Skip to content

Fix Qwen2IntegrationTest #45268

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
ydshieh merged 1 commit into from
Apr 6, 2026
Merged

Fix Qwen2IntegrationTest #45268

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
6 changes: 3 additions & 3 deletions tests/models/qwen2/test_modeling_qwen2.py
View file
Open in desktop
Original file line number Diff line number Diff line change
Expand Up @@ -81,14 +81,14 @@ def test_model_450m_logits(self):
out = model(input_ids).logits.float().cpu()
# Expected mean on dim = -1
EXPECTED_MEAN = Expectations({
(None, None): torch.tensor([[-2.2121, -1.6335, -1.4816, -1.5035, -1.9110, -1.8979, -1.9682, -2.1980]]),
(None, None): torch.tensor([[-2.2463, -1.6463, -1.4748, -1.4913, -1.9213, -1.9016, -1.9969, -2.1761]]),

@ydshieh ydshieh Apr 6, 2026

Copy link
Copy Markdown
Collaborator Author

Choose a reason for hiding this comment

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

@Abdennacer-Badaoui Maybe this will fail on AMD, if you want to double check. Although maybe I should use ("cuda", 8) and ("rocm", 9). But let's have double check anyway.

("xpu", 3): torch.tensor([[-2.2419, -1.6216, -1.4517, -1.4963, -1.9229, -1.8966, -1.9580, -2.1484]]),
}) # fmt: off

torch.testing.assert_close(out.mean(-1), EXPECTED_MEAN.get_expectation(), rtol=1e-2, atol=1e-2)
# slicing logits[0, 0, 0:30]
EXPECTED_SLICE = Expectations({
(None, None): torch.tensor([2.7344, 4.2812, 4.1562, 2.3906, 1.1875, 2.1562, 3.1719, 3.1406, 1.2891, 3.6094, 3.3125, 1.8203, 2.9219, 3.2344, 1.5938, 6.2500, 7.4062, 7.2188, 6.5938, 6.0312, 6.1562, 5.3750, 5.9688, 5.5938, 6.1250, 1.2656, 1.6016, 3.4062, 1.7891, 3.6406]),
(None, None): torch.tensor([2.6406, 4.3125, 3.9531, 2.2656, 1.0000, 2.0312, 3.2344, 2.9219, 1.1953, 3.3750, 3.1875, 2.0156, 2.8281, 3.2656, 1.4453, 6.0625, 7.2188, 7.0312, 6.4375, 5.8750, 6.0312, 5.2500, 5.8438, 5.5000, 5.9688, 1.2734, 1.5312, 3.2344, 1.6406, 3.4375]),

@ydshieh ydshieh Apr 6, 2026

Copy link
Copy Markdown
Collaborator Author

Choose a reason for hiding this comment

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

@Abdennacer-Badaoui Maybe this will fail on AMD, if you want to double check. Although maybe I should use ("cuda", 8) and ("rocm", 9). But let's have double check anyway.

("xpu", 3): torch.tensor([2.7500, 4.4062, 4.0625, 2.2656, 1.0859, 2.1094, 3.1719, 3.0781, 1.2656, 3.5312, 3.1719, 1.9062, 2.8750, 3.2812, 1.5156, 6.1562, 7.3125, 7.1250, 6.5312, 5.9688, 6.0938, 5.3438, 5.9375, 5.5938, 6.0938, 1.2344, 1.5391, 3.2969, 1.7266, 3.5312]),
}) # fmt: skip
torch.testing.assert_close(out[0, 0, :30], EXPECTED_SLICE.get_expectation(), rtol=1e-4, atol=1e-4)
Expand Down Expand Up @@ -185,7 +185,7 @@ def test_model_450m_long_prompt_sdpa(self):
def test_speculative_generation(self):
EXPECTED_TEXT_COMPLETION = Expectations(
{
("cuda", 8): "My favourite condiment is 100% real, organic, vegan and gluten free. I use it in my recipes and",
("cuda", 8): "My favourite condiment is 100% natural, organic, gluten-free, vegan, and vegetarian. I have been making",
("rocm", (9, 4)): "My favourite condiment is 100% natural, organic, gluten-free, vegan, and vegetarian. I have been making",
("xpu", 3): "My favourite condiment is 100% natural, organic, gluten-free, vegan, and vegetarian. I have been making",
}
Expand Down
Loading