[API/MemRef] Implement canonical stride validation for MemRefValue creation #252
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
jhalakpatel
force-pushed
the
jhalakp-memref-stride-check
branch
2 times, most recently
from
b7059ca to
238e993
Compare
jhalakpatel
changed the title
jhalakpatel
force-pushed
the
jhalakp-memref-stride-check
branch
from
238e993 to
ab28976
Compare
jhalakpatel
added a commit
to jhalakpatel/TensorRT-Incubator
that referenced
this pull request
Oct 4, 2024
MLIR-TensorRT requires strides for function arguments and results in canonical order. NVIDIA#252 adds a check to validate memref stride against a canonical stride order. In Tripy, memref strides are derived from framework DL Pack tensors. Creating a memref with a non-canonical DL Pack tensor stride throws an exception. Add a try-catch block to catch such an exception and augment with suggestions on creating a DL Pack tensor with canonical stride for Tripy-supported frameworks. Add unit tests to create a non-canonical stride tensor to validate exceptions and suggestions.
jhalakpatel
added a commit
to jhalakpatel/TensorRT-Incubator
that referenced
this pull request
Oct 4, 2024
MLIR-TensorRT requires strides for function arguments and results in canonical order. NVIDIA#252 adds a check to validate memref stride against a canonical stride order. In Tripy, memref strides are derived from framework DL Pack tensors. Creating a memref with a non-canonical DL Pack tensor stride throws an exception. Add a try-catch block to catch such an exception and augment with suggestions on creating a DL Pack tensor with canonical stride for Tripy-supported frameworks. Add unit tests to create a non-canonical stride tensor to validate exceptions and suggestions.
christopherbate
requested changes
Oct 4, 2024
jhalakpatel
force-pushed
the
jhalakp-memref-stride-check
branch
2 times, most recently
from
6204f3a to
ec624e7
Compare
christopherbate
requested changes
Oct 7, 2024
Collaborator
christopherbate
left a comment
christopherbate
left a comment
There was a problem hiding this comment.
The assertion has to be a parameter to the Python function as well, not just the C++ function.
jhalakpatel
force-pushed
the
jhalakp-memref-stride-check
branch
2 times, most recently
from
d26730f to
d67986e
Compare
christopherbate
approved these changes
Oct 8, 2024
Collaborator
christopherbate
left a comment
christopherbate
left a comment
There was a problem hiding this comment.
Some minor comments, otherwise LGTM.
Add optional stride validation in `MemRefValue::create` to compute canonical stride and compare against given strides while creaing a memref view from DLPack tensors. We need to handle special cases for zero-sized and unit-sized dimensions since frameworks deal with them arbitrarily while converting to the corresponding DLPack tensor. Add Python tests to verify both canonical and non-canonical stride validation.
jhalakpatel
force-pushed
the
jhalakp-memref-stride-check
branch
from
d67986e to
4a91333
Compare
jhalakpatel
added a commit
to jhalakpatel/TensorRT-Incubator
that referenced
this pull request
Oct 15, 2024
MLIR-TensorRT requires strides for function arguments and results in canonical order. NVIDIA#252 adds a check to validate memref stride against a canonical stride order. In Tripy, memref strides are derived from framework DL Pack tensors. Creating a memref with a non-canonical DL Pack tensor stride throws an exception. Add a try-catch block to catch such an exception and augment with suggestions on creating a DL Pack tensor with canonical stride for Tripy-supported frameworks. Add unit tests to create a non-canonical stride tensor to validate exceptions and suggestions.
jhalakpatel
added a commit
that referenced
this pull request
Oct 15, 2024
MLIR-TensorRT requires strides for function arguments and results in canonical order. #252 adds a check to validate memref stride against a canonical stride order. In Tripy, memref strides are derived from framework DL Pack tensors. Creating a memref with a non-canonical DL Pack tensor stride throws an exception. Add a try-catch block to catch such an exception and augment with suggestions on creating a DL Pack tensor with canonical stride for Tripy-supported frameworks. Add unit tests to create a non-canonical stride tensor to validate exceptions and suggestions.
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.
Add optional stride validation in
MemRefValue::createto compute canonical stride and compare against given strides while creaing a memref view from DLPack tensors. We need to handle special cases for zero-sized and unit-sized dimensions since frameworks deal with them arbitrarily while converting to the corresponding DLPack tensor. Add Python tests to verify both canonical and non-canonical stride validation.