[dsv4] support eplb

[SYChen123]

Motivation

Support EPLB for deepseek v4 (with megaMoE disabled).

Modifications

Add the eplb context in the forward of DeepseekV4Model. Without the context, the layer_idx used in eplb distribution gatherer is None which causes crash in prefill node and stats inaccuracy issue in decode node.

Accuracy Tests

Not relevant.

Speed Tests and Profiling

With this pr, eplb works normally on both prefill and decode.

Server launch command
SGLANG_OPT_SWA_SPLIT_LEAF_ON_INSERT=1 SGLANG_DEEPEP_NUM_MAX_DISPATCH_TOKENS_PER_RANK=128 python3 -m sglang.launch_server --trust-remote-code --model-path ./models/deepseek-ai/DeepSeek-V4-Pro/ --disable-cuda-graph --tp 8 --dp 8 --enable-dp-attention --enable-dp-lm-head --moe-a2a-backend deepep --chunked-prefill-size 32768 --enable-prefill-delayer --swa-full-tokens-ratio 0.1 --mem-fraction-static 0.87 --max-running-requests 512 --deepep-mode normal --disaggregation-mode prefill --disaggregation-transfer-backend mooncake --disaggregation-ib-device mlx5_4 --dist-init-addr 192.168.184.178:5757 --nnodes 1 --node-rank 0 --host 0.0.0.0 --port 9091 --enable-expert-distribution-metrics --enable-eplb --eplb-min-rebalancing-utilization-threshold 0.6 --eplb-rebalance-num-iterations 5000 --ep-num-redundant-experts 32 --ep-dispatch-algorithm dynamic --eplb-algorithm deepseek --init-expert-location ./expert_distribution_recorder.pt

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 #26363979763
Latest PR Test (Extra): ❌ Run #26363979729

[gemini-code-assist]

Warning

You have reached your daily quota limit. Please wait up to 24 hours and I will start processing your requests again!

[fzyzcjy]

general direction LGTM, cc @ch-wan for this assigned bug to double check