Zero Zeng
Zero Zeng
@ttyio Do you know? thanks!
Still fail, filed internal bug 4356383 for this.
As the warning say: some layers is force to running in INT32, you can not set those layers to FP32.
To avoid the warning, just don't set those layers precision to FP32, or just safely ignore it.
Filter with layer name or layer type?
Could you please try latest 9.2, IIRC we add support to opset 17 since TRT 8.6. Download from https://developer.nvidia.com/downloads/compute/machine-learning/tensorrt/9.2.0/tensorrt-9.2.0.5.linux.x86_64-gnu.cuda-11.8.tar.gz https://developer.nvidia.com/downloads/compute/machine-learning/tensorrt/9.2.0/tensorrt-9.2.0.5.linux.x86_64-gnu.cuda-12.2.tar.gz https://developer.nvidia.com/downloads/compute/machine-learning/tensorrt/9.2.0/tensorrt-9.2.0.5.ubuntu-22.04.aarch64-gnu.cuda-12.2.tar.gz
Looks like a usage issue to me, I would suggest using Polygraphy so that you don't need to implement the calibrator. see https://github.com/NVIDIA/TensorRT/tree/main/tools/Polygraphy/examples/cli/convert/01_int8_calibration_in_tensorrt A sample usage: ``` polygraphy convert model.onnx...
the dla standalone loadable can only be loaded with cuDLA, TRT cannot load it.
This looks like a bug, could you please provide a reproduce? Thanks!