feat(fine-tuning): add Prometheus monitoring and LoRA serving for fine-tuned models#369

Closed

zeyuyuyu wants to merge 16 commits intomainfrom

feat/fine-tuning-monitoring-and-serving

Commits on Feb 11, 2026

feat: storage-first flow - upload to 0G Storage + keep local backup

authored and
Wuying Created Local Users
committed
fix: handle raw JSONL files downloaded from 0G Storage

authored and
Wuying Created Local Users
committed
fix: handle raw JSONL from 0G Storage and auto-calculate fee

authored and
Wuying Created Local Users
committed
fix: address code review feedback for storage client

authored and
Wuying Created Local Users
committed
fix: upgrade 0g-storage-client to v1.2.2 and fix root hash encoding

authored and
Wuying Created Local Users
committed

Commits on Feb 24, 2026

feat(fine-tuning): add Prometheus monitoring and LoRA serving for fine-tuned models

authored and
Wuying Created Local Users
committed

Commits on Feb 25, 2026

Merge remote-tracking branch 'origin/main' into feat/fine-tuning-monitoring-and-serving

Wuying Created Local Users
authored and
Wuying Created Local Users
committed
fix: address code review feedback for monitoring and serving

authored and
Wuying Created Local Users
committed
refactor: address code review feedback — add godoc, DB limit, and cleanup duplicate Stop

authored and
Wuying Created Local Users
committed
refactor: address second round of review feedback

authored and
Wuying Created Local Users
committed

Commits on Feb 26, 2026

feat(fine-tuning): implement multi-tier LoRA caching with cold-storage offload/restore

Wuying Created Local Users
authored and
Wuying Created Local Users
committed
test(fine-tuning): add unit tests for serving module multi-tier LoRA caching

Wuying Created Local Users
authored and
Wuying Created Local Users
committed
docs(fine-tuning): add LoRA serving & multi-tier caching design document

Wuying Created Local Users
authored and
Wuying Created Local Users
committed

Commits on Feb 28, 2026

fix(fine-tuning): kill vLLM process group on Stop to prevent GPU memory leak

Wuying Created Local Users
authored and
Wuying Created Local Users
committed
feat(serving): add blocking wait mode for cold storage model loading

Wuying Created Local Users
authored and
Wuying Created Local Users
committed
feat(serving): add gpuMemoryUtilization config for vLLM

Wuying Created Local Users
authored and
Wuying Created Local Users
committed