milvus分批写入测试数据
创始人
2024-12-29 01:38:08
0

milvus分批写入测试数据

场景:

假如现在需要向milvus写入500W数据,调用milvus的insert api一次性写入500W数据是不现实的,内存也吃不消。

这时候需要分批写入,例如每次写入1000行,直至写完500W数据。

实现代码如下:

package main  import ( 	"context" 	"fmt" 	"github.com/milvus-io/milvus-sdk-go/v2/client" 	"github.com/milvus-io/milvus-sdk-go/v2/entity" 	"log" 	"math/rand" 	"strconv" )  const ( 	milvusAddr     = `192.168.230.71:19530` 	dim            = 512 	collectionName = "hello_iterator"  	msgFmt                                     = "==== %s ====\n" 	idCol, randomCol, addressCol, embeddingCol = "ID", "random", "address", "embeddings" 	totalRows                                  = 100005 	batchSize                                  = 1000 )  func main() { 	ctx := context.Background()  	log.Printf(msgFmt, "start connecting to Milvus") 	c, err := client.NewClient(ctx, client.Config{ 		Address: milvusAddr, 	}) 	if err != nil { 		log.Fatal("failed to connect to milvus, err: ", err.Error()) 	} 	defer c.Close()  	// delete collection if exists 	has, err := c.HasCollection(ctx, collectionName) 	if err != nil { 		log.Fatalf("failed to check collection exists, err: %v", err) 	} 	if has { 		//c.DropCollection(ctx, collectionName) 		log.Println("collection exists") 		return 	}  	// create collection 	log.Printf(msgFmt, fmt.Sprintf("create collection, `%s`", collectionName)) 	schema := entity.NewSchema().WithName(collectionName).WithDescription("hello_milvus is the simplest demo to introduce the APIs"). 		WithField(entity.NewField().WithName(idCol).WithDataType(entity.FieldTypeInt64).WithIsPrimaryKey(true).WithIsAutoID(false)). 		WithField(entity.NewField().WithName(randomCol).WithDataType(entity.FieldTypeDouble)). 		WithField(entity.NewField().WithName(addressCol).WithDataType(entity.FieldTypeVarChar).WithTypeParams(entity.TypeParamMaxLength, "50")). 		WithField(entity.NewField().WithName(embeddingCol).WithDataType(entity.FieldTypeFloatVector).WithDim(dim))  	if err := c.CreateCollection(ctx, schema, 1); err != nil { 		log.Fatalf("create collection failed, err: %v", err) 	} 	// 分批数 	groupNum := totalRows / batchSize 	if totalRows%batchSize > 0 { 		groupNum++ // 如果有余数,则需要多写一批 	} 	// 分批写入数据 	for i := 0; i < groupNum; i++ { 		// 计算当前批次的起始和结束索引 		start := i * batchSize 		end := start + batchSize 		if end > totalRows { 			end = totalRows // 确保不会超出总数 		} 		// insert data 		idList := make([]int64, 0, batchSize) 		randomList := make([]float64, 0, batchSize) 		addressList := make([]string, 0, batchSize) 		embeddingList := make([][]float32, 0, batchSize)  		// generate data 		for i := start; i < end; i++ { 			idList = append(idList, int64(i)) 		} 		for i := start; i < end; i++ { 			randomList = append(randomList, rand.Float64()) 		} 		for i := start; i < end; i++ { 			addressList = append(addressList, "wuhan"+strconv.Itoa(i)) 		} 		for i := start; i < end; i++ { 			vec := make([]float32, 0, dim) 			for j := 0; j < dim; j++ { 				vec = append(vec, rand.Float32()) 			} 			embeddingList = append(embeddingList, vec) 		} 		idColData := entity.NewColumnInt64(idCol, idList) 		randomColData := entity.NewColumnDouble(randomCol, randomList) 		addressColData := entity.NewColumnVarChar(addressCol, addressList) 		embeddingColData := entity.NewColumnFloatVector(embeddingCol, dim, embeddingList)  		if _, err := c.Insert(ctx, collectionName, "", idColData, randomColData, addressColData, embeddingColData); err != nil { 			log.Fatalf("failed to insert random data into `hello_milvus, err: %v", err) 		} 		log.Printf("inserted:%d\n", end) 	}  	if err := c.Flush(ctx, collectionName, false); err != nil { 		log.Fatalf("failed to flush data, err: %v", err) 	}  	// build index 	log.Printf(msgFmt, "start creating index HNSW") 	idx, err := entity.NewIndexHNSW(entity.COSINE, 15, 50) 	if err != nil { 		log.Fatalf("failed to create ivf flat index, err: %v", err) 	} 	if err := c.CreateIndex(ctx, collectionName, embeddingCol, idx, false); err != nil { 		log.Fatalf("failed to create index, err: %v", err) 	}  	log.Printf(msgFmt, "start loading collection") 	err = c.LoadCollection(ctx, collectionName, false) 	if err != nil { 		log.Fatalf("failed to load collection, err: %v", err) 	} } 

相关内容

热门资讯

第2分钟脚本!新玄龙辅助工具,... 第2分钟脚本!新玄龙辅助工具,福建天天开心辅助工具下载(竟然是真的辅助脚本)-哔哩哔哩1、下载好福建...
1分钟秘籍!hhpoker真的... 1分钟秘籍!hhpoker真的有透视吗(透视)竟然是真的有辅助教程(哔哩哔哩)1、hhpoker真的...
重大通报"蜜瓜大厅可... 重大通报"蜜瓜大厅可以装挂吗"一贯存在有辅助技巧(有挂分析)-哔哩哔哩蜜瓜大厅可以装挂吗是不是有人用...
透视办法!poker worl... 透视办法!poker world辅助器(透视)开挂透视工具(哔哩哔哩);1、poker world辅...
第五分钟脚本!微信超级三加一辅... 第五分钟脚本!微信超级三加一辅助,小程序牵手跑得有外开挂吗(切实真的有辅助app)-哔哩哔哩1、这是...
三分钟项目!wepoker透视... 三分钟项目!wepoker透视是真的吗(透视)总是存在有辅助开挂(哔哩哔哩)1、操作简单,无需wep...
盘点几款"情怀游戏字... 盘点几款"情怀游戏字牌辅助"都是是有辅助攻略(有挂方略)-哔哩哔哩1、情怀游戏字牌辅助辅助器安装包、...
透视课程!wepoker俱乐部... 透视课程!wepoker俱乐部辅助(透视)开挂透视辅助器(哔哩哔哩)暗藏猫腻,小编详细说明wepok...
两分钟辅助!爱来大菠萝怎么玩,... 两分钟辅助!爱来大菠萝怎么玩,皮皮透视辅助软件工具(一贯是有辅助平台)-哔哩哔哩皮皮透视辅助软件工具...
2分钟经验!pokerrrr2... 2分钟经验!pokerrrr2辅助(透视)一贯真的是有辅助辅助(哔哩哔哩)1、点击下载安装,poke...