[gemma4] infer from config instead of hardcoding

[eustlb]

What does this PR do?

As per title. Fix #45468

[HuggingFaceDocBuilderDev]

The docs for this PR live here. All of your documentation changes will be reflected on that endpoint. The docs are available until 30 days after the last update.

[mathceo]

@eustlb thanks for the clarification and the fix - context_size // 2 + 1 makes sense here given the blocked attention path and _rel_shift.

One thing that may still be useful is a small regression test for nondefault audio configs, so this does not silently stay tied to the default length again.

Something along these lines in tests/models/gemma4/test_modeling_gemma4.py:

def test_audio_rel_pos_encoding_uses_context_size_from_config(self):
    from transformers.models.gemma4.configuration_gemma4 import Gemma4AudioConfig
    from transformers.models.gemma4.modeling_gemma4 import Gemma4AudioRelPositionalEncoding

    config = Gemma4AudioConfig(
        hidden_size=32,
        attention_chunk_size=6,
        attention_context_left=5,
        attention_context_right=1,
        use_clipped_linears=False,
    )

    module = Gemma4AudioRelPositionalEncoding(config)
    hidden_states = torch.zeros(1, 3, config.hidden_size)

    pos = module(hidden_states)

    context_size = config.attention_chunk_size + config.attention_context_left - 1 + config.attention_context_right
    expected_len = context_size // 2 + 1

    self.assertEqual(pos.shape, (1, expected_len, config.hidden_size))

    position_ids = torch.arange(context_size // 2, -1, -1, device=hidden_states.device)[..., None]
    scaled_time = position_ids * module.inv_timescales.to(device=hidden_states.device)
    expected = torch.cat([torch.sin(scaled_time), torch.cos(scaled_time)], dim=-1).to(hidden_states.dtype)

    torch.testing.assert_close(pos, expected)

Happy to open a follow-up PR for the test as well if that is preferred.

[vasqu]

Thanks! Like @mathceo also mentioned, let's add a small regression test just in case 🫡

[mathceo]

I already implemented the regression test in #45607. Since #45606 is not my branch, I can't push directly there from my side. Feel free to cherry-pick the test commit from #45607
@eustlb

[eustlb]

thanks @mathceo! added you as a co-author here

[vasqu]

run-slow: gemma4

[github-actions]

[For maintainers] Suggested jobs to run (before merge)

run-slow: gemma4

[github-actions]

Workflow Run ⚙️

This comment contains run-slow, running the specified jobs:

models: ["models/gemma4"]
quantizations: []

[github-actions]

CI Results

Workflow Run ⚙️

Commit Info

Context	Commit	Description
RUN	32ada79d	workflow commit (merge commit)
PR	d404da97	branch commit (from PR)
main	bbb51c83	base commit (on main)

Model CI Report

❌ 1 new failed tests from this PR 😭

• gemma4:
  tests/models/gemma4/test_modeling_gemma4.py::Gemma4IntegrationTest::test_export_text_only (❌ ⟹ ❌)

[vasqu]

The different failure is a different PID in the test itself lol, merging