lix19937
commented
4 days ago | title | 备注 |
|---|---|
| Driveworks5.0-camera_replay | |
| Driveworks5.0-cgf | |
| Driveworks5.0-make user demo | |
| Engine file by trtexec different from user's cpp project | |
| Floating point computing capacity not match with Orin-x's datasheet | |
| GPU Compute Time in trtexec's output | |
| How is the performance of einsum operator | |
| How to debug foreign node | |
| How to do concat op quantization in qat | |
| How to further optimize batched gridsample2d | |
| How to get FLOP | |
| How to improve ptq accuracy for model with plugin | |
| How to improve the point cloud registration accuracy of CUDA ICP | |
| How to obtain DLA resource usage information | |
| Issue of sample_cgf_dwchannel | |
| Multi streams sync(stream capture) in enqueue | |
| Myelin 700 error in trtexec inference | |
| Myelin foreign node take much time | |
| Only 11 arm cores and 4MB L3 cache observed in orin-Devkit | |
| Orin DevKit can not mount u disk | |
| Orin Devkit can not reboot | |
| Orin DevKit report Illegal instruction with gdb | |
| Orin DevKit trtexec build plan error | |
| Some problems about dynamic graph(torch) convert TRT plan | |
| Some problems of QAT sample | |
| Some tests about cuda memory selection | |
| Technical exchange of transformer-based algorithm porting with TensorRT | |
| Technical exchange on cudaGraph | |
| TensorRT Sparse Convolution Support | |
| The forward inference of dcnv2 implemented with CUDA takes too long | |
| Time consumption of QAT | |
| TRT dynamic batch h w infer | |
| trtexec convert onnx error | |
| trtexec int8 time no less than fp32 | |
| Using multiple streams does not achieve concurrent execution | |
| Why after reformat, data is not change | |
| Why does the same fusion method take nearly twice as long between qat and ptq | |
| Why QAT2PTQ tool not work |