Hugging Face Pipeline behind Proxies - Windows Server OS - windows

Question

I am trying to use the Hugging face pipeline behind proxies.
Consider the following line of code
from transformers import pipeline
sentimentAnalysis_pipeline = pipeline("sentiment-analysis")
The above code gives the following error.
HTTPSConnectionPool(host='huggingface.co', port=443): Max retries exceeded with url: /distilbert-base-uncased-finetuned-sst-2-english/resolve/main/config.json (Caused by ProxyError('Your proxy appears to only use HTTP and not HTTPS, try changing your proxy URL to be HTTP. See: https://urllib3.readthedocs.io/en/1.26.x/advanced-usage.html#https-proxy-error-http-proxy', SSLError(SSLError(1, '[SSL: WRONG_VERSION_NUMBER] wrong version number (_ssl.c:1091)'))))
I tried to check the proxy on my machine having OS - "windows server 2016 Datacenter" using the following code.
import urllib.request
print(urllib.request.getproxies())
The output is as follows:
{'http': 'http://12.10.10.12:8080', 'https': 'https://12.10.10.12:8080', 'ftp': 'ftp://12.10.10.12:8080'}
However, as per the documentation from urlib3 page, the above setting is incompatible and the problem lies in the https setting :
{
"http": "http://127.0.0.1:8888",
"https": "https://127.0.0.1:8888" # <--- This setting is the problem!
}
and the right setting is
{ # Everything is good here! :)
"http": "http://127.0.0.1:8888",
"https": "http://127.0.0.1:8888"
}
How can we change the proxy setting from "https": "https://127.0.0.1:8888" to "https": "http://127.0.0.1:8888" in a windows OS?
I tried by setting the windows environment variable name as "https_proxy" and the variable values as http://127.0.0.1:8888. However, It is not working.

Answer 1

I found the solution and it is pretty simple. Include the following lines in your python script/notebook. Change the proxy_url and port as per your setting. I hope it helps, someone in the community.
import os
os.environ['HTTP_PROXY'] = 'http://proxy_url:proxy_port'
os.environ['HTTPS_PROXY'] = 'http://proxy_url:proxy_port'

Answer 2

Method 1 (recommended): pypac
pip install pypac
Monkeypatch the requests libary.
import requests
import pypac
def request(method, url, **kwargs):
with pypac.PACSession() as session:
return session.request(method=method, url=url, **kwargs)
requests.request = request
transformers should now work as expected.
from transformers import pipeline
sentimentAnalysis_pipeline = pipeline("sentiment-analysis")
Method 2: Disable SSL verification
WARNING: This method could expose you to malicious attacks.
Disabling SSL verification is a bad idea in general, but the story even worse here because (afaik) transformers may download code and run exec on it. This opens the door for a man-in-the-middle to execute arbitrary code on your machine.
This is probably a very bad idea unless you really know what you're doing or you don't care at all about your machine's security. Otherwise, do not use this method.
import requests
import functools
requests.request = functools.partial(requests, verify=False)
Explaination
Setting the HTTP_PROXY and HTTPS_PROXY environment variables might not be enough to get through your corporate firewall. It may be using a .pac to autoconfigure its proxy on your machine. Browsers pick file up and use it automatically, as do some development tools (e.g. JetBrains). The requests library does not appear to do so.
Fortunately, there is a library called PyPAC that can does this for you. But you'll need to monkeypatch requests to use PyPAC's request method rather than its own.
You don't need to patch requests.get since that function delegates to requests.request anyway.