Hacktoberfest 2026: the issues maintainers tagged for October, open and beginner-friendly. Browse Hacktoberfest issues

SA-3DAO leaderboard evaluation API returns HTTP 503 for all endpoints (CloudFront/awselb origin down); alternative submission route?

Open
  • Is something wrong?
#206 0 comments 0 reactions 0 assignees View on GitHub

Nobody has claimed this yet.

Assessment

Difficulty
4/5
Estimated time
3-5 days
Newbie friendliness
25/100
Issue type
Bug
Clarity
Mostly clear
Activity status
Active
Tech stack
aws, python
Domain
backend-api-design, infrastructure

Research direction

The report identifies the failing CloudFront URLs and the Hugging Face Space at facebook/sa3dao-leaderboard, but names no repository files or tests. Start by checking the Space and the API’s deployment or health status. Done means the maintainers provide an outage status or restore API access, and address whether an alternative submission route or offline evaluation code is available.

Written by the indexing model from the issue text.

Description

Summary

Since at least 2026-10-24, the SA-3DAO leaderboard evaluation API behind
https://huggingface.co/spaces/facebook/sa3dao-leaderboard returns HTTP 503 on
every request. The Space page itself loads fine (HTTP 200, static app,
RUNNING), but all calls to its backing API fail — so no submissions can be
created, and no leaderboard data can be read.

Evidence

Failed on 2026-10-24 09:51 UTC:

Response (truncated headers):

HTTP/2 503
Content-Type: text/html
Server: awselb/2.0
X-Cache: Error from cloudfront
Via: 1.1 98b0bd91dae8cd13230da037f5a4d5e0.cloudfront.net (CloudFront)
X-Amz-Cf-Pop: SEA900-P10
Date: Sat, 24 Oct 2026 09:51:13 GMT

<html><head><title>503 Service Temporarily Unavailable</title></head>
<body><center><h1>503 Service Temporarily Unavailable</h1></center></body></html>

No submission ID was returned; no archive was uploaded; no evaluation started.

Re-checked on 2026-10-25: still HTTP 503.

The outage looks origin-wide, not endpoint-specific

Every path on the distribution returns 503, including:

GET https://d3691cg3j7u5z0.cloudfront.net/            -> 503
GET https://d3691cg3j7u5z0.cloudfront.net/index.html  -> 503
GET https://d3691cg3j7u5z0.cloudfront.net/api/v1/     -> 503

Server: awselb/2.0 together with X-Cache: Error from cloudfront suggests
CloudFront cannot reach its ELB origin (e.g., empty or unhealthy target group).

Reproduction

Reproduced from two independent networks (a workstation and a separate
inference server), so this is unlikely to be a local-network or regional-CDN
issue. Minimal repro:

curl -s -D - "https://d3691cg3j7u5z0.cloudfront.net/api/v1/leaderboard/?limit=1"

The browser's Submit flow fails the same way ("Failed to fetch" on
POST /api/v1/submissions/).

Impact

We have completed inference and packaging for all 900 hidden benchmark objects
across five method variants (FP16 baseline, W4A4, mixed-precision, VQ W2A4,
strict W2A4), following the mesh.glb + OpenCV-convention pose.json submission
format. The packages (~19 GB total) are ready on disk, but we cannot create
any submission while the API is unavailable. Nothing has been uploaded
publicly.

Requests

  1. Status/ETA for the evaluation backend, if the outage is already known.
  2. An alternative submission route while the API is down — e.g., a temporary
    archive drop (presigned S3 link or similar) for server-side evaluation.
  3. Ideally, the official offline evaluation code (or a reference metrics
    implementation) for the 100 publicly released ground-truth samples, so
    local diagnostic scores can follow the official definitions in the meantime.

We are happy to provide full response headers, HAR files, or timing data if
useful. I searched existing issues and the HF Community discussions and found
no prior report of this outage — apologies in advance if this is a duplicate.

Environment

  • First observed: 2026-10-24 09:51 UTC; re-checked 2026-10-25 (still failing)
  • sam-3d-objects commit in use: f91db411c50efee93d8db7aeb323885650f6f722
    ("SA-3DAO Release - README Update (#190)", 2026-06-02)
  • Client: curl and browser fetch (both fail identically)
  • Team/contact: [[email protected]]
Dominant language
Python
Stars
7.6k
Forks
897
No merged PRs in 30d

Getting set up

First steps

  1. Read the whole issue, then the project's contributing guide.
  2. Comment on the issue to say you are picking it up — it saves two people doing the same work.
  3. Fork the repository and make your change on a branch.
  4. Open a pull request that references the issue number.

More from facebookresearch/sam-3d-objects

All issues in facebookresearch/sam-3d-objects

Similar issues

More Python issues

Get new issues in your inbox

A short digest of beginner-friendly GitHub issues.