cd /news/ai-infrastructure/spaces-aren-t-allocating-for-zerogpu… · home topics ai-infrastructure article
[ARTICLE · art-97597] src=discuss.huggingface.co ↗ pub= topic=ai-infrastructure verified=true sentiment=↓ negative

Spaces aren't allocating for ZeroGPU tonight

Hugging Face user AbstractPhil reports that their ZeroGPU Space 'AbstractPhil/alephllm-chat' builds successfully but never receives a hardware allocation, with the runtime API showing 'hardware.current: null' indefinitely since August 14, 2026. The issue also affects a predecessor Space, 'AbstractPhil/alephllm-chat2', which remains stuck with a hardware assignment it was asked to release. The user seeks an explanation and a fix from Hugging Face.

read2 min views1 publishedAug 15, 2026
Spaces aren't allocating for ZeroGPU tonight
Image: Discuss (auto-discovered)

Space never allocated: hardware.current: null

, builds succeed, app never runs

Claude wrote this up to save some time. Been a bit of a journey tonight trying to get it working, so far no dice.

Space: AbstractPhil/alephllm-chat

(Gradio SDK, ZeroGPU requested)

Account: AbstractPhil · Observed: 2026-08-14 → 2026-08-15

Symptom

The Space builds successfully on every commit and is never allocated a

machine. Runtime API reports:

stage    : APP_STARTING        (indefinitely)
hardware : {"current": null, "requested": "zero-a10g"}
errorMessage : null

current: null

persists for hours. Switching the request to cpu-basic

does not allocate either, so this is not ZeroGPU-specific.

Build is not the problem

Build log completes normally every time:

--> RUN --mount=target=/tmp/requirements.txt ... CACHED
--> COPY --link --chown=1000 ./ /home/user/app   DONE
--> Pushing image                                 DONE 0.8s
--> Exporting cache                               DONE 0.5s

The app is not the problem

The identical app.py

runs correctly outside Spaces (same Python 3.12,

same gradio, same spaces

package): it boots, serves /config

, and

answers requests. It is the canonical ZeroGPU shape — model loaded at

module level and moved to cuda

, @spaces.GPU

on the generator, plain

demo.launch()

. No threads, no locks, no asyncio, no coroutines, no

subprocesses of our own.

Related history on this account (may be the same underlying issue)

The predecessor Space (now AbstractPhil/alephllm-chat2

) spent ~16h

serving a two-day-old commit: new builds succeeded, containers

booted cleanly with Running on local URL

, but traffic kept going to

the old revision and runtime.sha

never advanced. Factory rebuild and

restart had no effect. Recreating the Space under a new name worked

once, then the new Space hit the current: null

state described above.

alephllm-chat2

is now D but still reports

hardware.current: zero-a10g

, with requested: cpu-basic

stuck in

flight. Re-issuing the hardware change and restarting the Space does

not apply it. It appears to be holding a hardware assignment it was

asked to release.

Similar reports

Ask

  • Why is no machine being allocated for AbstractPhil/alephllm-chat

despite a successful build and a pending hardware request?

  • Can the stuck hardware assignment on AbstractPhil/alephllm-chat2

(current: zero-a10g

while D, requested: cpu-basic

) be

released?

Additionally, what is this TWO LINKS PER POST thing for new users? I’m nowhere near a new user for huggingface.

Well on the plus side, the AlephLLM Mini-Beatrix is perfectly playable on cpu. Pretty fun ByteLM structured system turned out pretty smart.

── more in #ai-infrastructure 4 stories · sorted by recency
── more on @hugging face 3 stories trending now
sponsored brought to you by zahid.host 4,200+ EU-deployed projects
reading about agents? ship yours in a single git push.

Run your AI side-project on zahid.host

EU-based hosting, git-push deploys, automatic HTTPS, no cold starts. Free tier with a custom domain — perfect for shipping the agent you just read about.

$git push zahid main
Live at https://your-agent.zahid.host
Get free account → Pricing
from €0/mo · no card required
LIVE [news/spaces-aren-t-alloca…] indexed:0 read:2min 2026-08-15 ·