Spaces:
Running on Zero
Running on Zero
Commit History
22 days ago
-
Collect and freeze parent garbage before each worker fork (worker GC segfault); drop spacing Running
Verified signature513c139
23 days ago
-
Space consecutive GPU jobs by 50 s (back-to-back worker crash), bounded by caller identity
Verified signature872353a -
Fatal-signal stack dumps in workers; suppress core dumps (crash was costing ~300 s GPU)
Verified signaturebb0e4b3 -
Parent-side worker probe (/proc state, SIGUSR1 stack dump); watchdog covers stage prelude
Verified signature3b5f15c -
Bounded Hub sessions per stage download, one transport retry, stall diagnostics
Verified signaturef23d19d -
Fresh ZeroGPU worker per GPU job (reused worker died on a second staged run)
Verified signature2b5f99e -
Reference to video (Ref2VA), custom workflow nodes and chain templates
Verified signature95e8794 -
Mount Gradio OAuth routes for canvas sign-in
Verified signature928d017 -
HF OAuth sign-in; reject expired ZeroGPU identity before scheduling; no queue waits behind GPU jobs
Verified signature2e62205 -
Gradio 6.28 full-page canvas with H3 template gallery; calibrated GPU budgets
Verified signature153f177 -
Fix canvas result shape, surface bridge errors, clarify busy message
Verified signature3073d59 -
Add native WorkflowCanvas UI with named H3 templates
Verified signaturee0cd66a -
Use one allocation for sequential local Workflow Studio stages
Verified signature7d4352c -
Run conditioning and generation locally in one Workflow Studio request
Verified signature02a5e84 -
Add experimental shared H3 preset runner and automation client
Verified signature22f9ecd