Skip to content

Fixed assumption on out_shift for quantized linear #14789

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
merged 1 commit into from
Oct 7, 2025
+10 −10
Merged

Fixed assumption on out_shift for quantized linear #14789

Show file tree
Hide file tree
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
4 changes: 2 additions & 2 deletions backends/cadence/aot/ref_implementations.py
View file
Open in desktop
Original file line number Diff line number Diff line change
Expand Up @@ -330,8 +330,8 @@ def variant(
if out_shift.numel() != 1:
raise ValueError("out_shift must be a scalar")

if out_shift.dtype != torch.int64:
raise ValueError("out_shift must be an int64")
if out_shift.dtype != torch.int32:
raise ValueError("out_shift must be an int32")

_out_shift = int(out_shift.item())
_out_multiplier = int(out_multiplier[0].item())
Expand Down
16 changes: 8 additions & 8 deletions backends/cadence/aot/tests/test_ref_implementations.py
View file
Open in desktop
Original file line number Diff line number Diff line change
Expand Up @@ -172,7 +172,7 @@ def test_quantized_add(
torch.tensor(
[1073741824], dtype=torch.int32
), # out_multiplier (0.5 * 2^31)
torch.tensor([0], dtype=torch.int64), # out_shift
torch.tensor([0], dtype=torch.int32), # out_shift
0, # out_zero_point
torch.tensor([[0]], dtype=dtype), # expected_output
per_tensor,
Expand All @@ -197,7 +197,7 @@ def test_quantized_add(
torch.tensor(
[1073741824], dtype=torch.int32
), # out_multiplier (0.5 * 2^31)
torch.tensor([0], dtype=torch.int64), # out_shift
torch.tensor([0], dtype=torch.int32), # out_shift
0, # out_zero_point
torch.tensor([[-2, -8]], dtype=dtype), # expected_output
per_tensor,
Expand All @@ -220,7 +220,7 @@ def test_quantized_add(
torch.tensor(
[1073741824], dtype=torch.int32
), # out_multiplier (0.5 * 2^31)
torch.tensor([0], dtype=torch.int64), # out_shift
torch.tensor([0], dtype=torch.int32), # out_shift
0, # out_zero_point
torch.tensor([[0, 0]], dtype=dtype), # expected_output
per_tensor,
Expand All @@ -244,7 +244,7 @@ def test_quantized_add(
torch.tensor(
[1073741824], dtype=torch.int32
), # out_multiplier (0.5 * 2^31)
torch.tensor([0], dtype=torch.int64), # out_shift
torch.tensor([0], dtype=torch.int32), # out_shift
0, # out_zero_point
torch.tensor(
[[[0, -2, -4], [-2, -7, -12]]], dtype=dtype
Expand All @@ -270,7 +270,7 @@ def test_quantized_add(
torch.tensor(
[268435456], dtype=torch.int32
), # out_multiplier (1.0 * 2^31)
torch.tensor([0], dtype=torch.int64), # out_shift
torch.tensor([0], dtype=torch.int32), # out_shift
1, # out_zero_point
torch.tensor([[1, 1]], dtype=dtype), # expected_output
per_tensor,
Expand All @@ -295,7 +295,7 @@ def test_quantized_add(
torch.tensor(
[268435456], dtype=torch.int32
), # out_multiplier (1.0 * 2^31)
torch.tensor([0], dtype=torch.int64), # out_shift
torch.tensor([0], dtype=torch.int32), # out_shift
1, # out_zero_point
torch.tensor([[1, 1]], dtype=dtype), # expected_output
False,
Expand All @@ -317,7 +317,7 @@ def test_quantized_add(
[268435456], dtype=torch.int32
), # out_multiplier (0.125 * 2^31)
torch.tensor(
[1], dtype=torch.int64
[1], dtype=torch.int32
), # out_shift (shift=1, doubles the scale)
1, # out_zero_point
torch.tensor([[1, 2]], dtype=dtype), # expected_output
Expand All @@ -339,7 +339,7 @@ def test_quantized_add(
[268435456], dtype=torch.int32
), # out_multiplier (0.125 * 2^31)
torch.tensor(
[1], dtype=torch.int64
[1], dtype=torch.int32
), # out_shift (shift=1, doubles the scale)
1, # out_zero_point
torch.tensor([[1, 2]], dtype=dtype), # expected_output
Expand Down
Loading