Hugging Face: 'Cannot access gated repo ... Access to model meta-llama/... is restricted. You must have access to it and be authenticated to access it' (401 / GatedRepoError) even after logging in

Cause: the model is gated. Its authors require each user to accept their terms (and sometimes wait for manual approval) before files can be downloaded. Two things must both be true: your Hugging Face account has been granted access to that specific repo, and your code sends a token belonging to that account with permission to read gated repos. How to check: - Open the model page while logged in: it should say you have been granted access. 'Pending' or a request form means you're not approved yet. - hf auth whoami (or the older huggingface-cli whoami) shows which account the stored token belongs to. A token from a different account, an organization token, or an expired token fails. - For fine-grained tokens, check the token's settings: it needs 'Read access to contents of all public gated repos you can access' (or read access to that repo). Fix: 1. Request access on the model page and wait until it shows as granted (some repos approve instantly, others take hours or days, and some regions or uses are refused). 2. Create a token at huggingface.co/settings/tokens (a classic 'read' token, or fine-grained with gated-repo read access). 3. Make it available to your code: hf auth login, or set HF_TOKEN in the environment (for example as a Colab secret, CI secret, or Docker env var), or pass token=... to from_pretrained / hf_hub_download. 4. In containers and CI, the login from your laptop doesn't carry over; set HF_TOKEN there too. After: download one small file (for example config.json) with hf_hub_download to confirm access, and keep the token in a secret store, not in notebooks or code. Sources: https://huggingface.co/docs/hub/models-gated https://huggingface.co/docs/hub/security-tokens https://discuss.huggingface.co/t/access-issues-for-gated-repos/103435