File size: 3,285 Bytes
5c6c847 cf9d631 5c6c847 | 1 2 3 4 5 6 7 8 9 10 11 12 13 14 15 16 17 18 19 20 21 22 23 24 25 26 27 28 29 30 31 32 33 34 35 36 37 38 39 40 41 42 43 44 45 46 47 48 49 50 51 52 53 54 55 56 57 58 59 60 61 62 63 64 65 66 67 68 69 70 71 72 73 74 75 76 77 78 79 80 81 82 83 84 85 86 87 88 89 90 91 92 93 | # 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:
```bash
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):
```bash
hf auth login
```
or pass the token from an environment variable, without ever writing it to a file:
```bash
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:
```bash
hf auth whoami
```
## 2. Create the public model repository
```bash
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:
```bash
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
```bash
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/<your-username>/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.
|