Download scripts/upload.md from Holmes2011/ZeroPoint: direct link, hf CLI and curl.
- Browser
- Download file 3.29 kB
-
https://huggingface.co/Holmes2011/ZeroPoint/resolve/main/scripts/upload.md
- Command line
-
hf download hf://Holmes2011/ZeroPoint/scripts/upload.md
-
curl -L -o upload.md https://huggingface.co/Holmes2011/ZeroPoint/resolve/main/scripts/upload.md
Publishing ZeroPoint to the Hugging Face Hub
This note records the exact steps used to publish this release tree as a public model repository. It is for the maintainer performing the upload; users of the release do not need it. Never commit a real access token, and never print one in a terminal that records its output.
The commands below were checked against huggingface_hub 1.27.0 (hf CLI).
0. Environment: the local proxy is mandatory here
In this environment huggingface.co is only reachable through the local HTTP proxy at
http://127.0.0.1:7897. Set both proxy variables in every shell that talks to the Hub,
and pin the official endpoint:
export HTTPS_PROXY=http://127.0.0.1:7897
export HTTP_PROXY=http://127.0.0.1:7897
export HF_ENDPOINT=https://huggingface.co
If the proxy is not running on port 7897, the commands below hang and then fail with a
connection or SSL error — start the proxy first. Do not leave a mirror or an offline
HF_ENDPOINT in the environment.
1. Authenticate (token required)
Create a write token at https://huggingface.co/settings/tokens. Then either log in interactively (the token is not echoed):
hf auth login
or pass the token from an environment variable, without ever writing it to a file:
read -rs HF_TOKEN && export HF_TOKEN # paste at the prompt; input is hidden
hf auth login --token "$HF_TOKEN"
hf auth login --token "$HF_TOKEN" stores the token in the local Hugging Face config
(~/.cache/huggingface/), so the later commands pick it up automatically. Verify:
hf auth whoami
2. Create the public model repository
hf repos create <your-username>/ZeroPoint --type model --public
Replace <your-username> with the account (or organization) that owns the release. If the
repository already exists, skip this step and make sure its visibility is public in the
repository settings.
3. Upload the release tree
Upload the contents of repo/ so that README.md, LICENSE, docs/, models/,
experiments/, tests/, etc. land at the repository root:
cd "<release-root>" # the directory containing README.md, LICENSE, models/, docs/ (i.e. the parent of this scripts/ folder)
hf upload <your-username>/ZeroPoint . --type model \
--commit-message "ZeroPoint v0 research prototype: architecture, ZeroCG framework, comparison experiments"
This is a single-commit upload of the whole tree. It includes
tests/golden/zeropoint_ref.json (the fixed reference data required by the self-checks)
and must not include model weights, training runs, corpora, or tokens — see
.gitignore; the release tree itself contains none of those.
If the upload is interrupted, re-run the same command: files already uploaded are skipped and the commit is retried.
4. Verify the published repository
hf repos ls
hf auth whoami
# Confirm the public API view, through the proxy:
curl -sS --proxy http://127.0.0.1:7897 \
"https://huggingface.co/api/models/<your-username>/ZeroPoint" | head -c 2000
Then open <https://huggingface.co//ZeroPoint> in a browser and confirm
that README.md, LICENSE, docs/ZeroPoint.md, docs/ZeroPoint_EN.md, and the
tests/golden/ file are all present.