[Relay] Fix invalid shape function for "copy" operator #9749
Conversation
This file contains 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
mbs-octoml
requested review from
anijain2305,
areusch,
comaniac,
icemelon,
jroesch,
junrushao,
jwfromm,
kazum,
kparzysz-quic,
liangfu,
MarisaKirisame,
masahi,
mbrookhart,
merrymercy,
slyubomirsky,
tmoreau89,
tqchen,
vinx13,
wweic,
yzhliu,
zhiics and
ZihengJiang
as code owners
mbs-octoml
force-pushed
the
mbs-CORE-112
branch
2 times, most recently
from
80edf5b
to
34962cc
Compare
electriclilies
approved these changes
Dec 16, 2021
mbs-octoml
force-pushed
the
mbs-CORE-112
branch
from
f3af901
to
c8d0526
Compare
|
Thanks @electriclilies . Here goes another 6 hours of CI! |
mbs-octoml
force-pushed
the
mbs-CORE-112
branch
from
c8d0526
to
ba5c033
Compare
|
Dang, stepped on my own toes in #9759. Here we go again. |
The 'script' for of the shape function was ill-formed, resulting in a TIR shape function which did not assign to it's output, which in turn caused either OOM or assert fails as uninitialized dimensions worked their way downstream. That fix is in python/tvm/relay/op/tensor.py. Everything else is for testing and debugging as I tracked this down. Special thanks to Lily for helping me with the scalar vs tensor switch in the copy shape function. [This is CORE-112 in OctoML JIRA.]
mbs-octoml
force-pushed
the
mbs-CORE-112
branch
from
ba5c033
to
e10ae1a
Compare
masahi
approved these changes
Dec 18, 2021
ylc
pushed a commit
to ylc/tvm
that referenced
this pull request
Jan 7, 2022
The 'script' for of the shape function was ill-formed, resulting in a TIR shape function which did not assign to it's output, which in turn caused either OOM or assert fails as uninitialized dimensions worked their way downstream. That fix is in python/tvm/relay/op/tensor.py. Everything else is for testing and debugging as I tracked this down. Special thanks to Lily for helping me with the scalar vs tensor switch in the copy shape function. [This is CORE-112 in OctoML JIRA.]
ylc
pushed a commit
to ylc/tvm
that referenced
this pull request
Jan 13, 2022
The 'script' for of the shape function was ill-formed, resulting in a TIR shape function which did not assign to it's output, which in turn caused either OOM or assert fails as uninitialized dimensions worked their way downstream. That fix is in python/tvm/relay/op/tensor.py. Everything else is for testing and debugging as I tracked this down. Special thanks to Lily for helping me with the scalar vs tensor switch in the copy shape function. [This is CORE-112 in OctoML JIRA.]
qsqqsqqsq-intellif
pushed a commit
to qsqqsqqsq-intellif/tvm
that referenced
this pull request
Apr 29, 2022
The 'script' for of the shape function was ill-formed, resulting in a TIR shape function which did not assign to it's output, which in turn caused either OOM or assert fails as uninitialized dimensions worked their way downstream. That fix is in python/tvm/relay/op/tensor.py. Everything else is for testing and debugging as I tracked this down. Special thanks to Lily for helping me with the scalar vs tensor switch in the copy shape function. [This is CORE-112 in OctoML JIRA.]
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.
The
@scriptshape function forcopywas ill-formed, resulting in a TIR shapefunction which did not assign to its output. That in turn caused either OOM or
assert fails as uninitialized shape dimensions worked their way downstream.
That fix is in
python/tvm/relay/op/tensor.py.Special thanks to @electriclilies for helping me with the scalar vs tensor switch
in the shape function. It turns out
copyis used for scalars in some models.Everything else is for testing and debugging as I tracked
this down.
[This is CORE-112 in OctoML JIRA.]