邯郸网站建设产品介绍,小程序api文档,站酷官网,vi设计论文目录
一、overview
基础代码
核心API#xff1a;
二、核心概念
2.1 加载数据
从S3上读
从本地读#xff1a; 其他读取方式 读取分布式数据#xff08;spark#xff09; 从ML libraries 库中读取#xff08;不支持并行读取#xff09;
从sql中读取
2.2 变换数据…目录
一、overview
基础代码
核心API
二、核心概念
2.1 加载数据
从S3上读
从本地读 其他读取方式 读取分布式数据spark 从ML libraries 库中读取不支持并行读取
从sql中读取
2.2 变换数据
map
flat_map
Transforming batches
Shuffling rows
Repartitioning data
2.3 消费数据
1 按行遍历
2按batch遍历
3遍历batch时shuffle
4为分布式并行训练分割数据
2.4 保存数据
保存文件
修改分区数
将数据转换为python对象
将数据转换为分布式数据spark 今天来带大家一起来学习下ray中对数据的操作还是非常简洁的。
一、overview 基础代码
from typing import Dict
import numpy as np
import ray# Create datasets from on-disk files, Python objects, and cloud storage like S3.
ds ray.data.read_csv(s3://anonymousray-example-data/iris.csv)# Apply functions to transform data. Ray Data executes transformations in parallel.
def compute_area(batch: Dict[str, np.ndarray]) - Dict[str, np.ndarray]:length batch[petal length (cm)]width batch[petal width (cm)]batch[petal area (cm^2)] length * widthreturn batchtransformed_ds ds.map_batches(compute_area)# Iterate over batches of data.
for batch in transformed_ds.iter_batches(batch_size4):print(batch)# Save dataset contents to on-disk files or cloud storage.
transformed_ds.write_parquet(local:///tmp/iris/)
使用ray.data可以方便地从硬盘、python对象、S3上读取文件
最后写入云端
核心API 简单变换(map_batches()) 全局聚合和分组聚合(groupby()) Shuffling 操作 (random_shuffle(), sort(), repartition()). 二、核心概念
2.1 加载数据 从S3上读
import ray#加载csv文件
ds ray.data.read_csv(s3://anonymousair-example-data/iris.csv)
print(ds.schema())
ds.show(limit1)#加载parquet文件
ds ray.data.read_parquet(s3://anonymousray-example-data/iris.parquet)#加载image
ds ray.data.read_images(s3://anonymousray-example-data/batoidea/JPEGImages/)# Text
ds ray.data.read_text(s3://anonymousray-example-data/this.txt)# binary
ds ray.data.read_binary_files(s3://anonymousray-example-data/documents)#tfrecords
ds ray.data.read_tfrecords(s3://anonymousray-example-data/iris.tfrecords) 从本地读
ds ray.data.read_parquet(local:///tmp/iris.parquet)
处理压缩文件
ds ray.data.read_csv(s3://anonymousray-example-data/iris.csv.gz,arrow_open_stream_args{compression: gzip},
) 其他读取方式
import ray# 从python对象里获取
ds ray.data.from_items([{food: spam, price: 9.34},{food: ham, price: 5.37},{food: eggs, price: 0.94}
])ds ray.data.from_items([1, 2, 3, 4, 5])# 从numpy里获取
array np.ones((3, 2, 2))
ds ray.data.from_numpy(array)# 从pandas里获取
df pd.DataFrame({food: [spam, ham, eggs],price: [9.34, 5.37, 0.94]
})
ds ray.data.from_pandas(df)# 从py arrow里获取table pa.table({food: [spam, ham, eggs],price: [9.34, 5.37, 0.94]
})
ds ray.data.from_arrow(table) 读取分布式数据spark
import ray
import raydpspark raydp.init_spark(app_nameSpark - Datasets Example,num_executors2,executor_cores2,executor_memory500MB)
df spark.createDataFrame([(i, str(i)) for i in range(10000)], [col1, col2])
ds ray.data.from_spark(df)ds.show(3) 从ML libraries 库中读取不支持并行读取
import ray.data
from datasets import load_dataset# 从huggingface里读取不支持并行读取
hf_ds load_dataset(wikitext, wikitext-2-raw-v1)
ray_ds ray.data.from_huggingface(hf_ds[train])
ray_ds.take(2)# 从TensorFlow中读取不支持并行读取
import ray
import tensorflow_datasets as tfdstf_ds, _ tfds.load(cifar10, split[train, test])
ds ray.data.from_tf(tf_ds)print(ds)
从sql中读取
import mysql.connectorimport raydef create_connection():return mysql.connector.connect(useradmin,password...,hostexample-mysql-database.c2c2k1yfll7o.us-west-2.rds.amazonaws.com,connection_timeout30,databaseexample,)# Get all movies
dataset ray.data.read_sql(SELECT * FROM movie, create_connection)
# Get movies after the year 1980
dataset ray.data.read_sql(SELECT title, score FROM movie WHERE year 1980, create_connection
)
# Get the number of movies per year
dataset ray.data.read_sql(SELECT year, COUNT(*) FROM movie GROUP BY year, create_connection
)
Ray还支持从BigQuery和MongoDB中读取篇幅问题不赘述了。
2.2 变换数据
变换默认是lazy直到遍历、保存、检视数据集时才执行
map
import os
from typing import Any, Dict
import raydef parse_filename(row: Dict[str, Any]) - Dict[str, Any]:row[filename] os.path.basename(row[path])return rowds (ray.data.read_images(s3://anonymousray-example-data/image-datasets/simple, include_pathsTrue).map(parse_filename)
)
flat_map
from typing import Any, Dict, List
import raydef duplicate_row(row: Dict[str, Any]) - List[Dict[str, Any]]:return [row] * 2print(ray.data.range(3).flat_map(duplicate_row).take_all()
)# 结果
# [{id: 0}, {id: 0}, {id: 1}, {id: 1}, {id: 2}, {id: 2}]
# 原先的元素都变成2个
Transforming batches
from typing import Dict
import numpy as np
import raydef increase_brightness(batch: Dict[str, np.ndarray]) - Dict[str, np.ndarray]:batch[image] np.clip(batch[image] 4, 0, 255)return batch# batch_format指定batch类型可不加
ds (ray.data.read_images(s3://anonymousray-example-data/image-datasets/simple).map_batches(increase_brightness, batch_formatnumpy)
)
如果初始化较贵使用类而不是函数这样每次调用类的时候进行初始化。类有状态而函数没有状态。
并行度可以指定(min,max)来自由调整
Shuffling rows
import rayds (ray.data.read_images(s3://anonymousray-example-data/image-datasets/simple).random_shuffle()
)
Repartitioning data
import rayds ray.data.range(10000, parallelism1000)# Repartition the data into 100 blocks. Since shuffleFalse, Ray Data will minimize
# data movement during this operation by merging adjacent blocks.
ds ds.repartition(100, shuffleFalse).materialize()# Repartition the data into 200 blocks, and force a full data shuffle.
# This operation will be more expensive
ds ds.repartition(200, shuffleTrue).materialize()
2.3 消费数据
1 按行遍历
import rayds ray.data.read_csv(s3://anonymousair-example-data/iris.csv)for row in ds.iter_rows():print(row)
2按batch遍历
numpy、pandas、torch、tf使用不同的API遍历batch
# numpy
import ray
ds ray.data.read_images(s3://anonymousray-example-data/image-datasets/simple)
for batch in ds.iter_batches(batch_size2, batch_formatnumpy):print(batch)# pandas
import ray
ds ray.data.read_csv(s3://anonymousair-example-data/iris.csv)
for batch in ds.iter_batches(batch_size2, batch_formatpandas):print(batch)# torch
import ray
ds ray.data.read_images(s3://anonymousray-example-data/image-datasets/simple)
for batch in ds.iter_torch_batches(batch_size2):print(batch)# tf
import rayds ray.data.read_csv(s3://anonymousair-example-data/iris.csv)tf_dataset ds.to_tf(feature_columnssepal length (cm),label_columnstarget,batch_size2
)
for features, labels in tf_dataset:print(features, labels)3遍历batch时shuffle
只需要在遍历batch时增加local_shuffle_buffer_size参数即可。
非全局洗牌但性能更好。
import rayds ray.data.read_images(s3://anonymousray-example-data/image-datasets/simple)for batch in ds.iter_batches(batch_size2,batch_formatnumpy,local_shuffle_buffer_size250,
):print(batch)
4为分布式并行训练分割数据
import rayray.remote
class Worker:def train(self, data_iterator):for batch in data_iterator.iter_batches(batch_size8):passds ray.data.read_csv(s3://anonymousair-example-data/iris.csv)
workers [Worker.remote() for _ in range(4)]
shards ds.streaming_split(n4, equalTrue)
ray.get([w.train.remote(s) for w, s in zip(workers, shards)])
2.4 保存数据
保存文件
非常类似pandas保存文件唯一的区别保存本地文件时需要加入local://前缀。
注意如果不加local://前缀ray则会将不同分区的数据写在不同节点上
import rayds ray.data.read_csv(s3://anonymousray-example-data/iris.csv)# local
ds.write_parquet(local:///tmp/iris/)# s3
ds.write_parquet(s3://my-bucket/my-folder)
修改分区数
import os
import rayds ray.data.read_csv(s3://anonymousray-example-data/iris.csv)
ds.repartition(2).write_csv(/tmp/two_files/)print(os.listdir(/tmp/two_files/))
将数据转换为python对象
import rayds ray.data.read_csv(s3://anonymousray-example-data/iris.csv)df ds.to_pandas()
print(df)
将数据转换为分布式数据spark
import ray
import raydpspark raydp.init_spark(app_name example,num_executors 1,executor_cores 4,executor_memory 512M
)ds ray.data.read_csv(s3://anonymousray-example-data/iris.csv)
df ds.to_spark(spark)
