Creates, deploys, and debugs Hugging Face Spaces (Gradio, Docker, or Static SDKs, ZeroGPU, buckets, inference providers). Use when hosting an app on Spaces, porting a repo, fixing Space builds, or configuring ZeroGPU. Not for LoRA/QLoRA training (peft). Never set hardware via README YAML — use --flavor or hf spaces settings.
Creates, deploys, and debugs Hugging Face Spaces (Gradio, Docker, or Static SDKs, ZeroGPU, buckets, inference providers). Use when hosting an app on Spaces, porting a repo, fixing Space builds, or configuring ZeroGPU. Not for LoRA/QLoRA training (peft). Never set hardware via README YAML — use --flavor or hf spaces settings.
Create or host an application on Hugging Face Spaces.
Port existing code (GitHub repo, paper implementation, local script) onto a Space.
Debug a broken Space (build errors, runtime errors, ZeroGPU quota, silent fallbacks).
Configure hardware (ZeroGPU, dedicated GPU, cpu-basic) or persistent storage (buckets).
Set up inference providers as a zero-VRAM alternative to hosting a model.
Request a community grant for ZeroGPU on a non-PRO account.
Hugging Face Spaces host machine-learning applications — over 1M exist today. Each Space is a git repo. This skill covers creating, building, debugging, and maintaining them.
Note whoami flags: Check canPay and isPro — these gate hardware choices (dedicated GPUs require canPay=True; ZeroGPU requires PRO / Team / Enterprise).
Companion skill: The hf-cli skill teaches an agent every hf command. Install it with:
hf skills add hf-cli
Add --claude --global to install for Claude Code as well, user-level.
Procedure
1. Understand what a Space is
A Space is a git repo with three possible SDKs:
Gradio — most Spaces. Python, fast iteration, supports ZeroGPU.
Docker — arbitrary container. Use when you need a non-Python stack or a pre-built template (Streamlit, Argilla, Shiny, etc. — full list at https://huggingface.co/docs/hub/spaces-sdks-docker). Does support ZeroGPU.
not
Static — plain HTML, or a React/Svelte/Vue project built at deploy time. Use for in-browser ML (transformers.js / WebGPU / WebAssembly / onnxruntime-web), project pages, interactive reports, or Spaces that orchestrate other Spaces. No hardware needed.
Hardware tiers
Tier
Identifier
Cost
Notes
CPU basic
cpu-basic
Free
2 vCPU / 16 GB. Data viz, API-proxy, small CPU models.
ZeroGPU
zero-a10g
Free for creator
Dynamic per-request GPU on NVIDIA RTX PRO 6000 Blackwell (sm_120). Sizes: large (half MIG, 48 GB, 1× quota) and xlarge (full, 96 GB, 2× quota). Visitors consume their own daily quota (~5 min free / 40 min Pro / 60 min Enterprise). Gradio-only, PyTorch-first. Requires PRO / Team / Enterprise.
Dedicated GPU
T4, L4, A10G, L40S, A100, H200
Billed hourly to creator
Only creator can attach; requires canPay=True. Use when ZeroGPU genuinely doesn't fit.
Static
(none)
Free
No hardware needed.
List + pricing for dedicated GPUs:
hf spaces hardware
Non-PRO user wanting ZeroGPU? Create a cpu-basic Space, code the app for ZeroGPU, push, then request a community grant. See references/grants.md — load it whenever the user lacks PRO but needs GPU.
Before deciding how to build anything, search for prior art:
hf spaces search "<model name or task>" --sdk gradio --limit 10
If someone has built a similar Space, read its app.py and requirements.txt — that gives you the working pattern. Mention to the user what you found before committing to an approach.
3. Decide SDK and hardware
Follow the user's explicit request first. If they were vague:
Default for a public ML demo: Gradio + ZeroGPU. Use this unless something below applies.
The model's only inference path is non-PyTorch (ONNX / TF / JAX / vLLM as the MAIN model, with heavy init): dedicated GPU.
Exception: marginal non-torch tools (a small ONNX preprocessor, a TF utility) inside a torch-main pipeline are fine on ZeroGPU. The hijack only patches torch; init the non-torch lib inside @spaces.GPU and pay the short per-call init cost.
Tiny / CPU-bound model, or API-proxy Space:cpu-basic.
Browser-side ML or project page: Static.
Container with non-Python stack: Docker.
Sourcing the model
GitHub repo — clone locally to read structure. If it already has a Gradio demo, the minimal viable path is to adapt it onto ZeroGPU (load references/zerogpu.md now). Otherwise: read the README + inference code, prefer the PyTorch path, estimate VRAM (bf16 ≈ params_B × 2 GB; 48 GB fits ≤24B params at bf16, or much larger with quantization — see references/zerogpu.md for quantization on ZeroGPU).
HF model repo — read its README, follow any linked GitHub.
Paper / blog post — look for an official or unofficial implementation. Don't reimplement unless trivial or the user explicitly asks.
If the model genuinely won't fit, check Inference Providers as an alternative: load references/inference-providers.md. This avoids hosting the model at all.
--flavor selects hardware. zero-a10g is the (legacy) identifier for ZeroGPU. Omit for cpu-basic. Run hf spaces hardware for the full paid list and pricing.
Visibility: --public (anyone can view), --private (only you), --protected (app is reachable but git repo / Files tab is private).
--secrets KEY=val becomes an environment variable inside the Space and is not visible to visitors. Use for API keys, gated-repo tokens (HF_TOKEN=hf_…), etc. Can also be set later via hf spaces secrets set <id> KEY=val.
--env KEY=val is visible to visitors — use only for non-sensitive config (GRADIO_SSR_MODE=false, PYTORCH_CUDA_ALLOC_CONF=expandable_segments:True, etc.).
HARD RULE:hardware: in the README YAML is silently ignored — hardware is only set via --flavor at creation, or later via hf spaces settings <id> --hardware <name>.
5. Build the app
The Space now exists at https://huggingface.co/spaces/<namespace>/<name> but is empty.
README.md frontmatter (always required)
---title:...emoji:🚀# pick something representativecolorFrom:blue# red|yellow|green|blue|indigo|purple|pink|gray (only these)colorTo:indigosdk:gradio# gradio | docker | staticsdk_version:6.15.1# latest stable unless you have a reason*app_file:app.py# gradio only (docker / static use Dockerfile / index.html)short_description:...# ≤ 60 chars (server rejects longer)python_version:"3.12"# ZeroGPU officially supports 3.10.13 and 3.12.12startup_duration_timeout:30m# default; bump to 1h for big LLMs / heavy downloads---
* Reasons to use an older Gradio: a custom component pins it, or you're adapting an existing demo and don't want to rewrite for 5.x→6.x breaking changes. If you need a 5.x, pick 5.50.0 (latest of the series; still supports custom components).
import spaces # MUST come before torch / diffusers / transformersimport torch
import gradio as gr
from diffusers import DiffusionPipeline
pipe = DiffusionPipeline.from_pretrained("<repo>", torch_dtype=torch.bfloat16).to("cuda")
@spaces.GPU(duration=60)defgenerate(prompt):
return pipe(prompt).images[0]
gr.Interface(fn=generate, inputs=gr.Text(), outputs=gr.Image()).launch()
Three ZeroGPU rules (full treatment in references/zerogpu.md — load it whenever the Space targets ZeroGPU):
import spaces before torch / any CUDA-touching import. It monkey-patches torch.cuda.*; once CUDA is initialized in the main process, it's too late.
Load the model at module scope, .to("cuda") eagerly. ZeroGPU intercepts the call, packs weights to disk, and streams them into VRAM on the first @spaces.GPU entry. Lazy loading inside the decorator costs every user.
Decorate the function Gradio binds. Estimate duration to the realistic worst case (smaller = higher queue priority and tighter quota check). For input-dependent runtime, pass a callable.
requirements.txt
Do NOT list (preinstalled and platform-managed; pinning them causes resolution failures or silently breaks the ZeroGPU runtime):
ZeroGPU torch versions: Only 2.8.0, 2.9.1, 2.10.0, 2.11.0 are accepted. Default to leaving torch unpinned (the runtime preinstalls the latest). Only pin when a dep forces it.
Prebuilt CUDA-extension wheels (flash_attn, xformers, pytorch3d, nvdiffrast, diff_gaussian_rasterization, torchmcubes): use the prebuilt Blackwell wheels at https://huggingface.co/datasets/multimodalart/zerogpu-blackwell-wheels/tree/main/wheels. Full mapping + caveats in references/requirements.md — load it whenever you need to pin deps or source wheels.
Per-SDK depth references
SDK / topic
When to load
File
Gradio patterns (themes, gr.Examples, streaming, custom HTML components, gr.Server)
ZeroGPU specifics (decorator semantics, sizing, AoTI, generators, concurrency, pickle / gr.State across the worker boundary)
Space targets ZeroGPU
references/zerogpu.md
For built SPAs (Static), set app_build_command: npm run build and app_file: dist/index.html in frontmatter.
6. Iterate on the Space, not locally
Try to build a release candidate from the user quest locally and push it — then use the live URL as your test loop. The Space environment is the only one that matters; do not try to test locally. python3 -m py_compile app.py is the maximum local check worth doing before pushing.
Once pushed, pick the cheapest update mechanism for each change:
Hot-reload for pure Python edits.
hf upload for code-only files hot-reload can't touch.
Full rebuild only when requirements.txt / Dockerfile / README frontmatter actually changed.
Full ladder + footguns (hot-reload poisoning factory reboot, runtime.sha lag, etc.) in references/debugging.md — load it whenever you need to iterate or debug.
7. Permanent storage (buckets)
Spaces are stateless — /data is wiped on restart. If the Space needs to persist user uploads, generations, logs, or interact with a long-lived store, mount a bucket:
Buckets are paid storage; check canPay and confirm with the user. Full patterns (read-fast / write-durable, public bucket URLs, model-cache anti-pattern) in references/buckets.md — load it whenever the user needs persistent storage.
Verification
Don't trust RUNNING alone — the app can be running but broken. Four steps, in order:
A. Alive? Stage + hardware:
hf spaces info <ns>/<name> --expand runtime
B. Logs clean post-boot? Read the run log to confirm startup finished without warnings or silent fallbacks:
hf spaces logs <ns>/<name> --tail 200
Look for: model-load completion, no import warnings, no "falling back to CPU" / dtype downgrade messages, no RUNNING masking a half-broken app.
C. API actually responds. With logs still tailing in another terminal (hf spaces logs <ns>/<name> --follow), call the endpoint:
from gradio_client import Client, handle_file
import os
c = Client("<ns>/<name>", token=os.environ["HF_TOKEN"], httpx_kwargs={"timeout": 600})
print(c.view_api()) # discover endpoints — don't guess
result = c.predict(..., api_name="/generate")
D. Sniff output AND logs. HTTP 200 ≠ correct output. Check both:
And look at the run log emitted during the call — silent fallbacks (model snapping to a different size, missing optional dep, dtype downgrade) only show up there.
Full smoke-test patterns (streaming endpoints, OAuth-gated Spaces, gr.Server custom routes) in references/debugging.md.
Pitfalls
ZeroGPU-specific pitfalls
import spaces must come first. If any CUDA-touching import (torch, diffusers, transformers) runs before spaces, the monkey-patch fails silently and the Space will not get GPU allocation.
Lazy loading inside @spaces.GPU is an anti-pattern. Every user pays the model-load cost on every call. Load at module scope.
duration too large wastes quota; too small kills long runs. Estimate to realistic worst case. Smaller duration = higher queue priority.
ZeroGPU is Gradio-only and PyTorch-first. Docker Spaces and non-torch main models do not work on ZeroGPU.
Torch version mismatch. Only 2.8.0, 2.9.1, 2.10.0, 2.11.0 are accepted on ZeroGPU. Pin only when forced.
Pinning gradio, spaces, or huggingface_hub in requirements.txt causes resolution failures or silently breaks the ZeroGPU runtime. Never list them.
General Space pitfalls
hardware: in README YAML is silently ignored. Hardware is only set via --flavor at creation or hf spaces settings <id> --hardware <name> later.
--env values are visible to visitors. Never put secrets in --env; use --secrets instead.
short_description > 60 chars is rejected by the server. Keep it concise.
colorFrom / colorTo only accept:red, yellow, green, blue, indigo, purple, pink, gray. Any other value is rejected.
/data is wiped on restart. Spaces are stateless. Use buckets for persistence.
RUNNING does not mean working. Always verify with logs + API call + output sniff.
Hot-reload can poison a factory reboot. Use the cheapest correct rung from references/debugging.md — don't blindly hot-reload everything.
runtime.sha lag can cause stale builds. Full details in references/debugging.md.
Prebuilt CUDA wheels must come from the Blackwell wheel dataset (https://huggingface.co/datasets/multimodalart/zerogpu-blackwell-wheels/tree/main/wheels), not PyPI. See references/requirements.md.
Grep references/known-errors.md for the error string. Check if this is a known issue before trying your own fix — most common ZeroGPU / Gradio / dependency errors have a 1–2 line fix there. Load references/known-errors.md whenever you encounter an unfamiliar error.
Iterate using the cheapest rung from references/debugging.md. The vast majority of issues resolve with log-reading + smoke-test loops; interactive dev mode + SSH is a heavy-hammer last resort.
If you solve an error that wasn't in the known-errors list, suggest the user PR it back to this skill so future runs benefit.
Reference index
When to read
File
How ZeroGPU works + correct patterns (decorator, sizing, pickle, generators, real-time, AoTI)
references/zerogpu.md
Iterate + debug: logs, rung ladder, smoke testing (and dev mode + SSH as a last resort)
references/debugging.md
Error-string lookup — the single place for all error symptoms (Spaces, ZeroGPU, Gradio, deps)
