README.md

Friday 10th Nov 2023

09:00 - 10:30 > Hacking: Evaluation of AI models with benchmarks

• Test segmentation models and convert checkpoint.pth to colon.onnx > colon_seg_net_test.ipynb

10:45 - 12:00 > Hacking: Optmise, test and deploy AI models in clara-agx

How can choosing the right floating point precision improve performance with my application?

• See workflow to convert your images to video and gxf format
• See Inference Parameters (backend and enable_fp16=True/False) in colonoscopy_segmentation.py and further descriptions in this README

Models can also be optimised using tao-coverter

./tao-converter \
    -k tlt_encode \
    -d 3,544,960 \
    -w 40960M \
    -t fp16 \
    -o output_cov/Sigmoid,output_bbox/BiasAdd \
    -e engine.trt \
    resnet34_peoplenet_pruned_int8.etlt

• Parameters in tao-converter

-k: The key used to encode the .tlt model when doing the training. 
-d: Comma-separated list of input dimensions that should match the dimensions used for tao <model> export.
-w: Maximum workspace size for the TensorRT engine. The default value is 1073741824(1<<30).
-t: Desired engine data type, generates calibration cache if in INT8 mode. The default value is fp32. The options are {fp32, fp16, int8}.
-o: Comma-separated list of output blob names that should match the output configuration used for tao <model> export.
-e: Path to save the engine to. (default: ./saved.engine)
input_file

13:00 - 14:30 > Hacking: Clara AGX/IGX for intelligent Medical Instruments by Mikael Brudfors

• Mikael Brudfors introducing Clara AGX/IGX for intelligent Medical Instruments
• Q&As
• Keep hacking

14:45 - 16:30 > Tidied up documentation and present results

• Prepare documentation and discuss results
• Create PRs
• Review and merge PRS
• Group photo

References

https://docs.nvidia.com/deeplearning/tensorrt/quick-start-guide/index.html#abstract