[dynamo][fx] Don't emit call_function node to construct NamedTuple instances for Dynamo and make_fx tracing
#147145
Conversation
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
🔗 Helpful Links🧪 See artifacts and rendered test results at hud.pytorch.org/pr/147145
Note: Links to docs will display an error until the docs builds have been completed. ❌ 1 New Failure, 1 Unrelated FailureAs of commit 7be4215 with merge base f95bdf5 ( NEW FAILURE - The following job has failed:
BROKEN TRUNK - The following job failed but were present on the merge base:👉 Rebase onto the `viable/strict` branch to avoid these failures
This comment was automatically generated by Dr. CI and updates every 15 minutes. |
StrongerXi
added a commit
that referenced
this pull request
Feb 13, 2025
…` instances for Dynamo and `make_fx` tracing As title. This effectively undoes #49553, for Dynamo and `make_fx` tracing only (for `symbolic_trace` backward compatibility reasons). It heps enforce the invariant that Dynamo and `make_fx` graphs would always contain tensor ops -- so rather than having these `call_function` nodes to construct `NamedTuple`, we inline them directly as instance arguments to the user nodes. ghstack-source-id: d87a554 Pull Request resolved: #147145
StrongerXi
commented
Feb 13, 2025
torch/fx/proxy.py
Outdated
Comment on lines
145
to
151
| # Decides whether we represent `NamedTuple` instances into `call_function` | ||
| # nodes in the IR, or inline them directly as instance arguments to the user | ||
| # nodes. | ||
| _proxy_named_tuple: bool | ||
|
|
||
| def __init__(self): | ||
| self._proxy_named_tuple = True |
There was a problem hiding this comment.
@zou3519 not sure what's the best way to do it here, so this is just to get the proof-of-concept working.
StrongerXi
commented
Feb 13, 2025
| """ | ||
| Given the output arguments, generates the return statement of the FX function. | ||
| Note: The returned statement should not be indented. | ||
| """ | ||
| return f"return {repr(output_args)}" | ||
| return f"return {arg_repr(output_args)}" |
There was a problem hiding this comment.
All the arg_repr plumbing is to get this to apply.
This was referenced Feb 13, 2025
|
Abandoning this stack, see #147530. |
Sign up for free
to join this conversation on GitHub.
Already have an account?
Sign in to comment
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.
Stack from ghstack (oldest at bottom):
get_unique_name_wrthelper when applicable #146950nonstrict_trace#146367flat_apply#146714init=False#146713call_functionnode to construct dataclass instances for Dynamo andmake_fxtracing #147152call_functionnode to constructNamedTupleinstances for Dynamo andmake_fxtracing #147145As title. This effectively undoes #49553, for Dynamo and
make_fxtracing only (for
symbolic_tracebackward compatibility reasons).It heps enforce the invariant that Dynamo and
make_fxgraphs wouldalways contain tensor ops -- so rather than having these
call_functionnodes to construct
NamedTuple, we inline them directly as instancearguments to the user nodes.
cc @ezyang @SherlockNoMad @EikanWang @jgong5 @wenzhe-nrv @voznesenskym @penguinwu @Guobing-Chen @XiaobingSuper @zhuhaozhe @blzheng @jiayisunx @chenyang78 @kadeng @chauhang @amjames