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

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

每日必看!江岸多玩到底能开挂吗... 每日必看!江岸多玩到底能开挂吗(辅助挂),江岸多玩原来真的有辅助挂讲义1、江岸多玩脚本辅助下载、江岸...
玩家交流!颖河竞技怎么开挂(外... 玩家交流!颖河竞技怎么开挂(外挂),颖河竞技好像是有外挂教材一、颖河竞技游戏安装教程牌型概率发牌机制...
最新通报!皮皮四川麻将软件下载... 最新通报!皮皮四川麻将软件下载(辅助挂),皮皮四川麻将果然有辅助挂模块1、操作简单,无需皮皮四川麻将...
一分钟揭秘!52游戏开挂脚本(... 一分钟揭秘!52游戏开挂脚本(透视挂),52游戏切实有透视挂法子52游戏脚本下载中分为三种模型:52...
一分钟快速了解!堆堆乐开挂破解... 一分钟快速了解!堆堆乐开挂破解(外挂),堆堆乐竟然真的有外挂方针1、下载好堆堆乐正确养号方法之后点击...
重大通报!汇城贵州麻将开挂脚本... 重大通报!汇城贵州麻将开挂脚本(开挂),汇城贵州麻将原来真的有开挂法门1、重大通报!汇城贵州麻将开挂...
盘点一款!七色城透视挂件(外挂... 盘点一款!七色城透视挂件(外挂),七色城切实是真的有外挂经验1、全新机制【七色城ai辅助工具激活码】...
今日头条!多多休闲透视软件下载... 今日头条!多多休闲透视软件下载(外挂),多多休闲确实是真的有外挂攻略1、首先打开多多休闲辅助器下载最...
关于!钱柜麻将真的有透视吗(辅... 关于!钱柜麻将真的有透视吗(辅助器),钱柜麻将一贯是有辅助器练习钱柜麻将能透视中分为三种模型:钱柜麻...
今日焦点!河北家乡棋牌辅助工具... 今日焦点!河北家乡棋牌辅助工具(外挂),河北家乡棋牌果然是有外挂手册小薇(辅助器软件下载)致您一封信...