NVIDIA announces TensorRT 8.2 and Integrations with PyTorch and TensorFlow on Dec 02, 2021.
This Torch-TensorRT is a high-performance deep learning inference optimizer and runtime that delivers low latency, high-throughput inference for AI applications. TensorRT is used across several industries including healthcare, automotive, manufacturing, internet/telecom services, financial services, and energy.
I tested today using my gavatar image on colab notebook with the GPU device.
Am prelucrat un cod sursa exemplu existent de pe internet cu o un model RESNET known as Deep Residual Learning for Image Recognition, see this website.
model = models.resnet50(pretrained=True).to("cuda")I have a pretty good picture of the processing possibilities given for this topic and I can tell you today that this implementation of TensorRT is below my expectations.
However, there are some positive elements that can be used with this in the future.
The full exaemple and how can be used TensorRT with colab tool can be found on my GitHub repo with all colabs notebooks.