新版本来了! Milvus v1.1 发布!

来源: 投稿
作者: Zilliz
2021-05-17

在 Milvus 1.0 版本发布后的 2 个月,2021 年 5 月 7 日,Milvus 正式发布了 1.1 版本!

Milvus 1.1 版本新增诸多优化改进,修复大量漏洞,进一步丰富和完善了 Milvus 第一个长期支持(LTS)版本[1]。 以下是 Milvus 1.1 发版说明,想了解更多详情,请见 Milvus GitHub: https://github.com/milvus-io/milvus/releases/tag/v1.1.0 。如需做数据迁移的朋友,可使用 Milvus 数据迁移工具 -- Milvusdm

特别感谢以下社区贡献者,为此版本添砖加瓦:

 @BossZou @shengjun1985 @op-hunter  @matrixji @yhmo  @ericsyh @LocoRichard @del-zhenwu @XuanYang-cn  @fishpenguin

发版说明

New Features

  • #4564 Supports specifying partition in a get_entity_by_id() method call.

  • #4806 Supports specifying partition in a delete_entity_by_id() method call.

  • #4905 Adds the release_collection() method, which unloads a specific collection from cache.

Improvements

  • #4756 Improves the performance of the get_entity_by_id() method call.

  • #4856 Upgrades hnswlib to v0.5.0.

  • #4958 Improves the performance of IVF index training.

Fixed issues

  • #4778 Fails to access vector index in Mishards.

  • #4797 The system returns false results after merging search requests with different topK parameters.

  • #4838 The server does not respond immediately to an index building request on an empty collection.

  • #4858 For GPU-enabled Milvus, the system crashes on a search request with a large topK (> 2048).

  • #4862 A read-only node merges segments during startup.

  • #4894 The capacity of a Bloom filter does not equal to the row count of the segment it belongs to.

  • #4908 The GPU cache is not cleaned up after a collection is dropped.

  • #4933 It takes a long while for the system to build index for a small segment.

  • #4952 Fails to set timezone as "UTC + 5:30".

  • #5008 The system crashes randomly during continuous, concurrent delete, insert, and search operations.

  • #5010 For GPU-enabled Milvus, query fails on IVF_PQ if nbits ≠ 8.

  • #5050 get_collection_stats() returns false index type for segments still in the process of index building.

  • #5063 The system crashes when an empty segment is flushed.

  • #5078 For GPU-enabled Milvus, the system crashes when creating an IVF index on vectors of 2048, 4096, or 8192 dimensions.

备注

[1] Milvus 长期版本定义:https://milvus.io/cn/docs/v1.1.0/announcement.md

About Zilliz

Zilliz 以重新定义数据科学为愿景,致力于打造一家全球领先的开源技术创新公司,并通过开源和云原生解决方案为企业解锁非结构化数据的隐藏价值。 

Zilliz 构建了 Milvus 向量相似度搜索引擎,以加快下一代数据平台的发展。Milvus 目前是 LF AI & Data 基金会的孵化阶段项目,能够管理大量非结构化数据集。我们的技术在新药发现、计算机视觉、推荐引擎、聊天机器人等方面具有广泛的应用。 

展开阅读全文
7 收藏
分享
加载中
更多评论
0 评论
7 收藏
分享
返回顶部
顶部