[Refactor] Drop model placement attributes and parameters 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

Model modules copy parallel ranks and sizes onto self (self.tp_size = get_parallel().tp_size and similar). In 97 places no method of the class, its subclasses or its bases reads the attribute back, and no other code reads it off an instance. A few constructors also accept placement parameters that no caller sets.

Modifications

• Remove the 97 unread assignments across 53 model files. The Dots Note omni wrapper's tp_size, copied from its language model and itself never read, goes with the language model's. Three TODOs that only annotated a removed assignment go with it.
• Step3p5MLP accepted tp_size / tp_rank only to forward them to its linear layers, and neither caller passes them, so the layers always got their own defaults. OlmoeMoE, Grok1MoE and MixtralMoE accepted a tp_size they never use. Remove these parameters; every caller already passes by keyword without them.

Code outside the tree that read one of the removed attributes, for example model.tp_size, should read get_parallel() instead.

Accuracy Tests

Not applicable: nothing that is read changes.

• All 53 modified model files 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 #36772717146
Latest PR Test (Extra): 🚫 Run #36772716531
Latest PR Test (AMD ROCm 10): ❌ Run #36772717248