Skip to content

Enable non-contiguous quantize_4bit tests for the MPS backend #1994

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

Merged
matthewdouglas merged 1 commit into from
Jul 9, 2026
+4 −4
Merged

Enable non-contiguous quantize_4bit tests for the MPS backend #1994

Changes from all commits
Commits
Show all changes
1 commit
Select commit
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
8 changes: 4 additions & 4 deletions tests/test_ops.py
View file
Open in desktop
Original file line number Diff line number Diff line change
Expand Up @@ -435,8 +435,8 @@ def test_dequantize_blockwise_non_contiguous(self, device, dtype, blocksize):
@pytest.mark.parametrize("quant_type", ["fp4", "nf4"])
@pytest.mark.parametrize("blocksize", [64, 128, 256])
def test_quantize_4bit_non_contiguous(self, device, dtype, quant_type, blocksize):
if device != "cuda":
pytest.skip("Non-contiguous fix targets CUDA backend only")
if device not in ("cuda", "mps"):
pytest.skip("Non-contiguous input handling not implemented for this backend")

# Reproduce issue #1342: non-contiguous tensor from slicing
A_full = torch.randn(3, 4, 6, 256, dtype=dtype, device=device)
Expand All @@ -458,8 +458,8 @@ def test_quantize_4bit_non_contiguous(self, device, dtype, quant_type, blocksize
@pytest.mark.parametrize("blocksize", [64, 128, 256])
def test_quantize_4bit_roundtrip_non_contiguous(self, device, dtype, quant_type, blocksize):
"""End-to-end test: quantize non-contiguous, dequantize, compare with contiguous path."""
if device != "cuda":
pytest.skip("Non-contiguous fix targets CUDA backend only")
if device not in ("cuda", "mps"):
pytest.skip("Non-contiguous input handling not implemented for this backend")

A_full = torch.randn(3, 4, 6, 256, dtype=dtype, device=device)
A_noncontig = A_full[:, ::2, :, :]
Expand Down
Loading