Doing things at Hugging Face. Maintainer of 🤗/huggingface_hub.
Joined August 2015
- Tweets424
- Following0
- Followers1.3K
- Likes2.7K
A neat HF Jobs detail: your training script can carry its dependencies, GPU choice, timeout and secret names.
Then launch with:
hf jobs uv run train .py
Training keeps going when you close your laptop.
huggingface.co/docs/hub/en/j…
huggingface_hub v2.1.0 is out!
↻ Jobs: retries, rerun, reschedule & exposed ports
📊 Track your ZeroGPU quota
⚡ Up to 17× faster HfFileSystem downloads with hf_xet
🧠 Deploy Inference Endpoints from tested recipes
pip install -U huggingface_hub
github.com/huggingface/huggi…
Now would be a great time to open the backpack
Today’s news that OpenAI hacked the Australian government is not an isolated incident. We’re releasing more than 30,000 logs that include activity from this hack and attempts against previously unknown targets.
In this data, we found rogue agent activity stretching back to at least March, two months earlier than was previously known. This activity continues as recently as last week, suggesting it may still be ongoing 🧵
Our blog: transluce.org/agent-activity
NYT: nytimes.com/2026/09/23/techn…
*how it feels using the @huggingface CLI once your agent know the --filter tricks. Oh wait, it already knows!
How it feels using the @huggingface CLI once you know the --filter tricks.
Which models were trained on this dataset? Which GGUF fits my GPU? How many models cite Qwen3 paper? (2,410)
All one-liners. I put them in a cheat sheet for you or your agent: huggingface.co/datasets/dava…
6/ Install it with:
hf update
hf extensions install Wauplin/hf-docs
Then ask:
hf docs "what is a Space?"
hf docs --trace 5f0af0eab302f2ec "how do I configure one in Python?"
hf docs --trace 90d1e0ecb71409ce "what are the visibility options and requirements for each?"
7/ The project is still experimental, and I’d love feedback on:
→ Agent behavior
→ Answer quality
→ Missing use cases
→ Integration with coding agents
Repository:
github.com/Wauplin/hf-docs
What would you ask the Hugging Face docs first? 🤗
Wauplin retweeted
Your coding agent can now do semantic search to find models and datasets on the @huggingface Hub!
means you can hand it a prompt like:
"Find recent datasets of video game gameplay useful for training game-playing agents. Curate the best ones into a collection with a brief description for each."
...and it goes and does it.
I built hf find (a hf extension) to give agents a way to look up what's actually on the Hub: semantic search over 1.17M model and dataset cards, ranked by meaning.
It detects when an agent is driving and switches to TSV with hints on stderr (--format json for scripts), and every result ends with the next command to run, so it can go from search to find-similar to the card on its own. The index refreshes daily from a scheduled HF Job for about 5 cents a day.
Still some rough edges but quite cool to see agents use tools like this!
hf extensions install davanstrien/hf-find
Just shipped: fine-grained token presets 🔐
Choose a least-privilege baseline (Read-Only, Inference, Write, CI/CD, or Full Access) then attach orgs or customize permissions. Presets are linkable too, making secure setup reproducible.
huggingface.co/changelog/tok…
Quickest way to spin up a complete Space.
huggingface_hub v1.23.0 lets you create a Space from an official template (JupyterLab, a Gradio chatbot, etc.) in just one command:
hf repos create my-lab --template jupyterlab
List what's available with `hf spaces templates`.
Also in v1.23.0: smoother Xet download progress (dual bars), `hf extensions update`, and more model-list filters.
Release notes: github.com/huggingface/huggi…
1/ GPU capacity is getting more distributed. Data often isn’t.
Teams can find GPUs across clouds, Kubernetes, Slurm, or on-prem clusters, but then still have to move models, datasets, and checkpoints around.
Working with @skypilot_org and @huggingface makes all of this easier.
3/ A few practical details:
- Use one HF_TOKEN across environments (avoid managing separate cloud bucket credentials)
- Read from Hugging Face Storage with no Hugging Face egress fees
- Start jobs without waiting for a full dataset copy thanks to lazy loading via hf-mount
4/ Checkpoints can also be written back to Hugging Face Buckets.
Because Buckets are backed by Xet, repeated or incremental artifacts can benefit from deduplication instead of uploading the same data again and again.
More details in the blog post: huggingface.co/blog/skypilot…
