Troubleshooting for users
Record the target site, personal/organization identity, resource ID, time and error. Contact your deployment administrator or community support. Remove credentials from logs.
| Symptom | Check in order | Reference |
|---|---|---|
| Authentication fails | Site and address; Access Token versus API Key; expiry, refresh or revocation; request header | Access Token, API Keys |
| Repository missing or inaccessible | Namespace, type and ID; organization transfer; membership and effective permissions | Organizations |
| Deployment or training button unavailable | Complete model files and metadata; configured framework/image; administrator scan | Endpoint FAQ, Training FAQ |
| Upload fails | Write permissions, file size, service error and configured limits; preserve resume progress | Model upload, Dataset upload |
| Only LFS pointer files downloaded | Install Git LFS and run git lfs pull in the repository; check credentials, network and disk | Model download |
| Instance stays queued or fails to start | Logs; available resources/quotas; image/model downloads; storage and scheduling | Endpoint usage |
| Out of memory or model load failure | Precision/quantization, framework version, context length, concurrency and complete weights | Inference frameworks |
| API timeout or rate limit | URL, model, Key, permissions, service status and quota; preserve errors and avoid unbounded retries | API Keys |
| Instance runs after training | Export results, stop the instance and verify status and usage | Instance lifecycle |
When asking for help, provide software/client versions, minimal reproduction steps, expected and actual behavior, timestamp with timezone, resource ID and sanitized errors. Include a request ID when available, but never include a real Token or password.