Repository navigation
[torchlib] Cast int32 indices to int64 in aten::index before GatherND - #3221
Open
Devon Copp (djcopp) wants to merge 1 commit into
Open
Devon Copp (djcopp) wants to merge 1 commit into
Devon Copp (djcopp) wants to merge 1 commit into
Conversation
PyTorch allows int32 index tensors (e.g. table[idx] with idx.dtype == int32), but ONNX GatherND only accepts int64 indices, so the exported model was invalid: "Type 'tensor(int32)' of input parameter ... of operator (GatherND) is invalid". Cast non-int64 indices before building the GatherND index tensor, and add e2e tests for int32 and mixed int32/int64 advanced indexing. Co-Authored-By: Claude Opus 5.5 <noreply@anthropic.com>
|
Devon Copp (@djcopp) please read the following Contributor License Agreement(CLA). If you agree with the CLA, please reply with the following information.
Contributor License AgreementContribution License AgreementThis Contribution License Agreement (“Agreement”) is agreed to by the party signing below (“You”),
|
This branch has not been deployed
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
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Outdated suggestions cannot be applied.This suggestion has been applied or marked resolved.Suggestions cannot be applied from pending reviews.Suggestions cannot be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.
Description
_aten_index_onnxnow casts non-int64 index tensors to int64 before building theGatherNDindices.Added two tests to
e2e_ops_tests.py:test_index_with_int32_indicestest_index_with_mixed_int32_int64_indicesBoth fail on main and pass with this change. The existing
index_Tensor/_unsafe_indexop tests give the same results with and without the change (11 passed, 226 subtests).Motivation and Context
PyTorch accepts int32 index tensors (
table[idx]withidx.dtype == torch.int32), but ONNXGatherNDonly takes int64 indices. Sotorch.onnx.export(..., dynamo=True)currently writes a model that fails the checker:int32 indices are common in models written for GPUs and WebGPU, where int64 is slow or unsupported, so this comes up when people try to keep a graph int32 throughout.
🤖 Generated with Claude Code