Install any skill in seconds. Free to start, no credit card required.
Get Started Free →yangge521/model-quantization
模型量化加速方案,涵盖 FP16 半精度、INT8 量化、TensorRT/ONNX Runtime 部署优化
Passed 20 of 22 test cases with the skill, and 16 of 22 without it.
and 1 more in the full report
4 fixed16 passed either way2 still failing0 broke
Tokens, turns and time describe resource use, not a dollar price.
The instructions your model reads when the skill loads.
提供模型量化加速的完整方案,包括 FP16 半精度推理、INT8 量化、TensorRT 引擎转换、ONNX Runtime 部署优化。在保证精度的前提下最大化推理吞吐。
from ultralytics import YOLO
model = YOLO("yolo11n.pt")
results = model.predict(source="video.mp4", half=True, device=0)
FP16 将显存占用减半,推理速度提升 30-50%,精度损失通常小于 0.5%。NVIDIA T4 和 MTT S4000 均原生支持。
训练后量化(PTQ):使用校准数据集统计激活分布,将 FP32 权重和激活量化为 INT8。量化感知训练(QAT):在训练中模拟量化误差,精度更高但需重新训练。INT8 可将模型体积减小 4 倍,推理速度提升 2-3 倍。
from ultralytics import YOLO
model = YOLO("yolo11n.pt")
model.export(format="onnx", half=True, simplify=True)
model.export(format="engine", half=True, workspace=4)
test cases passed, vs 16 / 22 without
More skills from the same source and nearby in the registry, with their headline benchmark lift.