Add FloatTrueDiv and ToFloat to SYMPY_INTERP #128418
Closed
+2
−0
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
Summary: I admit I'm not 100% sure what I'm doing here. I'm hitting a bug in the FX graph cache when we try to evaluate a guards expression. We're creating guards that look like this: ``` Ne(CeilToInt(FloatTrueDiv(ToFloat(8*L['t0']) - 4.0, 8.0))*CeilToInt(FloatTrueDiv(ToFloat(8*L['t1']) - 4.0, 8.0)), CeilToInt(FloatTrueDiv(ToFloat(8*L['t1']) - 4.0, 8.0))) and ... ``` It looks like we have a facility to define these operators in the SYMPY_INTERP map and we're just missing FloatTrueDiv and ToFloat. What's surprsing to me is that we're only hitting this problem with the FX graph enabled. We can create such guards, but we've never actually evaluated any? Test Plan: `TORCHINDUCTOR_FX_GRAPH_CACHE=1 python benchmarks/dynamo/torchbench.py --ci --accuracy --timing --explain --inductor --device cuda --inference --bfloat16 --only detectron2_fcos_r_50_fpn` [ghstack-poisoned]
🔗 Helpful Links🧪 See artifacts and rendered test results at hud.pytorch.org/pr/128418
Note: Links to docs will display an error until the docs builds have been completed. ✅ No FailuresAs of commit e57be5c with merge base f2d7f23 ( This comment was automatically generated by Dr. CI and updates every 15 minutes. |
masnesral
added a commit
that referenced
this pull request
Jun 11, 2024
Summary: I admit I'm not 100% sure what I'm doing here. I'm hitting a bug in the FX graph cache when we try to evaluate a guards expression. We're creating guards that look like this: ``` Ne(CeilToInt(FloatTrueDiv(ToFloat(8*L['t0']) - 4.0, 8.0))*CeilToInt(FloatTrueDiv(ToFloat(8*L['t1']) - 4.0, 8.0)), CeilToInt(FloatTrueDiv(ToFloat(8*L['t1']) - 4.0, 8.0))) and ... ``` It looks like we have a facility to define these operators in the SYMPY_INTERP map and we're just missing FloatTrueDiv and ToFloat. What's surprsing to me is that we're only hitting this problem with the FX graph enabled. We can create such guards, but we've never actually evaluated any? Test Plan: `TORCHINDUCTOR_FX_GRAPH_CACHE=1 python benchmarks/dynamo/torchbench.py --ci --accuracy --timing --explain --inductor --device cuda --inference --bfloat16 --only detectron2_fcos_r_50_fpn` ghstack-source-id: 2da68de Pull Request resolved: #128418
|
fyi @oulgen |
eellison
reviewed
Jun 11, 2024
|
@eellison do you mean this: https://github.com/pytorch/pytorch/blob/main/torch/utils/_sympy/functions.py#L12-L28 |
ezyang
approved these changes
Jun 11, 2024
|
No test? |
|
@pytorchbot merge |
Merge startedYour change will be merged once all checks pass (ETA 0-4 Hours). Learn more about merging in the wiki. Questions? Feedback? Please reach out to the PyTorch DevX Team |
masnesral
added a commit
that referenced
this pull request
Jun 12, 2024
…in the codecache Summary: Add a few unit tests that exercise `produce_guards_expression` and `evaluate_guards_expression` (and specifically "ToFloat" "FloatTrueDiv" added in #128418) [ghstack-poisoned]
masnesral
added a commit
that referenced
this pull request
Jun 12, 2024
…in the codecache" Summary: Add a few unit tests that exercise `produce_guards_expression` and `evaluate_guards_expression` (and specifically "ToFloat" "FloatTrueDiv" added in #128418) cc voznesenskym penguinwu EikanWang jgong5 Guobing-Chen XiaobingSuper zhuhaozhe blzheng wenzhe-nrv jiayisunx peterbell10 ipiszy yf225 chenyang78 kadeng muchulee8 ColinPeppler amjames desertfire chauhang [ghstack-poisoned]
pytorchmergebot
pushed a commit
that referenced
this pull request
Jun 12, 2024
…cache (#128482) Summary: Add a few unit tests that exercise `produce_guards_expression` and `evaluate_guards_expression` (and specifically "ToFloat" "FloatTrueDiv" added in #128418) Pull Request resolved: #128482 Approved by: https://github.com/ezyang ghstack dependencies: #128418
TharinduRusira
pushed a commit
to TharinduRusira/pytorch
that referenced
this pull request
Jun 14, 2024
Summary: I admit I'm not 100% sure what I'm doing here. I'm hitting a bug in the FX graph cache when we try to evaluate a guards expression. We're creating guards that look like this: ``` Ne(CeilToInt(FloatTrueDiv(ToFloat(8*L['t0']) - 4.0, 8.0))*CeilToInt(FloatTrueDiv(ToFloat(8*L['t1']) - 4.0, 8.0)), CeilToInt(FloatTrueDiv(ToFloat(8*L['t1']) - 4.0, 8.0))) and ... ``` It looks like we have a facility to define these operators in the SYMPY_INTERP map and we're just missing FloatTrueDiv and ToFloat. What's surprsing to me is that we're only hitting this problem with the FX graph enabled. We can create such guards, but we've never actually evaluated any? Test Plan: `TORCHINDUCTOR_FX_GRAPH_CACHE=1 python benchmarks/dynamo/torchbench.py --ci --accuracy --timing --explain --inductor --device cuda --inference --bfloat16 --only detectron2_fcos_r_50_fpn` Pull Request resolved: pytorch#128418 Approved by: https://github.com/ezyang
TharinduRusira
pushed a commit
to TharinduRusira/pytorch
that referenced
this pull request
Jun 14, 2024
…cache (pytorch#128482) Summary: Add a few unit tests that exercise `produce_guards_expression` and `evaluate_guards_expression` (and specifically "ToFloat" "FloatTrueDiv" added in pytorch#128418) Pull Request resolved: pytorch#128482 Approved by: https://github.com/ezyang ghstack dependencies: pytorch#128418
ignaciobartol
pushed a commit
to ignaciobartol/pytorch
that referenced
this pull request
Jun 14, 2024
Summary: I admit I'm not 100% sure what I'm doing here. I'm hitting a bug in the FX graph cache when we try to evaluate a guards expression. We're creating guards that look like this: ``` Ne(CeilToInt(FloatTrueDiv(ToFloat(8*L['t0']) - 4.0, 8.0))*CeilToInt(FloatTrueDiv(ToFloat(8*L['t1']) - 4.0, 8.0)), CeilToInt(FloatTrueDiv(ToFloat(8*L['t1']) - 4.0, 8.0))) and ... ``` It looks like we have a facility to define these operators in the SYMPY_INTERP map and we're just missing FloatTrueDiv and ToFloat. What's surprsing to me is that we're only hitting this problem with the FX graph enabled. We can create such guards, but we've never actually evaluated any? Test Plan: `TORCHINDUCTOR_FX_GRAPH_CACHE=1 python benchmarks/dynamo/torchbench.py --ci --accuracy --timing --explain --inductor --device cuda --inference --bfloat16 --only detectron2_fcos_r_50_fpn` Pull Request resolved: pytorch#128418 Approved by: https://github.com/ezyang
ignaciobartol
pushed a commit
to ignaciobartol/pytorch
that referenced
this pull request
Jun 14, 2024
…cache (pytorch#128482) Summary: Add a few unit tests that exercise `produce_guards_expression` and `evaluate_guards_expression` (and specifically "ToFloat" "FloatTrueDiv" added in pytorch#128418) Pull Request resolved: pytorch#128482 Approved by: https://github.com/ezyang ghstack dependencies: pytorch#128418
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):
Summary: I admit I'm not 100% sure what I'm doing here. I'm hitting a bug in the FX graph cache when we try to evaluate a guards expression. We're creating guards that look like this:
It looks like we have a facility to define these operators in the SYMPY_INTERP map and we're just missing FloatTrueDiv and ToFloat. What's surprsing to me is that we're only hitting this problem with the FX graph enabled. We can create such guards, but we've never actually evaluated any?
Test Plan:
TORCHINDUCTOR_FX_GRAPH_CACHE=1 python benchmarks/dynamo/torchbench.py --ci --accuracy --timing --explain --inductor --device cuda --inference --bfloat16 --only detectron2_fcos_r_50_fpn