Add TensorRT export and inference notebooks

[zhiqwang]

Here is a small demo to surgeon the YOLOv5 ONNX model and then export to TensorRT engine.

Super parameters

model_path = "https://github.com/ultralytics/yolov5/releases/download/v6.0/yolov5n6.pt"
checkpoint_path = attempt_download(model_path)
onnx_path = "yolov5n6.onnx"
engine_path = "yolov5n6.engine"

score_thresh = 0.35
iou_thresh = 0.45
detections_per_img = 100

Surgeon the yolov5 ONNX models

from yolort.runtime.yolo_graphsurgeon import YOLOGraphSurgeon

yolo_gs = YOLOGraphSurgeon(
    model_path,
    version="r6.0",
    enable_dynamic=False,
)

yolo_gs.register_nms(
    score_thresh=score_thresh,
    nms_thresh=iou_thresh,
    detections_per_img=detections_per_img,
)

# Export the ONNX model
yolo_gs.save(onnx_path)

Build the TensorRT engine

from yolort.runtime.trt_helper import EngineBuilder

engine_builder = EngineBuilder()
engine_builder.create_network(onnx_path)
engine_builder.create_engine(engine_path, precision="fp32")

[codecov]

Codecov Report

Merging #253 (e8af511) into main (1d86967) will not change coverage.
The diff coverage is n/a.

@@           Coverage Diff           @@
##             main     #253   +/-   ##
=======================================
  Coverage   93.62%   93.62%           
=======================================
  Files          11       11           
  Lines         675      675           
=======================================
  Hits          632      632           
  Misses         43       43           

Flag	Coverage Δ
unittests	93.62% <ø> (ø)

Flags with carried forward coverage won't be shown. Click here to find out more.

---

Continue to review full report at Codecov.

Legend - Click here to learn more
Δ = absolute <relative> (impact), ø = not affected, ? = missing data
Powered by Codecov. Last update 1d86967...e8af511. Read the comment docs.