Error 401 Client Error: Unauthorized for url

🤗Hub
42

I get " HTTPError: 401 Client Error: Unauthorized for url: https://huggingface.co/Whisper%20Small%20manju%20Mal/resolve/main/config.json"

Did you solve your problem.

pls help

43

You may need to set a token for the model e processor.

44

Hi, @Deepakkori45
You must generate a “write” token under your access tokens page

write_token
write_token648×528 20.2 KB

then, filling that token by below cli command

huggingface-cli login

45

model = AutoModelForCausalLM.from_pretrained(
“meta-llama/Llama-2-7b-hf”,
token=‘’
)

try passing your token here it works like a charm

1 Like
46

This worked for me

47

Pass token as an extra arg. (..., token=hf_token)

Source: User access tokens

48

Greate!!! it works for me.

49

I want to create a Server-less inference endpoint from a model directly from the HUB. I have already accepted the model conditions. Here is the code I am using (reference):

import pandas as pd
import boto3
import sagemaker
import time
import json
import io
from io import StringIO
import base64
import re
import os

from sagemaker.image_uris import retrieve
from sagemaker.huggingface import HuggingFaceModel
import sagemaker 

from huggingface_hub import login

login(token = os.environ["HUGGING_FACE_READ_TOKEN"],add_to_git_credential=True)


sess = sagemaker.Session()

region = sess.boto_region_name
s3_client = boto3.client("s3", region_name=region)
sm_client = boto3.client("sagemaker", region_name=region)
sm_runtime_client = boto3.client("sagemaker-runtime")
sagemaker_role = sagemaker.get_execution_role()

model_name = "llama-3-1-8B"
hub = {
    'HF_MODEL_ID':'meta-llama/Meta-Llama-3.1-8B', # model_id from hf.co/models
    'HF_TASK':'question-answering', # NLP task you want to use for predictions
    'HF_API_TOKEN':os.environ["HUGGING_FACE_READ_TOKEN"],
}
# create Hugging Face Model Class
huggingface_model = HuggingFaceModel(
    env=hub,
    role=sagemaker_role, # iam role with permissions to create an Endpoint
    transformers_version="4.6", # transformers version used
    pytorch_version="1.7", # pytorch version used
    py_version="py36", # python version of the DLC
    name = model_name
)

predictor = huggingface_model.deploy(
    # initial_instance_count=1,
    # instance_type="ml.m5.xlarge",
    container_startup_health_check_timeout=300,
    serverless_inference_config=sagemaker.serverless.ServerlessInferenceConfig(memory_size_in_mb=4096,max_concurrency= 1)
)

I have tested with different permission set tokens but not to avail, the error seems to be related to a specific config file of the model:

requests.exceptions.HTTPError: 401 Client Error: Unauthorized for url: https://huggingface.co/meta-llama/Meta-Llama-3.1-8B/resolve/main/config.json

Python Version:

3.10.14

Sagemaker Version:

2.224.4

Let me know if I should post this issue on AWS instead. I want to use this option if possible or do I have to download the model and create a tar.gz to deploy it?

50

@afelipetu what kind of token did you create? is it a “read” token, or a fine-grained token (the default)? if the latter, you need to make sure to give the permission labeled “Read access to contents of all public gated repos you can access”

51

It is the default read permission, but the problem was that I only got the mail permissions for the 70B model. Even though the model I asked permission for was the 8B, the page says I have permission for it but it doesn’t work. Anyway, with 70B I can deploy, thanks.

1 Like
52

this solved my problem. for the learners out there like me, go to the huggingface tokens page, create a new token, select fine-grained, scroll down and search the specific model you want to access. copy the token and paste it in this command

huggingface-cli login --token (your_token_here) --add-to-git-credential

1 Like