Add addTypedMetadata to GenericTraceActivity - #1485
Closed
ryanzhang22 wants to merge 1 commit into
Closed
Conversation
Summary: One of the nice things about the old string-only JSON approach was that you could super easily add a new metadata field on the PyTorch side without doing additional work to reflect the change in Kineto. Since `GenericTraceActivity` ultimately stores a string key and `TypedValue`, we should support adding typed metadata dynamically by key. Note -- the catalog is still valuable and should remain the contract for fields Kineto explicitly reads or interprets. For example, NCCL metadata fields that are written on the PyTorch side and explicitly fetched via `getMetadataValue` in `output_json.cpp`. Right now we rely on strings happening to match between PyTorch and Kineto; using the catalog is a more official way of matching the two. For `GenericTraceActivity` metadata that is only access through a visitor class though (which NCCL metadata should eventually migrate to), it's not strictly necessary to go through the catalog. Differential Revision: D112555947
|
@ryanzhang22 has exported this pull request. If you are a Meta employee, you can view the originating Diff in D112555947. |
|
This pull request has been merged in 29ab453. |
pytorchmergebot
pushed a commit
to pytorch/pytorch
that referenced
this pull request
Jul 23, 2026
Includes the following commits: - Add XPU toggleCollectionDynamic mechanism (pytorch/kineto#1410) cb29370 - Explicitly wrap NCCL strings in output json (pytorch/kineto#1482) d62c7f5 - emit GPU_USER_ANNOTATION for user record_function ranges (pytorch/kineto#1393) a8fd013 - Re-select test_basic_chrome_trace_cpu on XPU (pytorch/kineto#1484) f451dfc - Fix CQS signal readability-avoid-const-params-in-decls in fbcode/kineto/libkineto (pytorch/kineto#1486) c4e1538 - Add NCCL collective metadata fields to the catalog (pytorch/kineto#1483) a31870c - Add addTypedMetadata to GenericTraceActivity (pytorch/kineto#1485) 29ab453 - Centralize MetadataField definitions (pytorch/kineto#1481) 9c49dca - Add XCCL collective communication activity tracing to XPU plugin (pytorch/kineto#1396) a97543e - Fix CQS signal modernize-use-scoped-lock in fbcode/kineto/libkineto (pytorch/kineto#1480) e804b8f - Un-deselect some profiler tests after fixing them (pytorch/kineto#1432) 446ac67 - Test controller async request arbitration (pytorch/kineto#1477) d579d88 - Annotation fixups (pytorch/kineto#1476) 994e6c5 Pull Request resolved: #190687 Approved by: https://github.com/Skylion007, https://github.com/scotts
youlituao
pushed a commit
to youlituao/pytorch
that referenced
this pull request
Jul 28, 2026
Includes the following commits: - Add XPU toggleCollectionDynamic mechanism (pytorch/kineto#1410) cb29370 - Explicitly wrap NCCL strings in output json (pytorch/kineto#1482) d62c7f5 - emit GPU_USER_ANNOTATION for user record_function ranges (pytorch/kineto#1393) a8fd013 - Re-select test_basic_chrome_trace_cpu on XPU (pytorch/kineto#1484) f451dfc - Fix CQS signal readability-avoid-const-params-in-decls in fbcode/kineto/libkineto (pytorch/kineto#1486) c4e1538 - Add NCCL collective metadata fields to the catalog (pytorch/kineto#1483) a31870c - Add addTypedMetadata to GenericTraceActivity (pytorch/kineto#1485) 29ab453 - Centralize MetadataField definitions (pytorch/kineto#1481) 9c49dca - Add XCCL collective communication activity tracing to XPU plugin (pytorch/kineto#1396) a97543e - Fix CQS signal modernize-use-scoped-lock in fbcode/kineto/libkineto (pytorch/kineto#1480) e804b8f - Un-deselect some profiler tests after fixing them (pytorch/kineto#1432) 446ac67 - Test controller async request arbitration (pytorch/kineto#1477) d579d88 - Annotation fixups (pytorch/kineto#1476) 994e6c5 Pull Request resolved: pytorch#190687 Approved by: https://github.com/Skylion007, https://github.com/scotts
youlituao
pushed a commit
to youlituao/pytorch
that referenced
this pull request
Jul 28, 2026
Includes the following commits: - Add XPU toggleCollectionDynamic mechanism (pytorch/kineto#1410) cb29370 - Explicitly wrap NCCL strings in output json (pytorch/kineto#1482) d62c7f5 - emit GPU_USER_ANNOTATION for user record_function ranges (pytorch/kineto#1393) a8fd013 - Re-select test_basic_chrome_trace_cpu on XPU (pytorch/kineto#1484) f451dfc - Fix CQS signal readability-avoid-const-params-in-decls in fbcode/kineto/libkineto (pytorch/kineto#1486) c4e1538 - Add NCCL collective metadata fields to the catalog (pytorch/kineto#1483) a31870c - Add addTypedMetadata to GenericTraceActivity (pytorch/kineto#1485) 29ab453 - Centralize MetadataField definitions (pytorch/kineto#1481) 9c49dca - Add XCCL collective communication activity tracing to XPU plugin (pytorch/kineto#1396) a97543e - Fix CQS signal modernize-use-scoped-lock in fbcode/kineto/libkineto (pytorch/kineto#1480) e804b8f - Un-deselect some profiler tests after fixing them (pytorch/kineto#1432) 446ac67 - Test controller async request arbitration (pytorch/kineto#1477) d579d88 - Annotation fixups (pytorch/kineto#1476) 994e6c5 Pull Request resolved: pytorch#190687 Approved by: https://github.com/Skylion007, https://github.com/scotts
aws-kingrj
pushed a commit
to amazon-contributing/upstream-to-pytorch
that referenced
this pull request
Jul 29, 2026
Includes the following commits: - Add XPU toggleCollectionDynamic mechanism (pytorch/kineto#1410) cb29370 - Explicitly wrap NCCL strings in output json (pytorch/kineto#1482) d62c7f5 - emit GPU_USER_ANNOTATION for user record_function ranges (pytorch/kineto#1393) a8fd013 - Re-select test_basic_chrome_trace_cpu on XPU (pytorch/kineto#1484) f451dfc - Fix CQS signal readability-avoid-const-params-in-decls in fbcode/kineto/libkineto (pytorch/kineto#1486) c4e1538 - Add NCCL collective metadata fields to the catalog (pytorch/kineto#1483) a31870c - Add addTypedMetadata to GenericTraceActivity (pytorch/kineto#1485) 29ab453 - Centralize MetadataField definitions (pytorch/kineto#1481) 9c49dca - Add XCCL collective communication activity tracing to XPU plugin (pytorch/kineto#1396) a97543e - Fix CQS signal modernize-use-scoped-lock in fbcode/kineto/libkineto (pytorch/kineto#1480) e804b8f - Un-deselect some profiler tests after fixing them (pytorch/kineto#1432) 446ac67 - Test controller async request arbitration (pytorch/kineto#1477) d579d88 - Annotation fixups (pytorch/kineto#1476) 994e6c5 Pull Request resolved: pytorch#190687 Approved by: https://github.com/Skylion007, https://github.com/scotts
MelindaTan
pushed a commit
to MelindaTan/pytorch
that referenced
this pull request
Jul 31, 2026
Includes the following commits: - Add XPU toggleCollectionDynamic mechanism (pytorch/kineto#1410) cb29370 - Explicitly wrap NCCL strings in output json (pytorch/kineto#1482) d62c7f5 - emit GPU_USER_ANNOTATION for user record_function ranges (pytorch/kineto#1393) a8fd013 - Re-select test_basic_chrome_trace_cpu on XPU (pytorch/kineto#1484) f451dfc - Fix CQS signal readability-avoid-const-params-in-decls in fbcode/kineto/libkineto (pytorch/kineto#1486) c4e1538 - Add NCCL collective metadata fields to the catalog (pytorch/kineto#1483) a31870c - Add addTypedMetadata to GenericTraceActivity (pytorch/kineto#1485) 29ab453 - Centralize MetadataField definitions (pytorch/kineto#1481) 9c49dca - Add XCCL collective communication activity tracing to XPU plugin (pytorch/kineto#1396) a97543e - Fix CQS signal modernize-use-scoped-lock in fbcode/kineto/libkineto (pytorch/kineto#1480) e804b8f - Un-deselect some profiler tests after fixing them (pytorch/kineto#1432) 446ac67 - Test controller async request arbitration (pytorch/kineto#1477) d579d88 - Annotation fixups (pytorch/kineto#1476) 994e6c5 Pull Request resolved: pytorch#190687 Approved by: https://github.com/Skylion007, https://github.com/scotts
This file contains hidden or bidirectional Unicode text that may be interpreted or compiled differently than what appears below. To review, open the file in an editor that reveals hidden Unicode characters.
Learn more about bidirectional Unicode characters
Sign up for free
to join this conversation on GitHub.
Already have an account?
Sign in to comment
1 participant
Add this suggestion to a batch that can be applied as a single commit.This suggestion is invalid because no changes were made to the code.Suggestions cannot be applied while the pull request is closed.Suggestions cannot be applied while viewing a subset of changes.Only one suggestion per line can be applied in a batch.Add this suggestion to a batch that can be applied as a single commit.Applying suggestions on deleted lines is not supported.You must change the existing code in this line in order to create a valid suggestion.Outdated suggestions cannot be applied.This suggestion has been applied or marked resolved.Suggestions cannot be applied from pending reviews.Suggestions cannot be applied on multi-line comments.Suggestions cannot be applied while the pull request is queued to merge.Suggestion cannot be applied right now. Please check back later.
Summary:
One of the nice things about the old string-only JSON approach was that you could super easily add a new metadata field on the PyTorch side without doing additional work to reflect the change in Kineto. Since
GenericTraceActivityultimately stores a string key andTypedValue, we should support adding typed metadata dynamically by key.Note -- the catalog is still valuable and should remain the contract for fields Kineto explicitly reads or interprets. For example, NCCL metadata fields that are written on the PyTorch side and explicitly fetched via
getMetadataValueinoutput_json.cpp. Right now we rely on strings happening to match between PyTorch and Kineto; using the catalog is a more official way of matching the two. ForGenericTraceActivitymetadata that is only access through a visitor class though (which NCCL metadata should eventually migrate to), it's not strictly necessary to go through the catalog.Differential Revision: D112555947