[GH-ISSUE #516] Does deepep-tests scripts support nsys profile? #333

New issue

Closed

opened 2026-10-26 09:58:43 +00:00 by bsakaguchi · 1 comment

bsakaguchi commented

2026-10-26 09:58:43 +00:00

Owner

Originally created by @fishautumn on GitHub (Dec 10, 2025).
Original GitHub issue: https://github.com/deepseek-ai/DeepEP/issues/516

it uses kineto to capture kernel durations. does it still work along with nsys profile -t cuda?

if it doesn't, can we add a command line option or some env variable to by-pass kineto data capturing and calculation, so nsys profile can work?

Originally created by @fishautumn on GitHub (Dec 10, 2025). Original GitHub issue: https://github.com/deepseek-ai/DeepEP/issues/516 it uses kineto to capture kernel durations. does it still work along with `nsys profile -t cuda`? if it doesn't, can we add a command line option or some env variable to by-pass kineto data capturing and calculation, so `nsys profile` can work?

bsakaguchi closed this issue

2026-10-26 09:58:44 +00:00

bsakaguchi commented

2026-10-26 09:58:46 +00:00

Author

Owner

@polarstormx commented on GitHub (Sep 18, 2026):

Current main already provides EP_USE_NVIDIA_TOOLS=1 to bypass the internal Kineto profiling, as documented in the README.

@polarstormx commented on GitHub (Sep 18, 2026): Current main already provides `EP_USE_NVIDIA_TOOLS=1` to bypass the internal Kineto profiling, as documented in the README.

bsakaguchi referenced this issue

2026-10-27 21:03:38 +00:00

[PR #333] [CLOSED] error #523