blocks|key|4292184|text|PyCUDA可能与WSGI+web服务器上下文不兼容。如果您使用某种消息队列(如Celery+)，您可能会让它工作，在这种队列中，HTTP请求将一个作业放在队列上，队列另一端的worker运行CUDA程序。|type|unstyled|depth|inlineStyleRanges|entityRanges|data|4292185|编辑:一种快速而简单的方法是使用Python+Subprocess+check_output+function|offset|length|4292186|在web请求中：|4292187|subprocess.check_output(['python',+'cudaFlask.py'])|style|CODE|4292188|entityMap|0|LINK|mutability|MUTABLE|url|https://docs.python.org/3/library/subprocess.html#subprocess.check_output^0|0|G|13|0|0|0|0|1F|0^^$0|@$1|2|3|4|5|6|7|T|8|@]|9|@]|A|$]]|$1|B|3|C|5|6|7|U|8|@]|9|@$D|V|E|W|1|X]]|A|$]]|$1|F|3|G|5|6|7|Y|8|@]|9|@]|A|$]]|$1|H|3|I|5|6|7|Z|8|@$D|10|E|11|J|K]]|9|@]|A|$]]|$1|L|3|-4|5|6|7|12|8|@]|9|@]|A|$]]]|M|$N|$5|O|P|Q|A|$R|S]]]]

PyCUDA may not be compatible with WSGI web server contexts.
You could possibly make it work if you use some kind of message queue like Celery, where the HTTP request places a job on the queue and the worker on the other side of the queue runs the CUDA program.

Edit: A quick and easy way would be to use <a href="https://docs.python.org/3/library/subprocess.html#subprocess.check_output" rel="nofollow noreferrer">Python Subprocess check_output function</a>

In the web request:

<code>subprocess.check_output(['python', 'cudaFlask.py'])</code>

blocks|key|4292330|text|解决了flask中的延迟加载和手动制作context的问题(即在PyCUDA中没有pycuda.autoinit+)。|type|unstyled|depth|inlineStyleRanges|offset|length|style|CODE|entityRanges|data|4292331|有关flask中的延迟加载，请参阅this。|4292332|我的views.py文件：|4292333|import+numpy+as+np
import+pycuda.driver+as+cuda
from+pycuda.compiler+import+SourceModule

def+index():
++++cuda.init()
++++device+=+cuda.Device(0)+#+enter+your+gpu+id+here
++++ctx+=+device.make_context()

++++mod+=+SourceModule("""
++++++++int+x+=+4;
++++""")

++++ctx.pop()+#+very+important

++++print+("done")
++++return+"success"|code-block|syntax|javascript|4292334|entityMap|0|LINK|mutability|MUTABLE|url|http://flask.pocoo.org/docs/1.0/patterns/lazyloading/^0|3|5|J|7|W|6|15|F|0|2|5|H|4|0|0|2|8|0|0^^$0|@$1|2|3|4|5|6|7|W|8|@$9|X|A|Y|B|C]|$9|Z|A|10|B|C]|$9|11|A|12|B|C]|$9|13|A|14|B|C]]|D|@]|E|$]]|$1|F|3|G|5|6|7|15|8|@$9|16|A|17|B|C]]|D|@$9|18|A|19|1|1A]]|E|$]]|$1|H|3|I|5|6|7|1B|8|@$9|1C|A|1D|B|C]]|D|@]|E|$]]|$1|J|3|K|5|L|7|1E|8|@]|D|@]|E|$M|N]]|$1|O|3|-4|5|6|7|1F|8|@]|D|@]|E|$]]]|P|$Q|$5|R|S|T|E|$U|V]]]]

Solved the issue with lazy loading in <code>flask</code> and making the <code>context</code> manually (i.e. without <code>pycuda.autoinit</code> in <code>PyCUDA</code>.

Refer <a href="http://flask.pocoo.org/docs/1.0/patterns/lazyloading/" rel="noreferrer">this</a> for lazy loading in <code>flask</code>.

My <code>views.py</code> file:

<pre><code>import numpy as np
import pycuda.driver as cuda
from pycuda.compiler import SourceModule

def index():
 cuda.init()
 device = cuda.Device(0) # enter your gpu id here
 ctx = device.make_context()

 mod = SourceModule("""
 int x = 4;
 """)

 ctx.pop() # very important

 print ("done")
 return "success"
</code></pre>

blocks|key|4292317|text|根据您的解决方案，我将代码从|type|unstyled|depth|inlineStyleRanges|entityRanges|data|4292318|def+print_device_info():+
++++(free,total)=drv.mem_get_info()
++++print("Global+memory+occupancy:%25f%25%25+free"%25(free*100/total))

++++for+devicenum+in+range(cuda.Device.count()):
++++++++device=drv.Device(devicenum)
++++++++attrs=device.get_attributes()

++++++++#Beyond+this+point+is+just+pretty+printing
++++++++print("\n===Attributes+for+device+%25d"%25devicenum)
++++++++for+(key,value)+in+attrs.iteritems():
++++++++++++print("%25s:%25s"%25(str(key),str(value)))|code-block|syntax|javascript|4292319|至|4292320|def+print_device_info():+
++++drv.init()
++++device+=+drv.Device(0)+#+enter+your+gpu+id+here
++++ctx+=+device.make_context()

++++(free,total)=drv.mem_get_info()
+++++print("Global+memory+occupancy:%25f%25%25+free"%25(free*100/total))
+++++attrs=device.get_attributes()

+++++#Beyond+this+point+is+just+pretty+printing
+++++print("\n===Attributes+for+device+%25d"%250)
+++++for+(key,value)+in+attrs.items():
++++++++print("%25s:%25s"%25(str(key),str(value)))
+++++ctx.pop()|4292321|它就像护身符一样起作用。非常感谢你分享你的解决方案，这真的让我很开心！|4292322|entityMap^0|0|0|0|0|0^^$0|@$1|2|3|4|5|6|7|O|8|@]|9|@]|A|$]]|$1|B|3|C|5|D|7|P|8|@]|9|@]|A|$E|F]]|$1|G|3|H|5|6|7|Q|8|@]|9|@]|A|$]]|$1|I|3|J|5|D|7|R|8|@]|9|@]|A|$E|F]]|$1|K|3|L|5|6|7|S|8|@]|9|@]|A|$]]|$1|M|3|-4|5|6|7|T|8|@]|9|@]|A|$]]]|N|$]]

According to your solution I changed my code from
<pre><code>def print_device_info(): 
 (free,total)=drv.mem_get_info()
 print(&quot;Global memory occupancy:%f%% free&quot;%(free*100/total))

 for devicenum in range(cuda.Device.count()):
 device=drv.Device(devicenum)
 attrs=device.get_attributes()

 #Beyond this point is just pretty printing
 print(&quot;\n===Attributes for device %d&quot;%devicenum)
 for (key,value) in attrs.iteritems():
 print(&quot;%s:%s&quot;%(str(key),str(value)))
</code></pre>
to
<pre><code>def print_device_info(): 
 drv.init()
 device = drv.Device(0) # enter your gpu id here
 ctx = device.make_context()

 (free,total)=drv.mem_get_info()
 print(&quot;Global memory occupancy:%f%% free&quot;%(free*100/total))
 attrs=device.get_attributes()

 #Beyond this point is just pretty printing
 print(&quot;\n===Attributes for device %d&quot;%0)
 for (key,value) in attrs.items():
 print(&quot;%s:%s&quot;%(str(key),str(value)))
 ctx.pop()
</code></pre>
and it works like a charm.
Thank you so much for sharing your solution, this really made my day !

blocks|key|4292374|text|以非线程模式启动Flask应用程序：|type|unstyled|depth|inlineStyleRanges|entityRanges|data|4292375|app.run(host=HOST,+port=PORT,+debug=False,threaded=False)|code-block|syntax|javascript|4292376|entityMap^0|0|0^^$0|@$1|2|3|4|5|6|7|I|8|@]|9|@]|A|$]]|$1|B|3|C|5|D|7|J|8|@]|9|@]|A|$E|F]]|$1|G|3|-4|5|6|7|K|8|@]|9|@]|A|$]]]|H|$]]

Starting the Flask application in non-threaded mode:
<pre><code>app.run(host=HOST, port=PORT, debug=False,threaded=False)
</code></pre>

I want to run a pyCUDA code on a <code>flask</code> server. The file runs correctly directly using <code>python3</code> but fails when the corresponding function is called using <code>flask</code>.

Here is the relevant code:

cudaFlask.py:

<pre><code>import pycuda.autoinit
import pycuda.driver as drv
import numpy

from pycuda.compiler import SourceModule

def cudaTest():
 mod = SourceModule("""
 int x = 4;
 """)

 print ("done")
 return

if __name__ == "__main__":
 cudaTest()
</code></pre>

server.py (only the part which calls the function):

<pre><code>@app.route('/bundle', methods=['POST'])
def bundle_edges():
 cudaTest()
 return "success"
</code></pre>

On running <code>python cudaFlask.py</code> I get the output <code>done</code> as expected but on starting the server and doing <code>POST</code> request at <code>website/bundle</code> I get the following error on the flask console:

<pre><code>pycuda._driver.LogicError: cuModuleLoadDataEx failed: invalid device context - 
</code></pre>

on the line <code>mod = SourceModule...</code>

Where am I going wrong?
There is a <a href="https://stackoverflow.com/q/49595175/7116413">similar question</a> out there but it has not been answered yet.

pyCUDA with Flask gives pycuda._driver.LogicError: cuModuleLoadDataEx

翻译质量差，导致语言生硬或混乱。

没有提供实际的解决方法或示例。

解答不清晰，无法理解或解决问题。

页面排版不美观，阅读体验差。

文章

问答

视频

学习中心

腾讯云实验室

直播

竞赛

腾讯云代码分析专区

腾讯iOA零信任安全管理系统专区

腾讯云架构师技术同盟交流圈

腾讯云数据库专区

腾讯云顾问专区

腾讯云原生专区

腾讯混元专区

腾讯云TCE专区

腾讯云Lighthouse专区

腾讯云HAI专区

腾讯云Edgeone专区

腾讯云存储专区

腾讯云智能专区

腾讯轻联专区 

腾讯云开发专区

TAPD专区

腾讯轻量云游戏服专区

腾讯云最具价值专家

腾讯云架构师技术同盟

腾讯云创作之星

腾讯云开发者先锋

腾讯云代码助手

云原生构建

TAPD 敏捷项目管理

Cloud Studio

SDK中心

API中心

命令行工具

涵盖代码开发、场景应用、自动测试全流程，助你从零构建专属AI助手

一站式MCP教程库，解锁AI应用新玩法

我想在flask服务器上运行pyCUDA代码。该文件可以直接使用python3正确运行，但在使用flask调用相应函数时会失败。相关代码如下：cudaFlask.py：import pycuda.autoinitimport pycuda.driver as drvimport numpyfrom pycuda.com...

问带烧瓶的pyCUDA提供pycuda._driver.LogicError: cuModuleLoadDataEx
EN

社区

活动

圈层

关于

腾讯云开发者

热门产品

热门推荐

更多推荐

问带烧瓶的pyCUDA提供pycuda._driver.LogicError: cuModuleLoadDataExEN

社区

活动

圈层

关于

腾讯云开发者

热门产品

热门推荐

更多推荐

问带烧瓶的pyCUDA提供pycuda._driver.LogicError: cuModuleLoadDataEx
EN