{"commit":{"body":[{"label":"Why","text":"Pose에도 검증된 FP16 입력과 CUDA sharing 경로가 필요했다."},{"label":"What","text":"모델별 dtype/sharing 선택을 공통화하고 Pose template과 Fire 보존 테스트를 연결했다."},{"label":"Impact","text":"Pose는 새 version4 engine과 함께 적용하며 Medium과 Fire 기존 계약은 보존한다."},{"label":"Verify","text":"host/image 각4tests 및 batch1~4 실행 확인과 단계별16채널 약10FPS 수용. 품질은 미평가다."},{"label":"Rollback","text":"대상 Stop 후 이전 image와 Pose mount 및 설정을 함께 복원한다. 이번에는 자동복원하지 않았다."}],"title":"feat(pose): FP16 입력 연결과 Tracking 16채널 수용"},"created_at":"2026-09-16T13:38:31+00:00","expires_at":"2026-09-17T13:38:31+00:00","job_id":"20260916-pose-fp16","paths":[{"change":"modified","content_sha256":"6ee2bbcaf40984c51e7c8795632bbf54d8782fbaab37987dcc54a108ef178202","path":"app/cfg_templates/nvinferserver_grpc_pose.tmpl"},{"change":"modified","content_sha256":"e26135df0bc9710d5b56a280e237c521ef43a578878502425adf25b4f1ca78c3","path":"app/inference_profiles.py"},{"change":"modified","content_sha256":"bbc8e9c31ea466935642be1375f338406e4e0036e7f935df63f33127d0ba7812","path":"app/medium_preprocess.py"},{"change":"modified","content_sha256":"5fd495517207c6b66abc7f2c14227bc83b104d2e12b532b8939d7bd69be24295","path":"tests/test_medium_preprocess.py"},{"change":"modified","content_sha256":"031b4c12cd00e2c7a48f397c55e5faf49ad9784c06e91b8906fa8e4b76cbe6d6","path":"docs/profiles/ltb1021/topics/runtime-lifecycle/README.ko.md"},{"change":"added","content_sha256":"b5f6e9cbe9d0b27b5aef32462fe084f72dfdeffc783e935a29e5073a4d9f4b40","path":"docs/profiles/ltb1021/topics/runtime-lifecycle/runs/20260916T223600+0900-run-pose-fp16-tracking.md"}],"plan_hash":"f7561899c4022979fd038986ed5ce5539da6f69a1619e8ffd9384c4b1ddeddf8","profile":"ltb1021","push":{"branch":"task/runtime-lifecycle-refactor","mode":"upstream","remote":"origin","remote_identity":"sha256:3a9ce41319a2ebea","upstream_head":"e0c4e77f002eb37c0f7c7243276c0a6a5c42d748","upstream_ref":"origin/task/runtime-lifecycle-refactor"},"repository":{"baseline_head":"e0c4e77f002eb37c0f7c7243276c0a6a5c42d748","branch":"task/runtime-lifecycle-refactor","worktree":"/workspace/contextra-analytics-runtime-lifecycle-refactor"},"schema_version":"ltb1021.codex_hub.git_closeout_manifest.v1","session_alias":"work-session","verify":["host/image render 4 tests PASS","batch1-4 shape dtype finite PASS","4/8/12/16 ascent PASS; stale/drop/errors0; replaced6","13 children reaped, original3 Pose running, detail Off","git diff --check PASS"]}
