CODEWITHSUNDEEP
python-3.xhuggingface-transformersbert-language-modelhuggingface-tokenizershuggingface
Nikita Malviya
Nikita Malviya

Reputation: 1429

SSLError: HTTPSConnectionPool(host='huggingface.co', port=443): Max retries exceeded with url: /dslim/bert-base-NER/resolve/main/tokenizer_config.json

I am facing below issue while loading the pretrained BERT model from HuggingFace due to SSL certificate error.

Error:

SSLError: HTTPSConnectionPool(host='huggingface.co', port=443): Max retries exceeded with url: /dslim/bert-base-NER/resolve/main/tokenizer_config.json (Caused by SSLError(SSLCertVerificationError(1, '[SSL: CERTIFICATE_VERIFY_FAILED] certificate verify failed: self signed certificate in certificate chain (_ssl.c:1108)')))

The line that is causing the issue is:

tokenizer = AutoTokenizer.from_pretrained("dslim/bert-base-NER")

Source code:

from transformers import AutoTokenizer, AutoModelForTokenClassification
from transformers import AutoModelForCausalLM, AutoTokenizer

tokenizer = AutoTokenizer.from_pretrained("dslim/bert-base-NER")
model = AutoModelForTokenClassification.from_pretrained("dslim/bert-base-NER")

I am expecting to download pre-trained models while running the code in jupyter lab on Windows.

Upvotes: 17

Views: 71164

Answers (9)

Harpreet
Harpreet

Reputation: 175

Please check if you have the access token to download the hugging face model.

Please refer the video https://www.youtube.com/watch?v=t-0s_2uZZU0 and check the information given between timestamps 1:44:18 - 1:45:34

Upvotes: 0

Satish
Satish

Reputation: 721

import os

os.environ['CURL_CA_BUNDLE'] = r"C:/examples/ssls/all.crt"
os.environ['REQUESTS_CA_BUNDLE'] = r"C:/examples/ssls/all.crt"

Note this error may come for other connections also.

  1. Download all those certificates from browser address browser padlock.

  2. Download the "chain" not just individual (it appeared 2nd option in combo while save). The question of this post has error related to chain. Because verification will happen till root CA. Probably it was downloaded as single cert. (refer screenshot)

  3. Multiple .crt files if you have concat all using concat command (comes with git bash)

    concat a.crt b.crt c.crt > all.crt.

enter image description here

Upvotes: 1

Trex
Trex

Reputation: 640

Here's a solution that worked for me:

  1. Access the huggingface.co certificate by clicking on the icon beside the web address in your browser (screenshot below) > 'Connection is secure' > Certificate is valid (click show certificate).

enter image description here

  1. Download the certificate 'Details' > Export. Export the entire certificate (change file type).

enter image description here

  1. Include the path to the certificate in your code:
import os
os.environ['CURL_CA_BUNDLE'] = 'C:/Users/xxxx/Downloads/certificates/huggingface.co.crt'
from transformers import AutoTokenizer
tokenizer = AutoTokenizer.from_pretrained('BAAI/bge-reranker-v2-m3')

Upvotes: 1

SOWNDARIYA M
SOWNDARIYA M

Reputation: 31

By setting the below environment variables it worked:

import os
os.environ['CURL_CA_BUNDLE'] = ''
os.environ['REQUESTS_CA_BUNDLE'] = ''

Upvotes: 1

oyste
oyste

Reputation: 61

Here is a solution if you want the actual certificate:

If you are on linux you can use this bash script I made to download the certificate file from Cisco Umberella, convert it to .crt and update the certificates folder. Script:

#!/bin/sh

curl -O https://d36u8deuxga9bo.cloudfront.net/certificates/Cisco_Umbrella_Root_CA.cer
openssl x509 -inform PEM -in Cisco_Umbrella_Root_CA.cer -out cisco.crt
rm -rf Cisco_Umbrella_Root_CA.cer
sudo mv cisco.crt /usr/local/share/ca-certificates/
sudo update-ca-certificates

After adding the new certificate from Cisco you still need to set the environment variable CURL_CA_BUNDLE before using the from_pretrained function. This is because the certifi package used by transformers for ca-certification defaults to some .pem file inside the certifi package. We instead set it to the directory your downloaded certificate was just moved to by update-ca-certificates. Here is how:

import os
from transformers import AutoModel, AutoTokenizer

os.environ["CURL_CA_BUNDLE"] = "/etc/ssl/certs/ca-certificates.crt"
model = AutoModel.from_pretrained("intfloat/multilingual-e5-small")
tokenizer = AutoTokenizer.from_pretrained("intfloat/multilingual-e5-small")

Now it should download the model without any security warnings. Hope this helped.

Upvotes: 6

Zaid Siddiqui
Zaid Siddiqui

Reputation: 83

I was working with my linux machine ,Looking at server of hugging face found that they were working fine ,it was just the issue of proxy directly export proxy in .zshrc work for me.

export https_proxy=http://127.0.0.1:7890
export http_proxy=http://127.0.0.1:7890
export all_proxy=socks5://127.0.0.1:7890

using source ~/.zshrc I tried downgrading request library but didn't worked ,by managing the proxy it did work for me .

Upvotes: 0

Pitts
Pitts

Reputation: 1

A working solution for this is Enable your device for development.

if you're writing software with Visual Studio on a computer for the first time, you will need to enable Developer Mode on both the development PC and on any devices you'll use to test your code.

Upvotes: 0

Timbus Calin
Timbus Calin

Reputation: 15033

A working solution to this is the following:

1)

 pip install requests==2.27.1

Please note that both prerequisites need to take place, just running the following code will not work unless you have that specific version of requests (I had to downgrade in my case from 2.29.0 to 2.27.1.

2)

import os
os.environ['CURL_CA_BUNDLE'] = ''

Upvotes: 24

Herb
Herb

Reputation: 171

Try to add the following in your main python file

import os

os.environ['CURL_CA_BUNDLE'] = ''

Upvotes: 11

Related Questions