Skip to content

Batch of updates - Nvidia GPU support & more

Valentina Kustikova requested to merge github/fork/maslovaz/develop into master

Created by: maslovaz

  • onnxruntime cpp - added cuda and tensorrt (FP16&FP32) support
  • onnxruntime python - added, with CPU, cuda and tensorrt (FP16&FP32) support
  • pytorch - fix cuda support, added tensorrt (FP16&FP32) support, change preprocessing to have ability to run language models and models with more than one input
  • tensorflow&tensorflowlite - add saved model support, move to tf2 API usage
  • added converter to TF2 saved model format and optimization with TRT for NVIDIA devices
  • fix preprocessing to cyclically feed inference script with limited data

Merge request reports