Description
Bug Description
Elementwise operations like floor_divide.Scalar uses float as default scalar type, this may cause error in such a case:
The graph is partitioned into torch/tensorrt-interleaved subgraphs, and floor_divide.Scalar accepts one INT-type tensor and one INT-type scalar as input, and the output is fed into the subsequent subgraph as its INT-type input.
However, currently the floor_divide.Scalar converts the scalar into float by default, and then output a float-type tensor, which does not fullfill the requirement of subsequent graph.
To Reproduce
Traceback (most recent call last):
File "reproduce.py", line 34, in
trt_model(img_input)
File "/root/.pyenv/versions/3.7.13/lib/python3.7/site-packages/torch/nn/modules/module.py", line 1130, in _call_impl
return forward_call(*input, **kwargs)
RuntimeError: The following operation failed in the TorchScript interpreter.
Traceback of TorchScript (most recent call last):
RuntimeError: [Error thrown at core/runtime/execute_engine.cpp:91] Expected inputs[pyt_idx].dtype() == expected_type to be true but got false
Expected input tensors to have type Int, found type float
Expected behavior
Environment
Build information about Torch-TensorRT can be found by turning on debug messages
- Torch-TensorRT Version (e.g. 1.0.0):
- PyTorch Version (e.g. 1.0):
- CPU Architecture:
- OS (e.g., Linux):
- How you installed PyTorch (
conda,pip,libtorch, source): - Build command you used (if compiling from source):
- Are you using local sources or building from archives:
- Python version:
- CUDA version:
- GPU models and configuration:
- Any other relevant information: