, you agree to our terms of service and privacy statement. We’ll occasionally send you account related emails.
Already on GitHub? Sign in to your account
Narek/storage upgrade #262
Conversation
440e570
to
bd11f64
Compare
Codecov ReportAttention:
Additional details and impacted files@@ Coverage Diff @@
## main #262 +/- ##
==========================================
- Coverage 77.26% 75.68% -1.58%
==========================================
Files 23 23
Lines 1869 1929 +60
Branches 480 505 +25
==========================================
+ Hits 1444 1460 +16
- Misses 220 247 +27
- Partials 205 222 +17
|
a9ad205
to
24f944b
Compare
4fb29f1
to
abeac7a
Compare
fd4316a
to
dcce645
Compare
uint32 neighbors_nr; | ||
uint32 neighbors_max; | ||
uint32 *neighbors; | ||
uint32 unused; | ||
|
||
LDB_VI_READ_NODE_CHUNK(vi_node, vi_node->vn_label, node_tape, &tape_pos, node_tape_size); | ||
LDB_VI_READ_NODE_CHUNK(vi_node, vi_node->vn_dim, node_tape, &tape_pos, node_tape_size); |
There was a problem hiding this comment.
Choose a reason for hiding this comment
The reason will be displayed to describe this comment to others. Learn more.
deleted because new binary index does not contain this information, so there is no room for inconsistency
…ment in UBSAN mode
When I ran the old command on a machine that had a really old lantern (v0.0.3), I noticed that it was never updated to the latest version and the old version ran instead. This makes sure that does not happen with our default docker command
Update usearch Remove hnswlib dependency Get closer to upstream usearch API Switch to using usearch_search_ef - new upstream compatible API in C api of usearch Update stroage level for external index generation
…ndomness Newer storage made the index slightly smaller which put old m=4 sift 1k index be near the border of a filled index page. As a result, depending on level distribution in a run, the regression test would have a differently sized index and result in a failure. Here we increase m slighly, to make it very unlikely a test will fail as a result of level randomness
96be1b7
to
3cc22f4
Compare
3cc22f4
to
282d056
Compare
We always create a new usearch instance per search, and always use it from a single thread, so no need to allocate resources for more threads
282d056
to
7534dc4
Compare
fb55a12
to
5164a93
Compare
5164a93
to
dc216a9
Compare
No description provided.