Hi! Could you please explain how to profile flops of converted tensorrt model?
veraj
2
Hi, @dara.vinogradova
Thanks for using Nsight Compute! You can check smsp__sass_thread_inst_executed_op* related metrics. And use --metrics option to profile.
Related topics
| Topic | Replies | Views | Activity | |
|---|---|---|---|---|
| Tensor Core Flops | 2 | 102 | December 24, 2025 | |
| Confusion about the (d/f/h)(mul/add/fma) count in the nsight compute | 6 | 1746 | January 16, 2024 | |
| Profile pytorch model using NCU | 1 | 1312 | July 1, 2022 | |
| Nsight Compute to measure metrics data | 1 | 589 | January 29, 2021 | |
| How to profile Flops when running inferences with TensorRT | 0 | 1271 | October 15, 2018 | |
| Metrics smsp__sass_thread_inst_executed_op* returns n/a | 8 | 2001 | August 2, 2019 | |
| How many FLOPs does one tensor_op_hmma instruction do? | 4 | 1049 | May 30, 2024 | |
| How to calculate TOPS (INT8) or TFLOPS (FP16) of each layer of a CNN using TensorRT | 7 | 12852 | September 12, 2021 | |
| How to see overall Flop(Floating point operations), not per second, using nsight compute? | 3 | 713 | July 18, 2023 | |
| NSIGHT Systems and NSIGHT Compute | 7 | 421 | March 20, 2025 |