Model Zoo

End-to-end compilation examples for popular models across vision, language, and generative AI. Each example shows how to compile the model with Torch-TensorRT and benchmark against eager PyTorch.

Vision

• Example: Compiling ResNet with dynamic shapes
• Example: Compiling BERT with torch.compile
• Example: Engine Caching (BERT)

Generative AI

• Example: Compiling GPT2 with torch.compile
• Example: Compiling SAM2 with the dynamo backend
• Example: TensorRT Plugin for RMSNorm in Llama2

For diffusion models see HuggingFace Models.

Notebooks

• Legacy notebooks