enh: Bidirectional mask with multiple images for Gemma3

[brb-nv]

Description

This MR:

• Updates masking utilities needed for context phase requests with image tokens in Gemma3 VLMs. Specifically, when there are multiple images in the same request, tokens of the same image attend to each other.
• Previously, all image tokens attend to each other (regardless of them coming from different images).
• Core change is an algorithm to identify contiguous blobs of ones separated by zeros.

This MR has a nice visualization of the attention mask for global attention and sliding window attention:
huggingface/transformers#38295

Background about custom mask:

• The function get_flashinfer_attention_mask will only be called for a batch when there's at least one context request in the batch with image tokens.
• In context phase, each sample's input_ids may have a mix of image (image_token_idx) and text tokens where tokens corresponding to an image appear as a contiguous blob.
  Example: torch.IntTensor([2, 3, 4, 5, img_idx, img_idx, img_idx, ..., img_idx, 100])
• While the text tokens attend to other tokens in a causal fashion, image tokens attend to others in a causal fashion and well as attend to other image tokens in a bidirectional manner. Hence, the need for custom masking.

Test Coverage

$ pytest tests/unittest/_torch/modeling/test_modeling_gemma3.py::TestGemma3::test_gemma3_flashinfer_mask -s -v
$ pytest tests/unittest/_torch/modeling/test_modeling_gemma3.py::TestGemma3::test_gemma3_global_context_mask_multi_image -s -v
$ pytest tests/unittest/_torch/modeling/test_modeling_gemma3.py::TestGemma3::test_gemma3_global_context_mask_single_image -s -v
$ pytest tests/unittest/_torch/modeling/test_modeling_gemma3.py::TestGemma3::test_gemma3_local_context_mask_multi_image -s -v
$ pytest tests/unittest/_torch/modeling/test_modeling_gemma3.py::TestGemma3::test_gemma3_local_context_mask_single_image -s -v

GitHub Bot Help

/bot [-h] ['run', 'kill', 'skip', 'reuse-pipeline'] ...

Provide a user friendly way for developers to interact with a Jenkins server.

Run /bot [-h|--help] to print this help message.

See details below for each supported subcommand.

Details

run [--disable-fail-fast --skip-test --stage-list "A10-1, xxx" --gpu-type "A30, H100_PCIe" --add-multi-gpu-test --only-multi-gpu-test --disable-multi-gpu-test --post-merge --extra-stage "H100_PCIe-[Post-Merge]-1, xxx"]

Launch build/test pipelines. All previously running jobs will be killed.

--disable-fail-fast (OPTIONAL) : Disable fail fast on build/tests/infra failures.

--skip-test (OPTIONAL) : Skip all test stages, but still run build stages, package stages and sanity check stages. Note: Does NOT update GitHub check status.

--stage-list "A10-1, xxx" (OPTIONAL) : Only run the specified test stages. Examples: "A10-1, xxx". Note: Does NOT update GitHub check status.

--gpu-type "A30, H100_PCIe" (OPTIONAL) : Only run the test stages on the specified GPU types. Examples: "A30, H100_PCIe". Note: Does NOT update GitHub check status.

--only-multi-gpu-test (OPTIONAL) : Only run the multi-GPU tests. Note: Does NOT update GitHub check status.

--disable-multi-gpu-test (OPTIONAL) : Disable the multi-GPU tests. Note: Does NOT update GitHub check status.

--add-multi-gpu-test (OPTIONAL) : Force run the multi-GPU tests. Will also run L0 pre-merge pipeline.

--post-merge (OPTIONAL) : Run the L0 post-merge pipeline instead of the ordinary L0 pre-merge pipeline.

--extra-stage "H100_PCIe-[Post-Merge]-1, xxx" (OPTIONAL) : Run the ordinary L0 pre-merge pipeline and specified test stages. Examples: --extra-stage "H100_PCIe-[Post-Merge]-1, xxx".

For guidance on mapping tests to stage names, see docs/source/reference/ci-overview.md.

kill

kill

Kill all running builds associated with pull request.

skip

skip --comment COMMENT

Skip testing for latest commit on pull request. --comment "Reason for skipping build/test" is required. IMPORTANT NOTE: This is dangerous since lack of user care and validation can cause top of tree to break.

reuse-pipeline

reuse-pipeline

Reuse a previous pipeline to validate current commit. This action will also kill all currently running builds associated with the pull request. IMPORTANT NOTE: This is dangerous since lack of user care and validation can cause top of tree to break.

[brb-nv]

/bot run

[tensorrt-cicd]

PR_Github #11683 [ run ] triggered by Bot

[tensorrt-cicd]

PR_Github #11683 [ run ] completed with state SUCCESS
/LLM/main/L0_MergeRequest_PR pipeline #8650 completed with status: 'SUCCESS'
Pipeline passed with automatic retried tests. Check the rerun report for details.

[brb-nv]

/bot run

[tensorrt-cicd]

PR_Github #11714 [ run ] triggered by Bot

[tensorrt-cicd]

PR_Github #11714 [ run ] completed with state SUCCESS
/LLM/main/L0_MergeRequest_PR pipeline #8671 completed with status: 'SUCCESS'
Pipeline passed with automatic retried tests. Check the rerun report for details.