Integrations / Hugging Face Hub

ConnectDocumented

Serve supported Hugging Face models on Badgr

Pick a model from the Hub and start it as a service. Check the licence and any access gate before you start.

The steps follow the tool's documented settings. No tested run is recorded for this page yet.

Why use Badgr for Hugging Face Hub

Badgr interface
Run
Where Hugging Face Hub fits
Model source
Why Badgr
Execute or serve models
  • Go from a model id on the Hub to a running service with one command, without choosing a GPU by hand.
  • A spending cap and a stop command bound what a service can cost.
  • You get an OpenAI-style address and key for the model you picked.

What you can do

  • Serve an open model you picked on the Hub
  • Try a model from a model card on a larger GPU than you own
  • Keep a private copy of a model available to your app

Quickstart: Serve a model by its Hub id

badgr serve Qwen/Qwen2.5-7B-Instruct --max-cost 5

How it works

  1. 1.You give Badgr a Hub model id
  2. 2.Badgr picks a GPU with enough memory and starts the service
  3. 3.You get an OpenAI-style address and key
  4. 4.Stop it with badgr down when you are done

What you get

  • A running service your app can call
  • A receipt when it stops

Requirements

  • The Badgr CLI, logged in, with a funded balance
  • A model id from the Hub (for example Qwen/Qwen2.5-7B-Instruct)
  • For gated models: accepted licence on the Hub and an access token
  • A model architecture the serving software supports

What it costs

  • A service bills for as long as it runs, until you stop it.
  • Pass --max-cost to bound spend. Larger models need larger GPUs, which cost more per hour.
  • Badgr shows the hourly rate before starting. This page does not quote one.

See pricing for how billing works.

Proof

No tested run is recorded for this integration yet. A receipt and example output will be added here after a real run, not before.

Troubleshooting

  • Gated model will not download

    Accept the licence on the model's Hub page and provide a token with read access. Without both, the download fails.

  • Out of GPU memory

    Choose a smaller or quantised model, or a GPU with more memory.

  • Architecture not supported

    Check the model card and the serving software's supported list. Run badgr serve --list-aliases for models known to work.

  • Licence does not allow your use

    Read the licence on the model card. Badgr does not change a model's terms.

Developer reference

Ready to start with Hugging Face Hub?

Every run or service has a spending limit you set before it starts.