PyTorch Developer Mailing List

Using Nsight Systems to profile GPU workload

hardware-backends NVIDIA CUDA

yasin August 22, 2024, 3:23pm 12

You can use --capture-range-end=stop instead.

Topic		Replies	Views	Activity
How profiling Pytorch Using Nsight Compute? NVIDIA CUDA	2	756	October 16, 2024
[RFC] Performance profiling at scale with detailed NVTX annotations compiler	0	535	July 10, 2024
Anomalous time gaps observed when using CUDA kernels in PyTorch	1	266	August 5, 2024
Profiling torch.compile compiler	0	253	April 10, 2025
Fast combined C++/Python/TorchScript/Inductor tracebacks performance	4	3193	August 17, 2023