Motivation

Implemented Mooncake decode-radix transfer support:

• Propagate decode_prefix_len through Mooncake metadata.
• Store it on prefill-side TransferInfo as Optional[int].
• Skip empty KV transfer for full prefix hits while still sending aux/state on the last chunk.
• Added Mooncake unit coverage and distributed decode-radix/GSM8K test coverage.
• Should be merged after [pd]: (Bug Fix) Incorrect out_cache_loc slicing in prepare_for_prebuilt #24230

TODO: Next PR will support hybrid tree (Mamba/SWA) on decode side

Collabrate with @ShangmingCai

Modifications

Accuracy Tests

GSM8K Accuracy

  model: qwen32b
  deocde_tree_run1  0.942
  decode_tree_run2  0.942
  no_deocde_tree    0.942     

Speed Tests and Profiling

Checklist

• Format your code according to the Format code with pre-commit.
• Add unit tests according to the Run and add unit tests.
• Update documentation according to Write documentations.
• Provide accuracy and speed benchmark results according to Test the accuracy and Benchmark the speed.
• Follow the SGLang code style guidance.

Review and Merge Process

1. Ping Merge Oncalls to start the process. See the PR Merge Process.
2. Get approvals from CODEOWNERS and other reviewers.
3. Trigger CI tests with comments or contact authorized users to do so.
   • Common commands include /tag-and-rerun-ci, /tag-run-ci-label, /rerun-failed-ci
4. After green CI and required approvals, ask Merge Oncalls or people with Write permission to merge the PR.