NVIDIA / NVIDIA/recsys-examples

[QST] nsys profile fails to generate profiling file on L20 with version 2603

Open
#397 1 comment 0 reactions 1 assignee View on GitHub

@JacoCheung is already working on this.

Since May 19, 2026.

question
Dominant language
Python
Stars
322
Forks
83
Avg merge
3d 3h
Merged PRs (30d)
8

Description

Environment

Machine: L20
Version: 2603 (latest)
CUDA Driver: 590.44.01
Docker Image: nvcr.io/nvidia/pytorch:26.02-py3 (as specified in Dockerfile),cuda 13.1

Steps to Reproduce

  1. run script:
    ./training/benchmark/scripts/run_all_experiments_local.sh \ --exp-file=training/benchmark/experiments.txt \ --nproc=2 --nsys
  2. When using nsys profile with -c cudaProfilerApi, no profiling file is generated. logged as:
    Processing events... Generated: No reports were generated.
  3. Expected output is A valid profiling report file (e.g., .nsys-rep or .sqlite) should be generated.
  4. The complete log is as follows:nsys.failed.logs.txt
  5. when nsys without -c cudaProfilerApi, through nsys-rep file created,but we see only the CPU timeline, without GPU timeline, while NSys GUI say error : Subscribe returned 999: CUPTI_ERROR_UNKNOWN
Image

Already Tried

  1. docker run with --privileged and --cap-add=SYS_ADMIN
  2. stop the dcgm server on the host
  3. downgrade or upgrade nsys version

Help Me ! Thanks For Your Answer

Contributor guide

Open the contributing guide

First steps

  1. Read the whole issue, then the project's contributing guide.
  2. Comment on the issue to say you are picking it up — it saves two people doing the same work.
  3. Fork the repository and make your change on a branch.
  4. Open a pull request that references the issue number.

Assessment

This issue has not been assessed yet.

Get new issues in your inbox

A short digest of beginner-friendly GitHub issues.