use default HF HUB token when checking for base model info

[noyoshi]

• This is important now that the models are gated, we need to use the token which started the model, since that should have permissions to read the base model repo!

[bi1101]

Hi, does this PR affect the Predibase API? I've been getting error with inferencing finetuned models through the OAI API since roughly 12 hours ago

{
    "error": "Request failed during generation: Server error: Unable to locate credentials",
    "error_type": "generation"
}