Skip to content

feat(profiler): support Ascend NPU profiling via torch_npu.profiler - #1096

Merged
DefTruth merged 4 commits into
mainfrom
dev
Aug 14, 2026
Merged

feat(profiler): support Ascend NPU profiling via torch_npu.profiler#1096
DefTruth merged 4 commits into
mainfrom
dev

Conversation

@DefTruth

Copy link
Copy Markdown
Member

No description provided.

brandneway and others added 4 commits July 26, 2026 09:44
…1093)

* feat(profiler): support Ascend NPU profiling via torch_npu.profiler

* feat(profiler): export full CANN output directory on NPU

Switch the NPU path from export_chrome_trace (single json) to
on_trace_ready=tensorboard_trace_handler + experimental_config, so the
full Ascend profiling output directory is produced: ASCEND_PROFILER_OUTPUT
(kernel_details.csv, step_trace_time.csv, trace_view.json, op_summary.csv,
op_statistic.csv, operator_details.csv, .db) plus PROF_XXX, FRAMEWORK and
logs. data_simplification=False keeps every generated file.

- profiler_level configurable via CACHE_DIT_NPU_PROFILER_LEVEL (Level0 default)
- sanitize worker_name to the allowed [A-Za-z0-9_-] charset
- locate the produced *_ascend_pt dir and expose it via trace_path
- CUDA path unchanged

---------

Co-authored-by: changetheway <guotaoyuan1@h-partners.com>
@DefTruth
DefTruth merged commit b904e00 into main Aug 14, 2026
4 checks passed
@DefTruth
DefTruth deleted the dev branch August 14, 2026 10:38
Sign up for free to join this conversation on GitHub. Already have an account? Sign in to comment

Labels

None yet

Projects

None yet

Development

Successfully merging this pull request may close these issues.

2 participants