Skip to content

Gemma 3 tests expect greedy decoding #36882

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
molbap merged 3 commits into from
Mar 21, 2025
Merged

Gemma 3 tests expect greedy decoding #36882

Changes from all commits
Commits
Show all changes
3 commits
Select commit Hold shift + click to select a range
ee917b8
tests expect greedy decoding
molbap Mar 21, 2025
69383a9
Merge branch 'main' into fix_gemma3_tests
molbap Mar 21, 2025
90d53d1
fix merge conflicts
molbap Mar 21, 2025
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
7 changes: 6 additions & 1 deletion tests/models/gemma3/test_modeling_gemma3.py
View file
Open in desktop
Original file line number Diff line number Diff line change
Expand Up @@ -567,7 +567,7 @@ def test_generation_beyond_sliding_window(self, attn_implementation: str):
input_size = inputs.input_ids.shape[-1]
self.assertTrue(input_size > model.config.sliding_window)

out = model.generate(**inputs, max_new_tokens=20)[:, input_size:]
out = model.generate(**inputs, max_new_tokens=20, do_sample=False)[:, input_size:]
output_text = tokenizer.batch_decode(out)

EXPECTED_COMPLETIONS = [" and I'm going to take a walk.\n\nI really enjoy the scenery, and I'", ", green, yellow, orange, purple, brown, black, white, gray.\n\nI'"] # fmt: skip
Expand Down Expand Up @@ -599,6 +599,11 @@ def test_generation_beyond_sliding_window_with_generation_config(self):
generation_config = GenerationConfig(max_new_tokens=5, min_new_tokens=5)
out = model.generate(**inputs, generation_config=generation_config)

out = model.generate(**inputs, generation_config=generation_config, do_sample=False)[:, input_size:]
output_text = tokenizer.batch_decode(out)
EXPECTED_COMPLETIONS = [" and I'm going to take a walk.\n\nI really enjoy the scenery, and I'", ", green, yellow, orange, purple, brown, black, white, gray.\n\nI'"] # fmt: skip
self.assertEqual(output_text, EXPECTED_COMPLETIONS)

# Generation works beyond sliding window
self.assertGreater(out.shape[1], model.config.sliding_window)
self.assertEqual(out.shape[1], input_size + 5)
Expand Down