[Refactor] Let FusedMoE's weight-loading helpers read the layer's MoE-TP rank

[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

The seven FusedMoE loading helpers (_load_w13, _load_w2, the scale and g_idx loaders, the GGUF and FP8 shared-expert paths) each take a tp_rank argument. Every call is inside FusedMoE, and every value traces back to tp_rank = self.moe_tp_rank in the two weight loaders.

Modifications

Read self.moe_tp_rank in the helpers, and drop the argument and the two locals.

A layer that shards differently stores that choice in self.moe_tp_rank itself: the Inkling shared experts and their loading helper keep the full-TP rank there. So the helpers see the same value as before.

Accuracy Tests

H200, compared with main at the top of this stack, which contains this PR; 4 greedy prompts × 64 tokens and GSM8K (200 questions):

model	config	greedy	GSM8K main / this stack
Qwen3-30B-A3B	--tp 2 --ep-size 2	identical	0.930 / 0.925
Qwen3-VL-30B-A3B-Instruct-AWQ (moe_wna16)	--tp 2	identical	0.945 / 0.945
Qwen3-VL-30B-A3B-Instruct-AWQ (moe_wna16)	--tp 2 --ep-size 2		fails to load on main / 0.935 (the load fix is #41807, below in this stack)

• unit/models, unit/layers/moe, unit/layers/quantization and test_runtime_context: the same results as the parent commit.
• 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 #36772813010
Latest PR Test (Extra): 🚫 Run #36772812682
Latest PR Test (AMD ROCm 10): ❌ Run #36772813054