Separate conversion pass for torch -> tcp.custom_op to support both kernel and codegen approaches
#19
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
sjain-stanford
changed the title
torch -> tcp.custom_optorch -> tcp.custom_op to support both kernel and codegen approaches
sanjoy
approved these changes
Nov 5, 2023
| @@ -97,6 +94,11 @@ class ConvertTorchToTcp : public ConvertTorchToTcpBase<ConvertTorchToTcp> { | |||
|
|
|||
| } // namespace | |||
|
|
|||
| std::unique_ptr<OperationPass<func::FuncOp>> createConvertTorchToTcpPass() { | |||
| llvm::ArrayRef<std::string> emptyArrayRef; | |||
| return std::make_unique<ConvertTorchToTcp>(/*convertTorchOps=*/emptyArrayRef); | |||
There was a problem hiding this comment.
Can this be std::make_unique<ConvertTorchToTcp>({})
There was a problem hiding this comment.
I initially had that, but hit this compiler error:
lib/Conversion/TorchToTcp/TorchToTcp.cpp:99:10: error: no matching function for call to 'make_unique'
return std::make_unique<ConvertTorchToTcp>({});
^~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~
/usr/lib/gcc/x86_64-linux-gnu/11/../../../../include/c++/11/bits/unique_ptr.h:961:5: note: candidate template ignored: substitution failure [with _Tp = mlir::tcp::(anonymous namespace)::ConvertTorchToTcp]: deduced incomplete pack <(no value)> for template parameter '_Args'
make_unique(_Args&&... __args)
^
sjain-stanford
force-pushed
the
sambhav/custom_op_pass
branch
from
15eceea to
580dd76
Compare
|
@navahgar I am about to land this under the assumption that it addresses your comment about tcp.gather codegen. Please let me know if it doesn't or if you have any other feedback, and I'll be glad to take it up in a follow-on. |
srinathava
pushed a commit
to srinathava/mlir-tcp
that referenced
this pull request
Jun 24, 2024
Replace the tcp custom op for `torch.tensorrt.execute_engine_variadic` to `tensorrt.execute_engine` for generic usage purpose. To test, from c/c repo: `bazel test @mlir-tcp//test/...`
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.
This allows switching between codegen vs kernels approaches, for torch ops that can lower to either tcp ops (codegen) or tcp.custom_op (kernel). Addresses #17 (comment) using Option 2.