支持python的分佈式計算框架Ray詳解

Posted on by

項目地址：https://github.com/ray-project/ray

1、簡介

Ray為構建分佈式應用程序提供瞭一個簡單、通用的API。Ray是一種分佈式執行框架,便於大規模應用程序和利用先進的機器學習庫。

Ray通過以下方式完成這項任務：

為構建和運行分佈式應用程序提供簡單的原語。

使最終用戶能夠並行化單個機器代碼，而代碼更改很少到零。

在核心Ray之上包含大量應用程序、庫和工具，以支持復雜的應用程序。

2、安裝

安裝方式比較簡單： pip install ray==1.4.1

[root@node2 ~]# pip install 'ray[default]'
Looking in indexes: https://mirrors.aliyun.com/pypi/simple/
Collecting ray[default]
  Downloading https://mirrors.aliyun.com/pypi/packages/13/ec/f727ddd3fbcdc6102eace62c9d5dd9d9ad8112d40eeb7de8783676aca24d/ray-1.4.1-cp36-cp36m-manylinux2014_x86_64.whl (51.6 MB)
     |████████████████████████████████| 51.6 MB 269 kB/s 
Collecting aiohttp
  Downloading https://mirrors.aliyun.com/pypi/packages/31/28/7b49246d7825c61e1a14306ab050e8eec48fcf5b009b19c8f6ea4e312467/aiohttp-3.7.4.post0-cp36-cp36m-manylinux2014_x86_64.whl (1.3 MB)
     |████████████████████████████████| 1.3 MB 121.5 MB/s 
Requirement already satisfied: dataclasses in /usr/local/lib/python3.6/site-packages (from ray[default]) (0.8)
Collecting protobuf>=3.15.3
  Downloading https://mirrors.aliyun.com/pypi/packages/53/4e/e2db88d0bb0bda6a879eea62fddbaf813719ce3770d458bc5580512d9c95/protobuf-3.17.3-cp36-cp36m-manylinux_2_5_x86_64.manylinux1_x86_64.whl (1.0 MB)
     |████████████████████████████████| 1.0 MB 136.1 MB/s 
Collecting jsonschema
  Downloading https://mirrors.aliyun.com/pypi/packages/c5/8f/51e89ce52a085483359217bc72cdbf6e75ee595d5b1d4b5ade40c7e018b8/jsonschema-3.2.0-py2.py3-none-any.whl (56 kB)
     |████████████████████████████████| 56 kB 7.1 MB/s 
Collecting aioredis
  Downloading https://mirrors.aliyun.com/pypi/packages/b0/64/1b1612d0a104f21f80eb4c6e1b6075f2e6aba8e228f46f229cfd3fdac859/aioredis-1.3.1-py3-none-any.whl (65 kB)
     |████████████████████████████████| 65 kB 6.3 MB/s 
Collecting pydantic>=1.8
  Downloading https://mirrors.aliyun.com/pypi/packages/2b/7c/7d0b3f2d7959b7193018896db236ded165f9bca1bb75f46f4c32fa6f4f9d/pydantic-1.8.2-cp36-cp36m-manylinux2014_x86_64.whl (10.2 MB)
     |████████████████████████████████| 10.2 MB 420 kB/s 
Collecting pyyaml
  Downloading https://mirrors.aliyun.com/pypi/packages/7a/5b/bc0b5ab38247bba158504a410112b6c03f153c652734ece1849749e5f518/PyYAML-5.4.1-cp36-cp36m-manylinux1_x86_64.whl (640 kB)
     |████████████████████████████████| 640 kB 19.1 MB/s 
Collecting numpy>=1.16
  Downloading https://mirrors.aliyun.com/pypi/packages/14/32/d3fa649ad7ec0b82737b92fefd3c4dd376b0bb23730715124569f38f3a08/numpy-1.19.5-cp36-cp36m-manylinux2010_x86_64.whl (14.8 MB)
     |████████████████████████████████| 14.8 MB 296 kB/s 
Collecting gpustat
  Downloading https://mirrors.aliyun.com/pypi/packages/b4/69/d8c849715171aeabd61af7da080fdc60948b5a396d2422f1f4672e43d008/gpustat-0.6.0.tar.gz (78 kB)
     |████████████████████████████████| 78 kB 7.3 MB/s 
Collecting redis>=3.5.0
  Downloading https://mirrors.aliyun.com/pypi/packages/a7/7c/24fb0511df653cf1a5d938d8f5d19802a88cef255706fdda242ff97e91b7/redis-3.5.3-py2.py3-none-any.whl (72 kB)
     |████████████████████████████████| 72 kB 968 kB/s 
Collecting aiohttp-cors
  Downloading https://mirrors.aliyun.com/pypi/packages/13/e7/e436a0c0eb5127d8b491a9b83ecd2391c6ff7dcd5548dfaec2080a2340fd/aiohttp_cors-0.7.0-py3-none-any.whl (27 kB)
Collecting py-spy>=0.2.0
  Downloading https://mirrors.aliyun.com/pypi/packages/9d/4d/1a9cbe9a0b543e6733cb38afe26451522a9ef8e4897b59e74cc76838f245/py_spy-0.3.7-py2.py3-none-manylinux1_x86_64.whl (3.1 MB)
     |████████████████████████████████| 3.1 MB 131 kB/s 
Requirement already satisfied: requests in /usr/local/lib/python3.6/site-packages (from ray[default]) (2.24.0)
Collecting colorama
  Downloading https://mirrors.aliyun.com/pypi/packages/44/98/5b86278fbbf250d239ae0ecb724f8572af1c91f4a11edf4d36a206189440/colorama-0.4.4-py2.py3-none-any.whl (16 kB)
Collecting filelock
  Downloading https://mirrors.aliyun.com/pypi/packages/93/83/71a2ee6158bb9f39a90c0dea1637f81d5eef866e188e1971a1b1ab01a35a/filelock-3.0.12-py3-none-any.whl (7.6 kB)
Collecting prometheus-client>=0.7.1
  Downloading https://mirrors.aliyun.com/pypi/packages/09/da/4e8471ff825769581593b5b84769d32f58e5373b59fccaf355d3529ad530/prometheus_client-0.11.0-py2.py3-none-any.whl (56 kB)
     |████████████████████████████████| 56 kB 4.2 MB/s 
Collecting msgpack<2.0.0,>=1.0.0
  Downloading https://mirrors.aliyun.com/pypi/packages/0c/0d/b1d9d32d03ce38ba5e2a37fbae850afd4530a14cc441e8335f1865a03705/msgpack-1.0.2-cp36-cp36m-manylinux1_x86_64.whl (272 kB)
     |████████████████████████████████| 272 kB 127.6 MB/s 
Requirement already satisfied: click>=7.0 in /usr/local/lib/python3.6/site-packages (from ray[default]) (7.1.2)
Collecting opencensus
  Downloading https://mirrors.aliyun.com/pypi/packages/18/59/12044123133d000f705383ad98579aeb0dd82d66b33a254a21b54bf0d6bb/opencensus-0.7.13-py2.py3-none-any.whl (127 kB)
     |████████████████████████████████| 127 kB 128.9 MB/s 
Collecting grpcio>=1.28.1
  Downloading https://mirrors.aliyun.com/pypi/packages/eb/32/332a2bb0cee0d4040331952c5ef3d040d119a15b3c622b60bdcbe768bded/grpcio-1.39.0-cp36-cp36m-manylinux2014_x86_64.whl (4.3 MB)
     |████████████████████████████████| 4.3 MB 417 kB/s 
Collecting colorful
  Downloading https://mirrors.aliyun.com/pypi/packages/b0/8e/e386e248266952d24d73ed734c2f5513f34d9557032618c8910e605dfaf6/colorful-0.5.4-py2.py3-none-any.whl (201 kB)
     |████████████████████████████████| 201 kB 113.9 MB/s 
Requirement already satisfied: six>=1.5.2 in /usr/local/lib/python3.6/site-packages (from grpcio>=1.28.1->ray[default]) (1.15.0)
Requirement already satisfied: typing-extensions>=3.7.4.3 in /usr/local/lib/python3.6/site-packages (from pydantic>=1.8->ray[default]) (3.10.0.0)
Collecting multidict<7.0,>=4.5
  Downloading https://mirrors.aliyun.com/pypi/packages/a1/35/b22524d6b9cacfb4c5eff413a069bbc17c6ea628e54da5c6c989998ced5f/multidict-5.1.0-cp36-cp36m-manylinux2014_x86_64.whl (141 kB)
     |████████████████████████████████| 141 kB 120.6 MB/s 
Collecting idna-ssl>=1.0
  Downloading https://mirrors.aliyun.com/pypi/packages/46/03/07c4894aae38b0de52b52586b24bf189bb83e4ddabfe2e2c8f2419eec6f4/idna-ssl-1.1.0.tar.gz (3.4 kB)
Collecting attrs>=17.3.0
  Downloading https://mirrors.aliyun.com/pypi/packages/20/a9/ba6f1cd1a1517ff022b35acd6a7e4246371dfab08b8e42b829b6d07913cc/attrs-21.2.0-py2.py3-none-any.whl (53 kB)
     |████████████████████████████████| 53 kB 3.9 MB/s 
Requirement already satisfied: chardet<5.0,>=2.0 in /usr/local/lib/python3.6/site-packages (from aiohttp->ray[default]) (3.0.4)
Collecting yarl<2.0,>=1.0
  Downloading https://mirrors.aliyun.com/pypi/packages/da/08/52b26b44bce7b818b410aee37c5e424c9ea420c557bca97dc2adac29b151/yarl-1.6.3-cp36-cp36m-manylinux2014_x86_64.whl (293 kB)
     |████████████████████████████████| 293 kB 130.8 MB/s 
Collecting async-timeout<4.0,>=3.0
  Downloading https://mirrors.aliyun.com/pypi/packages/e1/1e/5a4441be21b0726c4464f3f23c8b19628372f606755a9d2e46c187e65ec4/async_timeout-3.0.1-py3-none-any.whl (8.2 kB)
Requirement already satisfied: idna>=2.0 in /usr/local/lib/python3.6/site-packages (from idna-ssl>=1.0->aiohttp->ray[default]) (2.10)
Collecting hiredis
  Downloading https://mirrors.aliyun.com/pypi/packages/82/72/313fb6d30b7e413c8662447f9a0abe9105494ceb9266ecb08d442c14afbb/hiredis-2.0.0-cp36-cp36m-manylinux2010_x86_64.whl (84 kB)
     |████████████████████████████████| 84 kB 4.8 MB/s 
Collecting nvidia-ml-py3>=7.352.0
  Downloading https://mirrors.aliyun.com/pypi/packages/6d/64/cce82bddb80c0b0f5c703bbdafa94bfb69a1c5ad7a79cff00b482468f0d3/nvidia-ml-py3-7.352.0.tar.gz (19 kB)
Collecting psutil
  Downloading https://mirrors.aliyun.com/pypi/packages/da/82/56cd16a4c5f53e3e5dd7b2c30d5c803e124f218ebb644ca9c30bc907eadd/psutil-5.8.0-cp36-cp36m-manylinux2010_x86_64.whl (291 kB)
     |████████████████████████████████| 291 kB 131.6 MB/s 
Collecting blessings>=1.6
  Downloading https://mirrors.aliyun.com/pypi/packages/03/74/489f85a78247609c6b4f13733cbf3ba0d864b11aa565617b645d6fdf2a4a/blessings-1.7-py3-none-any.whl (18 kB)
Requirement already satisfied: setuptools in /usr/lib/python3.6/site-packages (from jsonschema->ray[default]) (39.2.0)
Collecting pyrsistent>=0.14.0
  Downloading https://mirrors.aliyun.com/pypi/packages/6c/19/1af501f6f388a40ede6d0185ba481bdb18ffc99deab0dd0d092b173bc0f4/pyrsistent-0.18.0-cp36-cp36m-manylinux1_x86_64.whl (117 kB)
     |████████████████████████████████| 117 kB 138.2 MB/s 
Requirement already satisfied: importlib-metadata in /usr/local/lib/python3.6/site-packages (from jsonschema->ray[default]) (4.3.0)
Requirement already satisfied: zipp>=0.5 in /usr/local/lib/python3.6/site-packages (from importlib-metadata->jsonschema->ray[default]) (3.4.1)
Collecting opencensus-context==0.1.2
  Downloading https://mirrors.aliyun.com/pypi/packages/f1/33/990f1bd9e7ee770fc8d3c154fc24743a96f16a0e49e14e1b7540cc2fdd93/opencensus_context-0.1.2-py2.py3-none-any.whl (4.4 kB)
Collecting google-api-core<2.0.0,>=1.0.0
  Downloading https://mirrors.aliyun.com/pypi/packages/aa/51/629a31d4e7db2bbc0b0b789ac4034108c68647bea9cf5b54c8366df47b8e/google_api_core-1.31.0-py2.py3-none-any.whl (93 kB)
     |████████████████████████████████| 93 kB 1.6 MB/s 
Collecting contextvars
  Downloading https://mirrors.aliyun.com/pypi/packages/83/96/55b82d9f13763be9d672622e1b8106c85acb83edd7cc2fa5bc67cd9877e9/contextvars-2.4.tar.gz (9.6 kB)
Collecting google-auth<2.0dev,>=1.25.0
  Downloading https://mirrors.aliyun.com/pypi/packages/d7/13/35b1e0a63e160ecad04985e60339382e48e6020d22f8328a2ab47226c910/google_auth-1.33.1-py2.py3-none-any.whl (152 kB)
     |████████████████████████████████| 152 kB 127.7 MB/s 
Collecting packaging>=14.3
  Downloading https://mirrors.aliyun.com/pypi/packages/3c/77/e2362b676dc5008d81be423070dd9577fa03be5da2ba1105811900fda546/packaging-21.0-py3-none-any.whl (40 kB)
     |████████████████████████████████| 40 kB 6.1 MB/s 
Collecting googleapis-common-protos<2.0dev,>=1.6.0
  Downloading https://mirrors.aliyun.com/pypi/packages/55/08/796a6bc0b550e2b7116041c953d3d5100016abea106131d71e5651826e7b/googleapis_common_protos-1.53.0-py2.py3-none-any.whl (198 kB)
     |████████████████████████████████| 198 kB 136.3 MB/s 
Collecting setuptools
  Downloading https://mirrors.aliyun.com/pypi/packages/bd/25/5bdf7f1adeebd4e3fa76b2e2f045ae53ee208e40a4231ad0f0c3007e4353/setuptools-57.4.0-py3-none-any.whl (819 kB)
     |████████████████████████████████| 819 kB 127.0 MB/s 
Requirement already satisfied: pytz in /usr/local/lib/python3.6/site-packages (from google-api-core<2.0.0,>=1.0.0->opencensus->ray[default]) (2020.1)
Collecting cachetools<5.0,>=2.0.0
  Downloading https://mirrors.aliyun.com/pypi/packages/bf/28/c4f5796c67ad06bb91d98d543a5e01805c1ff065e08871f78e52d2a331ad/cachetools-4.2.2-py3-none-any.whl (11 kB)
Collecting pyasn1-modules>=0.2.1
  Downloading https://mirrors.aliyun.com/pypi/packages/95/de/214830a981892a3e286c3794f41ae67a4495df1108c3da8a9f62159b9a9d/pyasn1_modules-0.2.8-py2.py3-none-any.whl (155 kB)
     |████████████████████████████████| 155 kB 21.5 MB/s 
Collecting rsa<5,>=3.1.4
  Downloading https://mirrors.aliyun.com/pypi/packages/e9/93/0c0f002031f18b53af7a6166103c02b9c0667be528944137cc954ec921b3/rsa-4.7.2-py3-none-any.whl (34 kB)
Collecting pyparsing>=2.0.2
  Downloading https://mirrors.aliyun.com/pypi/packages/8a/bb/488841f56197b13700afd5658fc279a2025a39e22449b7cf29864669b15d/pyparsing-2.4.7-py2.py3-none-any.whl (67 kB)
     |████████████████████████████████| 67 kB 7.2 MB/s 
Collecting pyasn1<0.5.0,>=0.4.6
  Downloading https://mirrors.aliyun.com/pypi/packages/62/1e/a94a8d635fa3ce4cfc7f506003548d0a2447ae76fd5ca53932970fe3053f/pyasn1-0.4.8-py2.py3-none-any.whl (77 kB)
     |████████████████████████████████| 77 kB 7.0 MB/s 
Requirement already satisfied: certifi>=2017.4.17 in /usr/local/lib/python3.6/site-packages (from requests->ray[default]) (2020.6.20)
Requirement already satisfied: urllib3!=1.25.0,!=1.25.1,<1.26,>=1.21.1 in /usr/local/lib/python3.6/site-packages (from requests->ray[default]) (1.25.10)
Collecting immutables>=0.9
  Downloading https://mirrors.aliyun.com/pypi/packages/4a/52/e64a14a99c509cbdfe0405e9f076aef0331cb9548a3efa1d5bacd524978a/immutables-0.15-cp36-cp36m-manylinux1_x86_64.whl (100 kB)
     |████████████████████████████████| 100 kB 18.2 MB/s 
Using legacy 'setup.py install' for idna-ssl, since package 'wheel' is not installed.
Using legacy 'setup.py install' for gpustat, since package 'wheel' is not installed.
Using legacy 'setup.py install' for nvidia-ml-py3, since package 'wheel' is not installed.
Using legacy 'setup.py install' for contextvars, since package 'wheel' is not installed.
Installing collected packages: pyasn1, setuptools, rsa, pyparsing, pyasn1-modules, protobuf, multidict, immutables, cachetools, yarl, packaging, idna-ssl, googleapis-common-protos, google-auth, contextvars, attrs, async-timeout, pyrsistent, psutil, opencensus-context, nvidia-ml-py3, hiredis, google-api-core, blessings, aiohttp, redis, pyyaml, pydantic, py-spy, prometheus-client, opencensus, numpy, msgpack, jsonschema, grpcio, gpustat, filelock, colorama, aioredis, aiohttp-cors, ray, colorful
  Attempting uninstall: setuptools
    Found existing installation: setuptools 39.2.0
    Uninstalling setuptools-39.2.0:
      Successfully uninstalled setuptools-39.2.0
    Running setup.py install for idna-ssl ... done
    Running setup.py install for contextvars ... done
    Running setup.py install for nvidia-ml-py3 ... done
    Running setup.py install for gpustat ... done
Successfully installed aiohttp-3.7.4.post0 aiohttp-cors-0.7.0 aioredis-1.3.1 async-timeout-3.0.1 attrs-21.2.0 blessings-1.7 cachetools-4.2.2 colorama-0.4.4 colorful-0.5.4 contextvars-2.4 filelock-3.0.12 google-api-core-1.31.0 google-auth-1.33.1 googleapis-common-protos-1.53.0 gpustat-0.6.0 grpcio-1.39.0 hiredis-2.0.0 idna-ssl-1.1.0 immutables-0.15 jsonschema-3.2.0 msgpack-1.0.2 multidict-5.1.0 numpy-1.19.5 nvidia-ml-py3-7.352.0 opencensus-0.7.13 opencensus-context-0.1.2 packaging-21.0 prometheus-client-0.11.0 protobuf-3.17.3 psutil-5.8.0 py-spy-0.3.7 pyasn1-0.4.8 pyasn1-modules-0.2.8 pydantic-1.8.2 pyparsing-2.4.7 pyrsistent-0.18.0 pyyaml-5.4.1 ray-1.4.1 redis-3.5.3 rsa-4.7.2 setuptools-57.4.0 yarl-1.6.3

3、單機

下面是不使用分佈式的代碼示例：

# test_ray.py
import ray
import cv2
import numpy
 
import time
 
#@ray.remote
def test(img):
    orb = cv2.AKAZE_create()
    kb = orb.detect(img, None)
    kp, des = orb.compute(img, kp)
    return des
 
# 不使用註解的方式
test_remote = ray.remote(test)
 
# 函數test 加註解時使用方式
# test_remote = test.remote(test)
 
img = cv2.imread("/tmp/test.jpg")
 
start_time = time.time()
 
for i in range(10):
    test(img)
 
print(time.time() - start_time)

運行：python test_ray.py

4、集群

如果需要啟動集群模式，則需要先啟動服務，選擇一臺機器作為主服務器，然後按蝦米那命令啟動：

[root@node1 ~]# ray start --head --port=8888      
Local node IP: 192.168.0.81
2021-07-28 09:58:44,051 INFO services.py:1274 -- View the Ray dashboard at http://127.0.0.1:8265
 
--------------------
Ray runtime started.
--------------------
 
Next steps
  To connect to this Ray runtime from another node, run
    ray start --address='192.168.0.81:8888' --redis-password='5241590000000000'
  
  Alternatively, use the following Python code:
    import ray
    ray.init(address='auto', _redis_password='5241590000000000')
  
  If connection fails, check your firewall settings and network configuration.
  
  To terminate the Ray runtime, run
    ray stop

可以看到裡面提示具體的用法。此時通過ray的web界面，端口8265訪問，看到已有一臺機器。

然後在另外兩臺機啟動客戶端節點，保證各個節點的ray版本一致,查看版本：

# pip freeze |grep ray

版本不一致的話，要升級為一致：pip install -U ray， 此處選擇的版本為1.4.1

從節點啟動方式(密碼可以在主節點啟動時顯示)：

ray start --address='192.168.0.81:8888' --redis-password='5241590000000000'

改寫程序，用於分佈式：

# test_ray.py
import ray
import cv2
import numpy
 
import time
 
@ray.remote
def test(img):
    orb = cv2.AKAZE_create()
    kb = orb.detect(img, None)
    kp, des = orb.compute(img, kp)
    return des
 
# 不使用註解的方式
#test_remote = ray.remote(test)
 
# 函數test 加註解時使用方式
test_remote = test.remote(test)
 
if __name__ == "__main__":
    ray.init(address='auto', _redis_password='5241590000000000')
    start_time = time.time()
    img = cv2.imread("/tmp/test.jpg")
    futures = [test_remote.remote(img) for i in range(100)]
    ray.get(futures)
    start_time = time.time()
    print(time.time() - start_time)

在主節點運行程序，程序會在3臺機器上啟動。停止ray服務的方式：ray stop.

到此這篇關於支持python的分佈式計算框架Ray的文章就介紹到這瞭,更多相關python的分佈式計算框架Ray內容請搜索WalkonNet以前的文章或繼續瀏覽下面的相關文章希望大傢以後多多支持WalkonNet！

推薦閱讀：