fix: 05_inference lit thresholds.yaml — streaming+conf_threshold #10

Open
poulpe wants to merge 0 commits from fix/05-inference-yaml-params into feature/auto-pipeline
Contributor

Signal

05_inference.py ignorait thresholds.yaml[inference] — mode windowed hardcode, sans --ply_conf_threshold.

Validation

GX049839_v2: streaming+conf=1.5+kf=1 -> 146M pts vs 0 pts (streaming+conf=2.5 defaut)

Patch

  • Ajoute _load_inference_cfg() qui lit config/thresholds.yaml
  • Mode/conf/kf/max_frames depuis config (streaming par defaut)
  • Ajoute --offload_to_cpu (stable RTX 3090 .84)

Auto-commit associe: keyframe_interval 6->1 sur feature/auto-pipeline (tag auto-iter-20260512-1637)

## Signal 05_inference.py ignorait thresholds.yaml[inference] — mode windowed hardcode, sans --ply_conf_threshold. ## Validation GX049839_v2: streaming+conf=1.5+kf=1 -> 146M pts vs 0 pts (streaming+conf=2.5 defaut) ## Patch - Ajoute _load_inference_cfg() qui lit config/thresholds.yaml - Mode/conf/kf/max_frames depuis config (streaming par defaut) - Ajoute --offload_to_cpu (stable RTX 3090 .84) Auto-commit associe: keyframe_interval 6->1 sur feature/auto-pipeline (tag auto-iter-20260512-1637)
poulpe added 1 commit 2026-05-12 16:39:09 +00:00
- Ajoute _load_inference_cfg() qui lit config/thresholds.yaml
- Mode/conf/keyframe_interval/max_frame_num depuis config (streaming par défaut)
- Valide par GX049839_v2: streaming+conf=1.5+kf=1 → 146M pts vs 0 pts en windowed sans conf_threshold
- Ajoute --offload_to_cpu (stable sur RTX 3090 .84)
This branch is already included in the target branch. There is nothing to merge.
View command line instructions

Checkout

From your project repository, check out a new branch and test the changes.
git fetch -u origin fix/05-inference-yaml-params:fix/05-inference-yaml-params
git checkout fix/05-inference-yaml-params
Sign in to join this conversation.
No Reviewers
1 Participants
Notifications
Due Date
No due date set.
Dependencies

No dependencies set.

Reference: floppyrj45/cosma-qc#10