TensorRT
TensorRT copied to clipboard
PyTorch/TorchScript/FX compiler for NVIDIA GPUs using TensorRT
select.val and select.src converter
Current `index` converter only supports `int` type while cannot handle `bool` mask indices. We need to expand index support to boolean masks, since this is a valid usage in Torch...
See the failures in the `test-py-dynamo-converters-cu121` [section here](https://github.com/pytorch/TensorRT/actions/runs/7350298746/job/20011958833?pr=2538). The bitwise converters are off by 1 in certain test cases, pointing to either implementation/functionality changes in TRT 9.x, or correctness/accuracy issues.
See the failures in the `test-py-dynamo-converters-cu121` [section here](https://github.com/pytorch/TensorRT/actions/runs/7350298746/job/20011958833?pr=2538). Some elementwise converters have mismatching data types.
**Is your feature request related to a problem? Please describe.** Currently if a network takes an input tensor which is used as shape for other layers in the graph, a...
## Bug Description 1) The graph has a conv node in pytorch and a TensorRT node. The conv node has weight and bias lifted as placeholders. Hence we are seeing...