[Refactor] Drop placement values nothing reads

[ch-wan]

This PR is part of a stack (oldest at bottom):

• [Feature] Add --attn-dp-size and deprecate --enable-dp-attention #41818
• [Refactor] Enter draft TP scopes by attention ownership and drop ModelRunner.tp_group #41817
• [Refactor] Add make_pp_layers so models stop handing their PP position down #41816
• [Refactor] Stop passing models' layers the placement they already read #41815
• [Refactor] Let FusedMoE's weight-loading helpers read the layer's MoE-TP rank #41814
• [Refactor] Keep only the TP and PP groups on the model runner #41813
• [Refactor] Drop model placement attributes and parameters nothing reads #41812
• -> [Refactor] Drop placement values nothing reads #41811
• [Fix] Make the Solar model constructible and runnable #41810
• [Fix] Size the prefill delayer's gather buffer to its TP group #41809
• [Fix] Pass the draft's attention ownership to DFLASH's eager LiLiCorr scope #41808
• [Fix] Shard MoE WNA16 and Quark INT4-FP8 weights by the MoE placement #41807
• [Fix] State the configured MoE-DP width in the weight-cache fingerprint #41806
• [Fix] Read TensorCast's WORLD placement under its current names #41805

Motivation

A number of classes store or pass parallel ranks, sizes and groups that nothing reads. Each one reads as if it mattered, and each is one more copy of the placement to keep consistent.

Modifications

Remove the unread values:

• linear layers passed tp_size to load_merged_column_weight, which no parameter class reads; the W8A8 INT8 MoE method, the humming linear weight, the FlashMLA backend (dcp_rank), the DSA indexer (cp_size) and the Triton / Flash3 vision attention backends (tp_size) kept values with no reader;
• VocabParallelEmbedding.get_sharded_to_full_mapping has no caller;
• BaseRunner stored tp_size and attn_tp_rank; the decode and prefill graph runners re-stored attn_tp_size, attn_tp_rank and dp_size right after BaseRunner.__init__ set them; four draft graph runners copied model_runner.tp_size. None of these is read;
• the n-gram worker kept an unread tp_rank;
• SchedulerRequestReceiver carried attn_tp_group and attn_cp_group that it never reads (their CPU groups stay);
• PrefillBootstrapQueue took and stored gloo_group and tp_size, and DecodePreallocQueue stored tp_size and dp_size, without reading them.

No read value changes.

Accuracy Tests

Not applicable: nothing that is read changes.

• All 19 modified modules import.
• test/registered/unit at this PR's head, compared with main: no new failures.

Speed Tests and Profiling

Not applicable.

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.

---

CI States

Latest PR Test (Base): 🚫 Run #36772686757
Latest PR Test (Extra): 🚫 Run #36772686777
Latest PR Test (AMD ROCm 10): 🚫 Run #36772686680