[Misc][Main2Main] Upgrade vLLM release dependency to v0.29.0

[zhao-stack]

What this PR does / why we need it?

Upgrade the supported vLLM release dependency from v0.28.0 to the official v0.29.0, alongside the fixed vLLM main commit inherited from #16216 (already merged).

| Reference | Commit |

|---|---|

| vllm-ascend rebase base | 9dc6704559ebe2809b5cb7b7fc44184bd1338b3c |

| PR head | 8c5d785931c495701bf1da8b5bc80b321ebd3cd0 |

| Fixed vLLM main | 84030bbe3d74d99bad477a3d2e37a973ccd8865c |

| Previous release: v0.28.0 | 2cf0a6915ce544dc493a0990f2ea38d81601128a |

| Target release: v0.29.0 | 98dff2a81d747d1dba01a47f939f48c3526d4206 |

• Use common implementations where v0.29.0 and the fixed main share KV-cache layouts, Mamba copy/group APIs, PCP handling and speculative-decoding contracts.

• Retain explicit vllm_version_is("0.29.0") branches for contracts that still differ, including RoPE, scheduler block snapshots, InputBatch, ReplaySSM, KV zeroing and DSpark PP handling.

• Remove obsolete v0.28.0 compatibility and adapt existing test fixtures. Version detection uses package versions and the explicit VLLM_VERSION override, with local-version suffix handling and UT environment isolation retained; no hard-coded release-SHA inference remains.

• Apply the Ascend PCP+DP validator patch only to v0.29.0, preserving other validations and EPLB platform binding. Rebuild dependent Pydantic schemas so nested configuration validation uses the patched validator. The global patch documentation records its rationale and removal criteria.

• Preserve [Feature][MRV2] Support GLM5.2 DSpark with PP #15747's Spec+PP protocol/partition handling after rebase; use the common exact-release selector and the real function-local DSpark sharing import.

This release-only upgrade does not require a new main old-to-new interface scan. The latest rebase incorporates #16905, which reverted #16544 at 9dc6704559ebe2809b5cb7b7fc44184bd1338b3c. The now-unnecessary V4.1 drafter import gate and tuple annotation have been removed. Other release adaptations, including #15747 Spec+PP handling, remain. Detailed contract evidence is retained below for review.

Per-file adaptations and exact upstream evidence

Per-file adaptation ledger

Evidence IDs refer to the exact source contract and upstream diff table below. Every row is syntax checked; branch-normalized AST comparison confirms unchanged main function bodies except the version identity helper and the explicitly retired propose argument/type annotations. Both supported versions completed CPU and NPU execution as recorded below.

| Ascend file / symbols | Disposition and upstream evidence | Verification |

|---|---|---|

| vllm_ascend/patch/worker/patch_v2/patch_dspark.py / _load_dspark_model_with_target_quant; tests/ut/patch/worker/test_patch_dspark_pp.py | Preserve rebased Ascend #15747 (82df9d871), including manual PP partition masking. vLLM #52809 diff (91a893de64722019ea2faf852e06cabe143b3490) moved _should_share inside the loader on both supported pins, so retain the earlier release fix: patch eagle_utils, never read/patch a nonexistent dspark_utils._should_share. v0.29 keeps its global PP guard; main has native PP via #50514 (d87a440f88e28e5b37f9b1e22ce214d0426d5352). | Existing UT now checks the real function-local import, absent module alias, and restoration on success/failure for both lanes; partition and PP assertions retained. Ruff/syntax pass; actual CPU/NPU pending. |

| vllm_ascend/worker/v2/pp_utils.py / use_legacy_spec_pp; tests/ut/worker/v2/test_pp_utils.py | #15747 added broad 0.28/0.29 routing and an obsolete 0.28 dev-build recognition path. For the two supported pins, #50514 exists only on fixed main. Route through vllm_version_is("0.29.0"); retain package/local-suffix and explicit environment-override semantics. No release-SHA inference or third release lane. | Existing UTs exercise the real uncached version helper with monkeypatch isolation, local suffix, fixed-main dev string, explicit override, and non-target versions. Isolated routing checks and Ruff/syntax pass; full CPU UT pending. |

| vllm_ascend/_310p/model_runner_310p.py
_prepare_inputs, _allocate_kv_cache_tensors | #51718, #56078; common contracts merged, differing release contracts explicitly gated as detailed below. | Static checked; prior-head dual-version results below; rebased CI pending |

| vllm_ascend/_310p/spec_decode/llm_base_proposer_310.py
set_inputs_first_pass | #56078; common contracts merged, differing release contracts explicitly gated as detailed below. | Static checked; prior-head dual-version results below; rebased CI pending |

| vllm_ascend/_310p/worker/v2/model_runner.py
initialize_kv_cache, _allocate_kv_cache_tensors, _prepare_inputs_310p | #51718/#54436; use shared standardized layouts and retain the v0.29.0 InputBatch gate. The rebase preserves vllm-ascend #16043's MTP copy tracking while removing only legacy v0.28.0 allocation paths. | Static checked; prior-head dual-version results below; rebased CI pending |

| vllm_ascend/_310p/worker/v2/rope.py
get_310p_rope_state | #56078; common contracts merged, differing release contracts explicitly gated as detailed below. | Static checked; prior-head dual-version results below; rebased CI pending |

| vllm_ascend/attention/attention_v1.py
module imports/dispatch | #52839; both lanes use the common PCP import. Rebase keeps current upstream graph code and removes only the v0.28.0 import branch. | Static checked; prior-head dual-version results below; rebased CI pending |

| vllm_ascend/attention/context_parallel/sfa_cp.py
module imports/dispatch | #52839; common contracts merged, differing release contracts explicitly gated as detailed below. | Static checked; prior-head dual-version results below; rebased CI pending |

| vllm_ascend/attention/indexer.py
module imports/dispatch | #52839; common contracts merged, differing release contracts explicitly gated as detailed below. | Static checked; prior-head dual-version results below; rebased CI pending |

| vllm_ascend/attention/mla_v1.py
module imports/dispatch | #52839; common contracts merged, differing release contracts explicitly gated as detailed below. | Static checked; prior-head dual-version results below; rebased CI pending |

| vllm_ascend/core/dyntra_lb_scheduler.py
module imports/dispatch | #51358, #54853; common contracts merged, differing release contracts explicitly gated as detailed below. | Static checked; prior-head dual-version results below; rebased CI pending |

| vllm_ascend/core/recompute_scheduler.py
module imports/dispatch, schedule | #51358, #54853; common contracts merged, differing release contracts explicitly gated as detailed below. | Static checked; prior-head dual-version results below; rebased CI pending |

| vllm_ascend/core/scheduler_profiling_chunk.py
module imports/dispatch | #51358, #54853; common contracts merged, differing release contracts explicitly gated as detailed below. | Static checked; prior-head dual-version results below; rebased CI pending |

| vllm_ascend/core/kv_cache_interface.py
get_kv_cache_compression_ratio, AscendMLAAttentionSpec, get_storage_block_size | #51718, #53906; common contracts merged, differing release contracts explicitly gated as detailed below. | Static checked; prior-head dual-version results below; rebased CI pending |

| vllm_ascend/distributed/kv_transfer/kv_pool/ascend_store/layerwise_cache_layout.py
_merge_specs | #51718; common contracts merged, differing release contracts explicitly gated as detailed below. | Static checked; prior-head dual-version results below; rebased CI pending |

| vllm_ascend/distributed/kv_transfer/kv_pool/kv_offload/preempt_offload/manager.py
_derive_cpu_config | #51718; common contracts merged, differing release contracts explicitly gated as detailed below. | Static checked; prior-head dual-version results below; rebased CI pending |

| vllm_ascend/distributed/kv_transfer/kv_pool/kv_offload/native/npu.py
create_worker | #52615; common contracts merged, differing release contracts explicitly gated as detailed below. | Static checked; prior-head dual-version results below; rebased CI pending |

| vllm_ascend/models/deepseek_mtp.py
load_weights | #53106; common contracts merged, differing release contracts explicitly gated as detailed below. | Static checked; prior-head dual-version results below; rebased CI pending |

| vllm_ascend/models/deepseek_v4/indexer.py
get_kv_cache_spec | #51718; common contracts merged, differing release contracts explicitly gated as detailed below. | Static checked; prior-head dual-version results below; rebased CI pending |

| vllm_ascend/models/glm5next/cache_config.py
make_tensor | #51718; common contracts merged, differing release contracts explicitly gated as detailed below. | Static checked; prior-head dual-version results below; rebased CI pending |

| vllm_ascend/models/glm5next/kv_cache.py
get_kv_cache_spec | #51718; common contracts merged, differing release contracts explicitly gated as detailed below. | Static checked; prior-head dual-version results below; rebased CI pending |

| vllm_ascend/models/kimi_k3_dspark.py
load_weights | #53106; common contracts merged, differing release contracts explicitly gated as detailed below. | Static checked; prior-head dual-version results below; rebased CI pending |

| vllm_ascend/models/layer/attention/layer.py
get_kv_cache_spec | #51718; common contracts merged, differing release contracts explicitly gated as detailed below. | Static checked; prior-head dual-version results below; rebased CI pending |

| vllm_ascend/patch/platform/patch_balance_schedule.py
module imports/dispatch | #51358, #54853; common contracts merged, differing release contracts explicitly gated as detailed below. | Static checked; prior-head dual-version results below; rebased CI pending |

| vllm_ascend/patch/platform/patch_kv_cache_coordinator.py
__init__ | #53614; common contracts merged, differing release contracts explicitly gated as detailed below. | Static checked; prior-head dual-version results below; rebased CI pending |

| vllm_ascend/patch/platform/patch_kv_cache_utils.py
module imports/dispatch, _get_max_layers_per_page_size, _ascend_max_memory_usage_bytes_from_groups, _ascend_get_kv_cache_config_from_groups | #51718; common contracts merged, differing release contracts explicitly gated as detailed below. | Static checked; prior-head dual-version results below; rebased CI pending |

| vllm_ascend/patch/platform/patch_use_v2_model_runner.py
module imports/dispatch, _patched_get_unsupported_features | #53853, #53183; common contracts merged, differing release contracts explicitly gated as detailed below. | Static checked; prior-head dual-version results below; rebased CI pending |

| vllm_ascend/patch/worker/patch_bind_kv_cache.py
bind_kv_cache | #51718; common contracts merged, differing release contracts explicitly gated as detailed below. | Static checked; prior-head dual-version results below; rebased CI pending |

| vllm_ascend/patch/worker/patch_bind_kv_cache.py::bind_kv_cache | v0.29.0 has no ReplaySSM ring-tracker helper; fixed main requires it after KV binding. Evidence: #52506, adebc41b7e9f1085d3f73434e23beb76883b9eb4, worker-utils diff. Keep an explicit v0.29.0 exclusion and preserve the fixed-main call. | Failure reproduced on v0.29.0; Historical release/main NPU validation passed; current-head CI pending |

| vllm_ascend/patch/worker/patch_mamba_utils.py
module imports/dispatch, _get_state_copy_funcs_for_layer | #53896; common contracts merged, differing release contracts explicitly gated as detailed below. | Static checked; prior-head dual-version results below; rebased CI pending |

| vllm_ascend/patch/worker/patch_v2/patch_attn_utils.py
module imports/dispatch | #51718; common contracts merged, differing release contracts explicitly gated as detailed below. | Static checked; prior-head dual-version results below; rebased CI pending |

| vllm_ascend/patch/worker/patch_v2/patch_dspark.py
_load_dspark_model_with_target_quant | #50514, #52809; v0.29.0 keeps get_pp_group module-bound but imports _should_share locally from Eagle utilities; fixed main removes the PP guard. Keep the release get_pp_group patch, share through the common Eagle utility, and delete the obsolete v0.28.0 dspark_utils._should_share patch. | Exact release failure reproduced; Historical release/main CPU/NPU validation passed; current-head CI pending |

| vllm_ascend/spec_decode/llm_base_proposer.py
model_returns_tuple, __init__, set_inputs_first_pass | #56078, #52861; common contracts merged, differing release contracts explicitly gated as detailed below. | Static checked; prior-head dual-version results below; rebased CI pending |

| vllm_ascend/utils.py
get_kv_cache_tensor_layers, register_ascend_customop, vllm_version_is | #52839, #51718, #52494; common contracts merged, differing release contracts explicitly gated as detailed below. | Static checked; prior-head dual-version results below; rebased CI pending |

| vllm_ascend/worker/model_runner_v1.py
_get_ascend_mamba_state_copy_funcs, _allocate_kv_cache_tensors, _prepare_inputs, _dummy_run, _reshape_kv_cache_tensors, get_kv_cache_spec, NaN helpers | #51718/#53896/#56078 plus #50323; remove v0.28.0 layouts/copy APIs and import the NaN helpers directly because v0.29.0 and fixed main expose the same contract. XD-RoPE remains explicitly gated to v0.29.0. | Static checked; historical dual-version CPU/NPU validation passed; current-head CI pending |

| vllm_ascend/worker/v2/aclgraph_utils.py
capture | #53869 makes pcp_manager common to both lanes. The rebase preserves vllm-ascend #16409's host-parameter-update revert and removes only the obsolete v0.28.0 capture branch. | Static checked; prior-head dual-version results below; rebased CI pending |

| vllm_ascend/worker/v2/attn_utils.py
get_kv_cache_spec, _allocate_kv_cache, _reshape_kv_cache_v2 | #51718; common contracts merged, differing release contracts explicitly gated as detailed below. | Static checked; prior-head dual-version results below; rebased CI pending |

| vllm_ascend/worker/v2/block_table.py
__init__, init_block_table_layout_tensors, compute_slot_mappings | #51718, #51031; common contracts merged, differing release contracts explicitly gated as detailed below. | Static checked; prior-head dual-version results below; rebased CI pending |

| vllm_ascend/worker/v2/model_runner.py
module imports/dispatch, prepare_inputs, __init__, sample_tokens, prepare_dummy_attn, execute_model | #50514, #54436, #52506, #55212, #53515; common contracts merged, differing release contracts explicitly gated as detailed below. | Static checked; prior-head dual-version results below; rebased CI pending |

| vllm_ascend/worker/v2/pcp_manager.py
partition_batch | #53515; common contracts merged, differing release contracts explicitly gated as detailed below. | Static checked; prior-head dual-version results below; rebased CI pending |

| vllm_ascend/worker/v2/sample/gumbel.py
module imports/dispatch | #54282; common contracts merged, differing release contracts explicitly gated as detailed below. | Static checked; prior-head dual-version results below; rebased CI pending |

| vllm_ascend/worker/v2/spec_decode/__init__.py
init_speculator | #49811; common contracts merged, differing release contracts explicitly gated as detailed below. | Static checked; prior-head dual-version results below; rebased CI pending |

| vllm_ascend/worker/v2/spec_decode/autoregressive/speculator.py
propose | #53694; common contracts merged, differing release contracts explicitly gated as detailed below. | Static checked; prior-head dual-version results below; rebased CI pending |

| vllm_ascend/worker/v2/spec_decode/dflash/speculator.py
module imports/dispatch, propose | #53694, #52188; common contracts merged, differing release contracts explicitly gated as detailed below. | Static checked; prior-head dual-version results below; rebased CI pending |

| vllm_ascend/worker/v2/spec_decode/dspark/speculator.py
propose | #53694; common contracts merged, differing release contracts explicitly gated as detailed below. | Static checked; prior-head dual-version results below; rebased CI pending |

| vllm_ascend/worker/worker.py
_scale_kv_cache_memory_for_multi_group, wake_up | #51718, #53508; common contracts merged, differing release contracts explicitly gated as detailed below. | Static checked; prior-head dual-version results below; rebased CI pending |

Exact upstream evidence

| Upstream change | Full commit SHA / direct diff | Actual supported contracts and branch decision |

|---|---|---|

| #51718 | 8bdc70ec7b379279ec0152343239c2d50aced687
vllm/v1/kv_cache_interface.py | Both use layers/layer_stride/block_stride/offset, tokens_per_state, CircularBufferSpec and standardized backing; retire shared_by/compress_ratio allocation branches. |

| #52839 | 58e5ee0158b6a264c3506f00480e108a34b33ee3
vllm/v1/attention/ops/pcp.py | Both import PCP operations from vllm.v1.attention.ops.pcp. |

| #53896 | e126687a9a828d513c01a07cd69f025f27d63280
vllm/v1/worker/mamba_utils.py | Both use Mamba copy-function dictionaries and unwrap UniformTypeKVCacheSpecs. |

| #53106 | 1fe3a1571ac67581478a11743e55a306de1d136f
vllm/model_executor/models/utils.py | Both use WeightsMapper instead of AutoWeightsLoader skip_prefixes/skip_substrs. |

| #53906 | 98ed0856f31fa3aaf5e27464e2b4ef5a8ee6b2f5
vllm/v1/kv_cache_interface.py | Only pinned main has the optional MLA storage_block_size dataclass field; release keeps the Ascend derived property, using tokens_per_state. |

| #56078 | 719284fe158f1be8a9dd92953295fc9d49015730
vllm/config/model.py | Release retains uses_xdrope_dim and three M-RoPE dimensions; pinned main uses mrope_num_dims and unified RoPE. |

| #52615 | 138d137b5b955a5ebee98dfc946ecdd65d7b87ce
vllm/v1/kv_offload/cpu/spec.py | Release uses num_blocks/kv_bytes_per_block; main uses num_chunks/kv_bytes_per_chunk. |

| #51358 | 6b110badbb22d3f66c7218b71138f13b7a6b3419
vllm/v1/core/sched/output.py | Release now has boundary_state_offloads and KVConnectorBlockState; remove partial_tail_offloads plumbing. |

| #54853 | 0b066293f3c738a0cbd3a087bf893f2f4dcd61f2
vllm/v1/core/sched/output.py | Release constructor takes block_ids snapshots; main takes req_ids and resolve_block_ids. Keep exact release snapshot membership and main lazy-resolution membership. |

| #53614 | 144e79c8106da23141ac010394b782f730cc7fe8
vllm/v1/core/kv_cache_coordinator.py | Only main configures drop_eagle_checkpoint_block for replay-aligned Mamba checkpoints. |

| #50514 | d87a440f88e28e5b37f9b1e22ce214d0426d5352
vllm/v1/worker/gpu/spec_decode/dspark/utils.py | Release retains module-level PP/share symbols and Ascend PP workaround; main has the subsequent PP integration. |

| #52809 | 91a893de64722019ea2faf852e06cabe143b3490
vllm/v1/worker/gpu/spec_decode/dspark/utils.py | Between v0.28.0 and v0.29.0, _should_share moved from a DSpark module binding to a function-local import from Eagle utilities. The shared Eagle patch remains effective; the old DSpark-module read/write must be removed. |

| #54436 | 6bafc049aae6c26e210162630914ee9177a4b586
vllm/v1/worker/gpu/input_batch.py | Release InputBatch requires max_seq_len_np; main removed it. |

| #52506 | adebc41b7e9f1085d3f73434e23beb76883b9eb4
vllm/v1/worker/gpu/model_runner.py | Only main accepts valid_dummy_state_slots/valid_state_slots capture arguments. |

| #55212 | 83990f5bcc0b5eb08b2f1fd1b109fa8a31b122f3
vllm/v1/worker/gpu/model_runner.py | Release prepares DCP local sequence lengths before partitioning; main initializes DCP metadata afterwards. |

| #53515 | b1fbbc2ade51e3826bc92e4733c9c692ee21d42d
vllm/v1/worker/gpu/pcp_manager.py | Both accept padded_num_tokens for persistent PCP input buffers. |

| #53869 | b3af042abd8fe5a297ec3ec72db276fd661a67b3
vllm/v1/worker/gpu/cudagraph_utils.py | Both accept pcp_manager during graph capture. |

| #51031 | 0ecc284790e5403f74b899524ef82ecb69f83cb3
vllm/v1/worker/gpu/block_table.py | Both distinguish KV and kernel block sizes during DCP slot mapping. |

| #54282 | fe755c88995ad468882517b6c4bdd60138d46a3a
vllm/v1/worker/gpu/sample/gumbel.py | Both gumbel sampling APIs include is_drafting. |

| #52188 | d1e3eee6fb8ed3623241ef5c8e3ac533f775bff9
vllm/v1/worker/gpu/spec_decode/dflash/speculator.py | Both DFlash input kernels take cp_rank/CP_SIZE/CP_INTERLEAVE. |

| #53694 | 5acc1c4e4b8730298cbff7a4a7c68c814dc24fd7
vllm/v1/worker/gpu/spec_decode/autoregressive/speculator.py | Both propose APIs take DPSyncState; remove the obsolete token-count argument and retain replicated-PCP synchronization. |

| #49811 | 01af92e175407231b1433b0aef01a1b9c983d955
vllm/v1/worker/gpu/spec_decode/extract_hidden_states.py | Both support extract_hidden_states on MRV2; remove old unsupported dispatch/skip. |

| #53508 | 479eeb32d2b432dbb4e442fbd3f94ca2eca35d67
vllm/v1/worker/gpu_model_runner.py | Both remove post_kv_cache_wake_up; retire release-only call. |

| #52494 | 3ff4f02dfe69abc1a0375d1ea8d8d5cb25609fcc
vllm/models/kimi_k3/amd/mla.py | Only main provides KimiK3MultiHeadLatentAttentionWrapper; retain release exclusion. |

| #52861 | b09bd69b5bf14911abef9a0e8e493b83c8a38fa6
vllm/v1/spec_decode/llm_base_proposer.py | Both include DeepseekV32MTPModel in the two-hidden-state architecture set. |

| #54713 | b28c3e1568bfae930f61d4b24940e47528c85d4a
vllm/v1/core/single_type_kv_cache_manager.py | Only main takes replay_boundaries in compressed-prefix hit lookup; preserve release calls without that keyword. |

| #42785 | 442d36031ce710aa2a777c353becb660c57ab2bd
vllm/v1/worker/encoder_cudagraph.py | Only main capture callers pass axis_keys; preserve the existing Ascend rejection of nonempty axes. |

| #52358 | 8f816a3f665489d7f0d222115d4f72ebab01076b
vllm/v1/worker/gpu/model_runner.py | Both ExecuteModelState have dp_sync; only main has cudagraph_stats. |

| #52789 | 9eb9d9d3953959695108600c8ed33d36bc6a1e5f
vllm/v1/core/sched/scheduler.py | Release already has mamba_has_prefill_checkpoint_blocks; later main also has fine-grained prefix-cache state. |

| #51251 | 7bbbf7c8e5040f7ebd374e8cbc657e01af1136dd
vllm/config/vllm.py | v0.29.0 and main expose ec_manager_config; retire the old release-only ScoreEncoder configuration skip. |

| #53240 | b2db227a7c4c5e55f85524a094b607e5d27408b4
vllm/model_executor/layers/fused_moe/routed_experts_capturer.py | v0.29.0 supports MRV2 routed-expert capture and unwraps uniform cache groups; retire the old release-only replay skip. |

| #53853 | e376d45e82cb7e220da430e3179e81eb0922cf56
vllm/config/vllm.py | Both delegate PCP compatibility validation to the PCP manager. |

| #53183 | 4aab2b0ebed20343efe543c633f71b3c1336d5b8
vllm/config/vllm.py | Both expose the V1 unsupported-feature helper used by the existing Ascend MRV1 feature filter. |

Additional inherited contracts

| vllm-ascend change | Why it is required | Upstream cause and direct link | Lane | Verification |

|---|---|---|---|---|

| _310p/worker/v2/model_state.py: remove tuple-return and RecoverSSM v0.28 fallbacks | v0.29.0 and fixed main both return dict[MambaSpec, list[int]] from get_mamba_groups and both initialize recoverssm; keeping the old fallback would preserve an unsupported third contract | v0.29.0 mamba groups, fixed-main mamba groups, v0.29.0 RecoverSSM, fixed-main RecoverSSM | both, common implementation | Existing constructor UT now asserts the parent-created RecoverSSM value is retained; Ruff and compileall pass |

| worker/v2/model_runner.py: always forward kv_cache_allocation_context | v0.29.0 and fixed main both accept this keyword; the new-base gate was specifically for v0.28.0 | v0.29.0 signature, fixed-main signature, vllm-ascend #16791 | both, common implementation | Existing UT continues to assert the exact context object reaches the parent; Ruff and compileall pass |

| _310p/worker/v2/model_runner.py: select the release KV-zeroing contract with vllm_version_is("0.29.0") | v0.29.0 exposes KVCacheGroupSpec.is_eagle_group but lacks SpeculativeConfig.use_eagle_block_drop; fixed main added the method | vLLM #53388 diff, commit 481839ad9e5ebf87aecb54fa5c9d986bd5ea4b81; v0.29.0 group field, fixed-main helper | release differs from main | Existing two-path KV-zeroing UT retained and renamed for v0.29.0; Ruff and compileall pass |

| existing DFlash kernel UT: remove v0.28-only kwarg omission | the current Ascend kernel accepts the CP arguments and the only excluded lane was v0.28.0, which this PR replaces | vllm-ascend #15098 | both, common invocation | Existing NPU test remains enabled with all assertions; Current-head CI pending |

| Ascend change | Why / upstream cause | Lane | Verification |

|---|---|---|---|

| patch/platform/patch_parallel_config.py, registration, and global patch documentation | Allow Ascend PCP+DP by removing the generic GPU restriction, following vLLM #54523, commit 7c2f1ff4958eaf0818405e9192c71608fe4a16b1. Preserve dynamic parallel_config.current_platform lookup and rebuild ParallelConfig → SpeculativeConfig → VllmConfig. | v0.29.0 only | Existing configuration/EPLB UTs and historical PCP+DP NPU execution passed; current-head CI pending. |

| ops/triton/v2/sample/categorical_sample.py, existing Gumbel UT | Both supported contracts require is_drafting, from #54282, fe755c88995ad468882517b6c4bdd60138d46a3a; preserve the inherited optimized kernel and use a common wrapper. | Both | Existing positive drafting assertion retained; current-head CI pending. |

| patch/platform/patch_kv_cache_coordinator.py, existing prefix-cache UT | Both support cache_hit_alignment_tokens, introduced by #53598, 2ba984a5d06db414f3b2474fe9338faf6cd80a1c; retire the v0.28-only write-mask branch. | Both | Existing assertions retained; current-head CI pending. |

Rebase and retired-fallback evidence

| Change | Exact source evidence | Decision |

|---|---|---|

| tests/e2e/conftest.py::PROMPT_CONFIGS[HunyuanOCR] | vLLM #53272, d53b1c2efc0ca7161c3844f51ca9acbcdb7129d5 removes native Hunyuan V1/VL before v0.29.0; both exact supported sources lack it. The old conditional came from vllm-ascend #14898, e5118d151314ae18e56c0b63aa8dd00d294adc22. | Remove the v0.28.0 selector and use the existing exclusion for both supported lanes. This does not migrate the protected SFA PCP skip or add a v0.29.0 accuracy skip. |

| vllm_ascend/worker/model_runner_v1.py NaN helper imports and tests/ut/worker/test_model_runner_v1_nan_detection.py | vLLM #50323, 12292d94b25869be2af6b6d4f8eea6c2445e935f; both exact sources export nans_to_dict, gpu_sync_allowed, and raise_if_nan_logits. The v0.28.0 fallback originated in vllm-ascend #14898, e5118d151314ae18e56c0b63aa8dd00d294adc22. | Delete import-failure/None fallbacks and the existing UT's obsolete availability skip; assertions remain unchanged and execute on both lanes. |

| vllm_ascend/worker/v2/aclgraph_utils.py | vllm-ascend #16409, 799801feef347469d5e9b39374b9210e0d4f7431. | Preserve the upstream revert while resolving the real rebase conflict; do not reintroduce the reverted graph-update behavior. |

| vllm_ascend/_310p/worker/v2/model_runner.py | vllm-ascend #16043, d4d2957e5208c2f464d4625c05920bd29ea233cb. | Preserve the newly merged 310P MTP copy tracking and remove only the v0.28.0 descriptor/allocation branch. |

Does this PR introduce any user-facing change?

Yes. The supported release changes from vLLM v0.28.0 to v0.29.0, and Ascend PCP+DP is enabled on v0.29.0. The fixed main commit remains supported. The vllm-ascend package version is unchanged.

Obsolete v0.28-only skips are removed where the contracts are now supported. No skip is migrated to v0.29.0, and no golden data or precision threshold is changed. Existing unrelated exclusions remain exclusions, not passes.

How was this patch tested?

Current-head validation: run 35420046256 completed successfully on fd876c5269fb6b39afa04e5a52433505e8786e22: 39 successful jobs and 6 skipped jobs, no failures.

• Pre-commit and mypy passed. Fixed-main CPU UT: 5078 passed, 67 skipped, 17 warnings. Actual vLLM checkout 84030bbe3d74d99bad477a3d2e37a973ccd8865c, installed 0.1.dev1+g84030bbe3.empty; Ascend checkout is the PR head. CPU base_sha is empty; the log says HEAD is up to date, so no unprinted integration SHA is inferred.
• All 32 NPU jobs passed, 16 per lane. Raw-log session totals per lane: 562 passed, 35 skipped, 1 xfailed (execution counts, not deduplicated cases). Actual checkouts: main 84030bbe3d74d99bad477a3d2e37a973ccd8865c, release 98dff2a81d747d1dba01a47f939f48c3526d4206; installed versions 0.1.dev1+g84030bbe3.empty and 0.29.0+empty. Each log confirms Ascend head fd876c5269fb6b39afa04e5a52433505e8786e22, integration base e139b7d573d3769fd1407d5027d7d4831f5469f0, and HEAD is up to date.
• PCP+DP and PCP+PP+MTP suite: main A3 four-card part4 and release A3 four-card part4 both passed. This does not establish a root cause for historical HCCL failures.
• Skipped and expected-failure cases are not passes. Current-head release CPU verification remains missing: the existing CPU job runs fixed main only. Historical release CPU results do not fill this gap. No workflow changes or additional skips were introduced.

Latest rebase: head fd876c5269fb6b39afa04e5a52433505e8786e22 onto upstream main e139b7d573d3769fd1407d5027d7d4831f5469f0. The old-head green results below are historical; new-head CI has completed successfully; see the verified results below. The fixed vLLM main and release pins remain unchanged. PR-wide Ruff, formatting and Python AST checks passed for 95 Python files; no workflow changes.

Rebase reconciliation:

• Preserve upstream #16853, 8f2e3fed73328148ea603ccfe5764542fa7cbbd2, without migrating its 0.28-only validator/dispatch gates to 0.29. Both supported pins already call the PCP manager for dispatch token counts. Restore the version-helper import needed by the inherited method. Its inherited 0.28-only skipped tests are not counted as passes.

• Adapt the existing unsupported-feature UT to preserve the upstream list: both supported pins delegate PCP checks to the manager via vLLM #53853, e376d45e82cb7e220da430e3179e81eb0922cf56; no obsolete 0.28 string filtering is reinstated.

• Adapt the existing KVPP allocation-entry UT from #15514, b64959a11ea69a51409f64e4e68da8e0924dec42, to the common allocation entry already used by both supported pins after vLLM #51718. Preserve cache-view assertions and the upstream dtype changes. No test functions or skips added.

• [Feature] Add DeepSeek V4.1 framework support and Engram host offload #16544 remains reverted by Revert "[Feature] Add DeepSeek V4.1 framework support and Engram host offload (#16544)" #16905; its V4.1 compatibility workaround remains removed.

Historical validation before this rebase:

• Current head 8c5d785931c495701bf1da8b5bc80b321ebd3cd0: run 35350119603 completed successfully: 40 successful jobs, 6 skipped jobs, no failures. Pre-commit and mypy passed. Local Ruff lint/format and AST checks passed for all 95 changed Python files.

• Actual main CPU: job 105617611031 reports 5074 passed, 37 skipped, 17 warnings. Logs confirm the exact PR head, vLLM checkout 84030bbe3d74d99bad477a3d2e37a973ccd8865c, and installed 0.1.dev1+g84030bbe3.empty. CPU base_sha is empty; no integration merge is inferred.

• Actual dual-version NPU: All 32 NPU job logs were audited, 16 per lane. Fixed main installed 0.1.dev1+g84030bbe3.empty; official release checkout 98dff2a81d747d1dba01a47f939f48c3526d4206 installed 0.29.0+empty. Each lane reports 562 passed, 35 skipped, 1 xfailed across pytest session summaries (execution counts, not deduplicated unique tests). Logs confirm checkout of this PR head and integration base 9dc6704559ebe2809b5cb7b7fc44184bd1338b3c; the rebase step reports HEAD is up to date.

• PCP combinations: The context-parallel suite containing PCP+DP and PCP+PP+MTP passes on both main A3 part4 and release A3 part4.

• Release CPU gap: Current-head actual release CPU remains missing because the workflow has only a main CPU entry. Version-parameterized UTs and historical release CPU runs do not replace actual release installation. Historical run 35213834751 at head 2560da9c7c2628f1b92b9305a1e80c5856bce74f passed 4923 tests with 37 skips per lane; its temporary CPU workflow was reverted.

• Backup: [Misc] Backup green v0.29.0 upgrade snapshot from #16393 #16898 remains unchanged with no test CI triggered. Local branch codex/backup-16393-before-16905 preserves previous green 7ad4c82ee; its results are not used as current-head validation.

Skipped/xfail cases and jobs are not counted as passes. Earlier intermittent HCCL errors have no proven root cause; no speculative fix is included. This run passes existing CI, but complete dual-version CPU validation remains outstanding.

• vLLM main: vllm-project/vllm@84030bb

[gemini-code-assist]

Summary of Changes

Hello, I'm Gemini Code Assist¹! I'm currently reviewing this pull request and will post my feedback shortly. In the meantime, here's a summary to help you and other reviewers quickly get up to speed!

This pull request updates the vLLM release dependency to v0.29.0 to align with upstream changes. It refactors internal Ascend-specific logic to support the new release contracts while preserving compatibility with the pinned main branch. The changes focus on removing legacy version-gated code paths and standardizing KV cache management and attention backend interfaces to match the v0.29.0 release.

Highlights

• vLLM Dependency Upgrade: Upgraded the supported vLLM release dependency from v0.28.0 to v0.29.0, while maintaining the existing vLLM main pin.
• Contract Retirement: Retired obsolete v0.28.0 compatibility branches and gates in favor of standardized v0.29.0 contracts for shared PCP imports, KV descriptors, and Mamba state handling.
• Test Adaptation: Updated existing test suites to align with v0.29.0 API changes, including the removal of version-specific skip boundaries and the standardization of KV cache layout assertions.

New Features

🧠 You can now enable Memory (public preview) to help Gemini Code Assist learn from your team's feedback. This makes future code reviews more consistent and personalized to your project's style. Click here to enable Memory in your admin console.

Using Gemini Code Assist

The full guide for Gemini Code Assist can be found on our documentation page, here are some quick tips.

Invoking Gemini

You can request assistance from Gemini at any point by creating a comment using either /gemini <command> or @gemini-code-assist <command>. Below is a summary of the supported commands on the current page.

Feature	Command	Description
Code Review	/gemini review	Performs a code review for the current pull request in its current state.
Pull Request Summary	/gemini summary	Provides a summary of the current pull request in its current state.
Comment	@gemini-code-assist	Responds in comments when explicitly tagged, both in pull request comments and review comments.
Help	/gemini help	Displays a list of available commands.

Customization

To customize the Gemini Code Assist for GitHub experience, repository maintainers can create a configuration file and/or provide a custom code review style guide (such as PEP-8 for Python) by creating and adding files to a .gemini/ folder in the base of the repository. Detailed instructions can be found here.

Limitations & Feedback

Gemini Code Assist may make mistakes. Please leave feedback on any instances where its feedback is incorrect or counterproductive. You can react with 👍 and 👎 on @gemini-code-assist comments. If you're interested in giving your feedback about your experience with Gemini Code Assist for GitHub and other Google products, sign up here.

Footnotes

1. Review the Privacy Notices, Generative AI Prohibited Use Policy, Terms of Service, and learn how to configure Gemini Code Assist in GitHub here. Gemini can make mistakes, so double check it and use code with caution. ↩

[gemini-code-assist]

Code Review

Suggested PR Title:

[CI][Misc] Bump verified vLLM version to v0.29.0 and remove v0.28.0 compatibility code

Suggested PR Summary:

### What this PR does / why we need it?
This PR bumps the verified vLLM version from `v0.28.0` to `v0.29.0` and cleans up legacy compatibility code, workarounds, and test skips that were specific to `v0.28.0`. It simplifies the codebase by removing version-gated branches (e.g., `vllm_version_is("0.28.0")`) across various modules including attention, schedulers, speculators, and model runners.

### Does this PR introduce _any_ user-facing change?
No.

### How was this patch tested?
Tested via existing unit tests and end-to-end tests updated to align with `v0.29.0` specifications.

I have reviewed the changes and have no additional feedback to provide.

[github-actions]

👋 Hi! Thank you for contributing to the vLLM Ascend project. The following points will speed up your PR merge:‌‌

• A PR should do only one thing, smaller PRs enable faster reviews.
• Every PR should include unit tests and end-to-end tests ‌to ensure it works and is not broken by other future PRs.
• Write the commit message by fulfilling the PR description to help reviewer and future developers understand.

If CI fails, you can run linting and testing checks locally according Contributing and Testing.

[github-actions]

This pull request has conflicts, please resolve those before we can evaluate the pull request.

[github-actions]

This pull request has conflicts, please resolve those before we can evaluate the pull request.

[github-actions]

This pull request has conflicts, please resolve those before we can evaluate the pull request.

[github-actions]

This pull request has conflicts, please resolve those before we can evaluate the pull request.