Milvus

Compatibility changes

- The `load_collection()` method supports preloading specified partitions. [4307](https://github.com/milvus-io/milvus/issues/4307)

Improvements

- Optimizes the process of indexing and querying. [4454](https://github.com/milvus-io/milvus/issues/4454)

Fixed issues

- Calling `load_collection()` and `search()` in multiple threads causes Milvus to hang. [4378](https://github.com/milvus-io/milvus/issues/4378)
- Milvus only searches the default partition if the search parameter `partition_tags` contains `_default`. [4484](https://github.com/milvus-io/milvus/issues/4484)

> See [CHANGELOG](https://github.com/milvus-io/milvus/blob/0.10.5/CHANGELOG.md) for more information.

Improvements

- Improves the query performance in situations where no entities are deleted. [3775](https://github.com/milvus-io/milvus/issues/3775)
- Reduces the time to create a large number of partitions. [4318](https://github.com/milvus-io/milvus/issues/4318)

Fixed issues

- The server crashes when searching with IVF_PQ on GPU. [3626](https://github.com/milvus-io/milvus/issues/3626)
- The performance of query on IVF_SQ8H worsens. [3903](https://github.com/milvus-io/milvus/issues/3903)
- The server crashes when deleting entities. [3906](https://github.com/milvus-io/milvus/issues/3906)
- The server occasionally hangs when creating and dropping partitions continuously. [4012](https://github.com/milvus-io/milvus/issues/4012)
- An out-of-memory error occurs when searching on GPU. [4174](https://github.com/milvus-io/milvus/issues/4174)
- Memory leaks when building an IVF index. [4318](https://github.com/milvus-io/milvus/issues/4318)

> See [CHANGELOG](https://github.com/milvus-io/milvus/blob/0.10.4/CHANGELOG.md) for more information.

Improvements


- Allows the user to specify `metric_type` when searching in a collection with FLAT index. [3213](https://github.com/milvus-io/milvus/issues/3213)
- Adds more valid values to the indexing parameter `m` of IVF_PQ: An `m` is valid so long as it can be divided by `dim`, the number of vector dimensions. [3254](https://github.com/milvus-io/milvus/issues/3254)
- Expands the range of `nprobe` to [1, 16384] for all IVF indexes. [3606](https://github.com/milvus-io/milvus/issues/3606)
- Expands the range of `top_k` in an embeddings search to [1, 16384]. [3639](https://github.com/milvus-io/milvus/issues/3639)
- Optimizes the memory release strategy to avoid memory overuse. [3536](https://github.com/milvus-io/milvus/issues/3536)

Fixed issues


- Milvus running on multiple GPU devices and using IVF_SQ8H crashes if GPU cache is set too low. [3742](https://github.com/milvus-io/milvus/issues/3742)
- Milvus reports a CUDA error when quitting, if the index is IVF_SQ8H. [3760](https://github.com/milvus-io/milvus/issues/3760)

> See [CHANGELOG](https://github.com/milvus-io/milvus/blob/0.10.3/CHANGELOG.md) for more information.

Improvements


- Improved IVF search performance when both `nq` and `nprobe` are large. [2653](https://github.com/milvus-io/milvus/issues/2653)


Fixed issues

- Inaccurate calculation of the memory allocated for index files. [2890](https://github.com/milvus-io/milvus/issues/2890)
- Search result sets based on IVF_PQ index with IP metric are incorrectly merged in situations where multiple segments are produced for the inserted data. [2952](https://github.com/milvus-io/milvus/issues/2952)
- The server crashes if you set `cache.cache_size` smaller than the size of an index file in a multi-GPU scenario. [3012](https://github.com/milvus-io/milvus/issues/3012)
- Search result sets based on IP metric are incorrectly merged, if you insert vectors to multiple partitions in Mishards. [3133](https://github.com/milvus-io/milvus/issues/3133)

> See [CHANGELOG](https://github.com/milvus-io/milvus/blob/0.10.2/CHANGELOG.md) for more information.

Fixed issues

- Occasionally, the result count does not match the query count. [2578](https://github.com/milvus-io/milvus/issues/2578)
- IVF_PQ does not support IP on GPU-enabled Milvus. [2585](https://github.com/milvus-io/milvus/issues/2585)
- Milvus reports "illegal instruction" when running on some legacy CPUs. [2598](https://github.com/milvus-io/milvus/issues/2598)
- Updated the range of HNSW settings. [2637](https://github.com/milvus-io/milvus/issues/2637)
- Milvus processes crash when building an index. [2642](https://github.com/milvus-io/milvus/issues/2642)
- The default ANNOY settings do not match the settings specified in the document. [2649](https://github.com/milvus-io/milvus/issues/2649)
- Milvus does not respond under a stress test. [2692](https://github.com/milvus-io/milvus/issues/2692)
- The precision of the returned vectors is incorrect in an HTTP interface. [2752](https://github.com/milvus-io/milvus/issues/2752)
- The upper limit of `nprobe` is incorrect on the GPU-enabled Milvus. [2767](https://github.com/milvus-io/milvus/issues/2767)
- The number of the vectors changes occasionally after building an index. [2768](https://github.com/milvus-io/milvus/issues/2768)

> See [CHANGELOG](https://github.com/milvus-io/milvus/blob/0.10.1/CHANGELOG.md) for more information.

Compatibility changes

- Updates the Milvus configuration file. [2510](https://github.com/milvus-io/milvus/issues/2510)

Improvements

- Optimizes the index building performance when many small segments exist. [2373](https://github.com/milvus-io/milvus/issues/2373)
- FAISS is upgraded to 1.6.3. [2381](https://github.com/milvus-io/milvus/issues/2381)
- Optimizes dropping collection performance when too many partitions exist. [2394](https://github.com/milvus-io/milvus/issues/2394)
- Optimizes the implementation of k-selection algorithm for GPU-enabled Milvus. [2466](https://github.com/milvus-io/milvus/issues/2466)

Fixed issues

- The search performance degrades on Milvus 0.9.1. [2429](https://github.com/milvus-io/milvus/issues/2429)

> See [CHANGELOG](https://github.com/milvus-io/milvus/blob/0.10.0/CHANGELOG.md) for more information.