Skip to content

Enable ONNX backend tests on [U]INT4 types in Cast #27326

New issue

Have a question about this project? Sign up for a free GitHub account to open an issue and contact its maintainers and the community.

Sign up for GitHub

By clicking “Sign up for GitHub”, you agree to our terms of service and privacy statement. We’ll occasionally send you account related emails.

Already on GitHub? Sign in to your account

Open
mcollinswisc wants to merge 3 commits into
base: main
Choose a base branch
from
+2 −12
Open

Enable ONNX backend tests on [U]INT4 types in Cast #27326

Show file tree
Hide file tree
Changes from all commits
Commits
Show all changes
3 commits
Select commit Hold shift + click to select a range
da5523c
Remove skip of tests on cast from 4-bit
mcollinswisc Feb 12, 2026
5cb48d3
Remove old TODO on [u]int4 support
mcollinswisc Feb 12, 2026
fefeb36
Clarify comment on why to filter int4 tests in python
mcollinswisc Feb 12, 2026
File filter

Filter by extension

Filter by extension
Conversations
Failed to load comments.
Loading
Jump to
Jump to file
Failed to load files.
Loading
Diff view
Diff view
1 change: 0 additions & 1 deletion onnxruntime/core/providers/cpu/tensor/cast_op.cc
View file
Open in desktop
Original file line number Diff line number Diff line change
Expand Up @@ -1087,7 +1087,6 @@ ONNX_CPU_OPERATOR_VERSIONED_KERNEL(
.MayInplace(0, 0), // allocation planner will check input and output sizes match before inplacing
Cast);

// TODO(adrianlizarraga): Implement support for int4 and uint4.
Copy link
Contributor Author

@mcollinswisc mcollinswisc Feb 12, 2026

Choose a reason for hiding this comment

The reason will be displayed to describe this comment to others. Learn more.

@adrianlizarraga Does the support added since cd7112f meet the intention of this TODO?

ONNX_CPU_OPERATOR_VERSIONED_KERNEL(
Cast,
21,
Expand Down
10 changes: 0 additions & 10 deletions onnxruntime/test/onnx/TestCase.cc
View file
Open in desktop
Original file line number Diff line number Diff line change
Expand Up @@ -948,16 +948,6 @@ std::unique_ptr<std::set<BrokenTest>> GetBrokenTests(const std::string& provider
{"slice_neg_steps",
"Type parameter (Tind) bound to different types (tensor(int64) and tensor(int32) in node ()."},
{"cast_BFLOAT16_to_FLOAT", "Unexpected input data type"},
{"cast_FLOAT16_to_INT4", "Skipped until onnxruntime/cmake/external/onnx points to onnx 1.19 which should include @onnx/onnx/pull/7074"},
{"cast_FLOAT16_to_UINT4", "Skipped until onnxruntime/cmake/external/onnx points to onnx 1.19 which should include @onnx/onnx/pull/7074"},
{"cast_FLOAT_to_INT4", "Skipped until onnxruntime/cmake/external/onnx points to onnx 1.19 which should include @onnx/onnx/pull/7074"},
{"cast_FLOAT_to_UINT4", "Skipped until onnxruntime/cmake/external/onnx points to onnx 1.19 which should include @onnx/onnx/pull/7074"},
{"cast_INT4_to_FLOAT", "Skipped until onnxruntime/cmake/external/onnx points to onnx 1.19 which should include @onnx/onnx/pull/7074"},
{"cast_INT4_to_FLOAT16", "Skipped until onnxruntime/cmake/external/onnx points to onnx 1.19 which should include @onnx/onnx/pull/7074"},
{"cast_INT4_to_INT8", "Skipped until onnxruntime/cmake/external/onnx points to onnx 1.19 which should include @onnx/onnx/pull/7074"},
{"cast_UINT4_to_FLOAT", "Skipped until onnxruntime/cmake/external/onnx points to onnx 1.19 which should include @onnx/onnx/pull/7074"},
{"cast_UINT4_to_FLOAT16", "Skipped until onnxruntime/cmake/external/onnx points to onnx 1.19 which should include @onnx/onnx/pull/7074"},
{"cast_UINT4_to_UINT8", "Skipped until onnxruntime/cmake/external/onnx points to onnx 1.19 which should include @onnx/onnx/pull/7074"},
{"loop13_seq", "Creation of empty sequences is currently not supported in the test runner"},
{"sequence_insert_at_front", "shape mismatch, expect {4} got {3}"},
{"cast_FLOAT_to_BFLOAT16", "expect uint16 got bfloat16"},
Expand Down
3 changes: 2 additions & 1 deletion onnxruntime/test/testdata/onnx_backend_test_series_filters.jsonc
View file
Open in desktop
Original file line number Diff line number Diff line change
Expand Up @@ -320,7 +320,8 @@
"^test_ai_onnx_ml_tree_ensemble_single_tree_cpu",
"^test_ai_onnx_ml_tree_ensemble_set_membership_cuda",
"^test_ai_onnx_ml_tree_ensemble_single_tree_cuda",
// ORT Cast(21) implementation doesn't support int4 yet
// ORT's sess.run() does not support ml_dtypes (int4/uint4); the Cast implementation itself works (but the test
// would need to use OrtValue API, or the ORT Python API needs to support ml_dtypes).
"^test_cast_INT4_to_FLOAT_cpu",
"^test_cast_INT4_to_INT8_cpu",
"^test_cast_UINT4_to_FLOAT_cpu",
Expand Down