v12 YOLO26s (207 labels): weights.onnx

use onnxruntime-gpu + cuda128 torch

v10: chute_config.yml (fine-tuned from v9c, 163 labels)

v10: class_names.txt (fine-tuned from v9c, 163 labels)

v10: miner.py (fine-tuned from v9c, 163 labels)

v10: weights.onnx (fine-tuned from v9c, 163 labels)

upgrade chute_config to onnxruntime-gpu + cost exclusions

v8 yolo26s pseudo-gt (manifest order, identity cls_remap, conf=0.35)

initial commit