[DeepSeek-V4][fix] Detect MTP routed-expert layout separately

[waynehacking8]

Description

#16433 merged the base mixed-precision construction and loading fix for nvidia/DeepSeek-V4-Pro-NVFP4, superseding the earlier construction commits in this PR. This PR now contains only the remaining MTP-specific fix, authored by @d3nb.

The checkpoint stores dense routed experts as NVFP4 (U8) but leaves MTP routed experts at the base model's MXFP4 layout (I8). _set_deepseek_v4_routed_moe_quant_config() previously detected one layout from layers.0 and assigned it to every MoE layer, so MTP speculative decoding received NVFP4 configuration and crashed in fused_moe scale loading.

The loader now probes the MTP expert header separately and assigns MXFP4 or NVFP4 configuration to the MTP layer range only when its layout differs from the dense experts. Matching or absent MTP headers retain the existing dense configuration.

Related: #16196, #16433.

Test Coverage

• @d3nb validated the equivalent change on the actual 851 GB nvidia/DeepSeek-V4-Pro-NVFP4 checkpoint using 4x B300, TP4, and the TRT-LLM MoE backend.
• Weights loaded completely, the server generated successfully with MTP=1 and MTP=3, and observed MTP acceptance length was approximately 2.86.
• The original commit and authorship are preserved after rebasing onto current main.

PR Checklist

• Scope reflects the remaining non-superseded fix
• Commit is signed off
• Validated on real weights and hardware

[coderabbitai]

Walkthrough

DeepSeek V4 mixed-precision handling now derives FP8 block-scale configurations, dequantizes fused-A weights for unquantized destinations, and applies consistent quantization settings to attention, shared-expert, and MTP projections.

Changes

DeepSeek V4 mixed-precision quantization

Layer / File(s)	Summary
Mixed-precision configuration resolution tensorrt_llm/_torch/models/modeling_deepseekv4.py	Adds helpers for FP8 block-scale configuration and updates per-layer expert quantization resolution.
Fused-A weight loading tensorrt_llm/_torch/models/modeling_deepseekv4.py	Dequantizes fused-A checkpoint weights when the destination module lacks weight_scale.
Projection quantization wiring tensorrt_llm/_torch/models/modeling_deepseekv4.py	Applies derived mixed-precision configuration to attention, shared-expert, and MTP projections.

Estimated code review effort: 4 (Complex) | ~45 minutes

🚥 Pre-merge checks | ✅ 4 | ❌ 1

❌ Failed checks (1 warning)

Check name	Status	Explanation	Resolution
Docstring Coverage	⚠️ Warning	Docstring coverage is 41.67% which is insufficient. The required threshold is 80.00%.	Write docstrings for the functions missing them to satisfy the coverage threshold.

✅ Passed checks (4 passed)

Check name	Status	Explanation
Linked Issues check	✅ Passed	The changes match #16196 by fixing excluded-module FP8 loading, per-layer NVFP4 selection, and checkpoint-native expert key handling.
Out of Scope Changes check	✅ Passed	The modified loader and quantization paths all support the DeepSeek-V4 NVFP4 loading fix; no unrelated code changes are evident.
Title check	✅ Passed	The title is concise and accurately describes the remaining fix: detecting MTP routed-expert layout separately.
Description check	✅ Passed	The description follows the template well, with clear problem/solution, test coverage, and checklist notes.

✨ Finishing Touches

🧪 Generate unit tests (beta)

• Create PR with unit tests

---

_{Comment @coderabbitai help to get the list of available commands.}

[waynehacking8]

@mikeiovine I rebased onto current main after #16433 and narrowed the PR to the remaining MTP layout commit; it is mergeable and DCO is green. When convenient, could you trigger /bot run for the rebuilt head?