[Bugfix][KV Offload] Keep Mamba block span unscaled under DCP

[jongukc]

Purpose

Native KV offloading currently multiplies every cache group's token span by the decode context parallelism world size. This is correct for attention KV, which is sharded across DCP ranks, but Mamba state is replicated and keeps its original block size.

With 16 token attention and Mamba blocks under DCP=2, the current configuration derives 32 tokens per block for both groups. The correct spans are 32 tokens for attention and 16 tokens for Mamba. For a 64 token prefix, the incorrect geometry associates offload keys with the wrong Mamba state blocks, so a later cache hit can restore recurrent state from an earlier boundary.

This change scales tokens_per_block by DCP only for AttentionSpec. Mamba and other non-attention groups retain their declared block size.

Duplicate-work check

I searched open and closed issues and PRs for native offloading, OffloadingConnector, DCP, Mamba, hybrid KV cache, and block size mismatches.
I found no matching report or fix.

Test plan

tests/v1/kv_connector/unit/offloading_connector/test_config.py

Test results

• Before the fix, 16 token attention and Mamba groups under DCP=2 both derive a 32 token offload span.
• After the fix, the groups derive the expected 32 token attention span and 16 token Mamba span.
• The corrected alignment selects Mamba state blocks at the 32 token and 64 token boundaries.
• tests/v1/kv_connector/unit/offloading_connector/test_config.py: 39 passed.

AI Assistance

OpenAI Codex was used to assist with investigation, implementation, testing, and PR preparation. The human submitter must review every changed line and be prepared to explain and defend the change end-to-end.

---

Essential Elements of an Effective PR Description Checklist

• The purpose of the PR, such as "Fix some issue (link existing issues this PR will resolve)".
• The test plan, such as providing test command.
• The test results, such as pasting the results comparison before and after, or e2e results
• (Optional) The necessary documentation update, such as updating supported_models.md and examples for a new model.

[claude]

Claude Code Review

This pull request is from a fork — automated review is disabled. A repository maintainer can comment @claude review to run a one-time review.

[orozery]

Thanks @jongukc !