test(TPU): relax numerical tests for TPUs + parallel TPU testing#2609
Merged
test(TPU): relax numerical tests for TPUs + parallel TPU testing#2609
Conversation
avik-pal
force-pushed
the
ap/tpu_test_fixes
branch
2 times, most recently
from
ab2b836 to
efd8382
Compare
Collaborator
Author
|
@sbrantq can you take a look at the probprog tests for TPU https://github.com/EnzymeAD/Reactant.jl/actions/runs/22694756046/job/65798539411?pr=2609, most are likely just tolerance issues |
avik-pal
force-pushed
the
ap/tpu_test_fixes
branch
4 times, most recently
from
119ee80 to
9d5d92a
Compare
Member
Looks like JAX used to generate reference results is falling back to CPU here, giving different rng results. I guess the fix would be only doing pointwise check when Reactant is using a CPU backend (will commit a fix shortly) |
avik-pal
force-pushed
the
ap/tpu_test_fixes
branch
2 times, most recently
from
8c2a9c5 to
0bf22d5
Compare
avik-pal
changed the title
wsmoses
reviewed
Mar 5, 2026
| end | ||
| else | ||
| if RunningOnTPU | ||
| @warn "Skipping MultiRotate test on TPU" |
Member
There was a problem hiding this comment.
what crashes here, will concurrently investigate?
Collaborator
Author
There was a problem hiding this comment.
wsmoses
reviewed
Mar 5, 2026
wsmoses
approved these changes
Mar 5, 2026
giordano
reviewed
Mar 5, 2026
avik-pal
force-pushed
the
ap/tpu_test_fixes
branch
2 times, most recently
from
fd1ad4f to
9eda871
Compare
avik-pal
force-pushed
the
ap/tpu_test_fixes
branch
from
50c6653 to
4da2f49
Compare
Collaborator
Author
|
@sbrantq should we disable these on TPU? Seems like an XLA internal bug │ F0306 03:13:48.227292 15978 shape_util.cc:1214] Check failed: return_shape->IsTuple() Invalid index {0} for shape u32[2,2]{1,0}
│ *** Check failure stack trace: ***
│ @ 0x7a3282749064 absl::log_internal::LogMessage::SendToLog()
│ @ 0x7a3282749018 absl::log_internal::LogMessage::Flush()
│ @ 0x7a32820fa4d4 xla::ShapeUtil::GetSubshape()
│ @ 0x7a327527bdfb xla::ShapeTree<>::CopySubtreeFrom()
│ @ 0x7a327527a843 xla::HloReplicationAnalysis::ComputeHloReplicationOnComputation()
│ @ 0x7a327527b1da xla::HloReplicationAnalysis::ComputeHloReplicationOnComputation()
│ @ 0x7a3275279d60 xla::HloReplicationAnalysis::ComputeHloReplicationOnComputation()
│ @ 0x7a3275279d60 xla::HloReplicationAnalysis::ComputeHloReplicationOnComputation()
│ @ 0x7a3275279d60 xla::HloReplicationAnalysis::ComputeHloReplicationOnComputation()
│ @ 0x7a327527bf3a xla::HloReplicationAnalysis::ComputeHloReplication()
│ @ 0x7a327527e33b xla::HloReplicationAnalysis::Run()
│ @ 0x7a327527e245 xla::HloReplicationAnalysis::Run()
│ @ 0x7a327523ce75 xla::AllReduceSimplifier::RunImpl()
│ @ 0x7a327e7320e5 xla::HloPassPipeline::RunPassesInternal<>()
│ @ 0x7a327e73192f xla::HloPassPipeline::RunImpl()
│ @ 0x7a3273223fca xla::HloPassFix<>::RunOnChangedComputationsOnce()
│ @ 0x7a3273223681 xla::HloPassFix<>::RunToFixPoint()
│ @ 0x7a3273223260 xla::HloPassFix<>::RunImpl()
│ @ 0x7a327e7320e5 xla::HloPassPipeline::RunPassesInternal<>()
│ @ 0x7a327e73192f xla::HloPassPipeline::RunImpl()
│ @ 0x7a32732196c9 xla::jellyfish::(anonymous namespace)::HloOptimizeThroughLayoutAssignment()::$_0::operator()()
│ @ 0x7a327df59c93 absl::internal_any_invocable::LocalInvoker<>()
│ @ 0x7a32825a42d6 Thread::ThreadBody()
│ @ 0x7a39edd17aa4 (unknown)
│ @ 0x7a39edda4c6c (unknown)
│ https://symbolize.stripped_domain/r/?trace=7a3282749063,7a3282749017,7a32820fa4d3,7a327527bdfa,7a327527a842,7a327527b1d9,7a3275279d5f,7a3275279d5f,7a3275279d5f,7a327527bf39,7a327527e33a,7a327527e244,7a327523ce74,7a327e7320e4,7a327e73192e,7a3273223fc9,7a3273223680,7a327322325f,7a327e7320e4,7a327e73192e,7a32732196c8,7a327df59c92,7a32825a42d5,7a39edd17aa3,7a39edda4c6b&map=
│
│ [13014] signal 6 (-6): Aborted
│ in expression starting at /__w/Reactant.jl/Reactant.jl/test/probprog/mcmc_logpdf.jl:38 |
avik-pal
force-pushed
the
ap/tpu_test_fixes
branch
from
8dfb352 to
1ac45b8
Compare
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
4 participants
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.
Uh oh!
There was an error while loading. Please reload this page.