nemo-mbridge-perf-moe-comm-overlap

bởi nvidia

Tương tác giao tiếp song song chuyên gia MoE trong Megatron Bridge. Bao gồm tối ưu hóa dispatch/combine, các backend flex dispatcher, và lập lịch expert wgrad.

npx skills add https://github.com/nvidia/skills --skill nemo-mbridge-perf-moe-comm-overlap

MoE Communication Overlap

For the higher-level overview, see:

  • @docs/training/communication-overlap.md
  • @skills/nemo-mbridge-perf-moe-comm-overlap/card.yaml

Quick Decision

Use MoE communication overlap when:

  • EP > 1
  • token dispatch or combine time is visible in the profile
  • the run is already correct and you are now tuning throughput

Avoid turning it on as an early bring-up step. It is easier to validate after the dispatcher, routing mode, and recompute plan are already stable.

Enablement

cfg.comm_overlap.overlap_moe_expert_parallel_comm = True

# Optional: delayed wgrad for additional overlap
cfg.comm_overlap.delay_wgrad_compute = True

# IMPORTANT: disable shared expert overlap when using dispatch overlap
cfg.model.moe_shared_expert_overlap = False

Prerequisites

  • expert_model_parallel_size > 1
  • num_moe_experts > 1
  • moe_token_dispatcher_type must be "alltoall" or "flex"
  • Precision: BF16 or FP16
  • If PP is used, VPP (virtual_pipeline_model_parallel_size) must be set (non-None)

Flex dispatcher activation

Setting moe_flex_dispatcher_backend alone does not activate flex dispatch. You must also set moe_token_dispatcher_type = "flex".

Recompute And CUDA Graph Interaction

  • Full recompute is not a good companion for the overlap path.
  • delay_wgrad_compute adds further constraints if CUDA-graph scopes include attention or MoE-router work.
  • In practice, selective recompute is the safer pairing when overlap is enabled.

Measured Evidence

HybridEP production-shape validation

A 2026-07-25 controlled Qwen3 30B-A3B pretraining comparison used 16 H100 GPUs, BF16, sequence length 4096, TP=1, PP=1, CP=1, EP=16, MBS=1, GBS=1024, forced-balanced routing, HybridEP, and Transformer Engine CUDA-graph scopes moe_router and moe_preprocess. The only performance change was plain EP overlap; delayed wgrad stayed disabled.

CaseSteady windowStep timeModel TFLOPS/GPU
EP overlap offiterations 5-2024.7138s244.039
EP overlap on, search runiterations 5-2021.0725s286.208
EP overlap on, independent validationiterations 41-5020.9920s287.305

The independent result reduced step time by 15.059% and increased throughput by 17.729% over the reproduced baseline. Loss remained finite, no iterations were skipped or NaN, and rank-0 peak allocated memory was 62.166 GiB.

A same-method rank-0 Nsight Systems comparison captured 463,348 kernels in each case:

Profile metricOverlap offOverlap on
Communication concurrent with GEMM/attention9.079ms3,958.997ms
Communication time hidden by compute0.11%36.55%
GPU-active interval union22.821s21.221s
HybridEP dispatch-with-permute NVTX4.253s1.767s
HybridEP metadata-preprocess NVTX3.109s0.670s

This is direct evidence that the gain came from hiding exposed HybridEP dispatch/combine work, not from changing the dispatcher, routing, graph scopes, batch shape, or parallel layout.

Correctness-first alltoall smoke

A 2026-05-18 current-main H100 x16 smoke on Qwen3 30B-A3B mock pretraining used EP=16, alltoall, global batch size 1024, CUDA graphs disabled, and moe_permute_fusion=false because the PyTorch 25.11 / TE / Triton stack failed in Transformer Engine fused permutation in prior bring-up.

Results were directional rather than release-grade:

  • no EP overlap: 41.25s steady-state mean over iterations 3-8
  • EP overlap: 31.31s steady-state mean over iterations 3-8
  • EP overlap plus delay_wgrad_compute: 31.20s steady-state mean over iterations 3-8

Treat this as evidence that EP overlap can help an inter-node alltoall MoE shape when communication is exposed. It is not proof that delayed wgrad is a separate win, and it does not validate the fused permutation path. An earlier 2026-05-16 short smoke on the same shape showed the same pattern.

Code Anchors

  • Overlap validation: src/megatron/bridge/training/comm_overlap.py
  • Flex dispatcher backend: src/megatron/bridge/training/flex_dispatcher_backend.py
  • Config: src/megatron/bridge/training/config.py
  • Unit tests: tests/unit_tests/training/test_comm_overlap.py
  • DeepEP tests: tests/unit_tests/training/test_deepep.py

Pitfalls

  1. Shared expert overlap conflict: moe_shared_expert_overlap and overlap_moe_expert_parallel_comm can conflict. Disable shared expert overlap when using the dispatch overlap path.

  2. PP without VPP: MoE overlap requires VPP when pipeline parallelism is active. Without it, the overlap scheduling cannot interleave correctly.

  3. Flex != backend flag: moe_flex_dispatcher_backend="deepep" alone does nothing if moe_token_dispatcher_type is still "alltoall".

  4. Conservative recipe defaults: Most public recipes leave MoE overlap disabled. You need to explicitly enable it via overrides.

  5. Performance gains are workload-dependent: overlap helps most when dispatch communication is already a visible slice of step time. It is not guaranteed to help every small or lightly loaded EP run.

  6. Summed kernel time is not wall time: concurrent kernels can run longer because they contend for SMs or bandwidth, so overlap may increase summed per-stream kernel duration while reducing the exposed interval union and end-to-end step time.

Verification

Look for overlap-related log messages during initialization. The comm overlap validation in comm_overlap.py will raise if prerequisites are not met, so a clean startup confirms the feature is active.

For a short performance-harness smoke, keep the command shape explicit and vary only one overlap knob at a time:

uv run python scripts/performance/run_script.py \
  -m qwen \
  -mr qwen3_30b_a3b \
  --task pretrain \
  -g h100 \
  -c bf16 \
  -ng 16 \
  -gn 8 \
  --max_steps 8 \
  --cuda_graph_impl none \
  --moe_flex_dispatcher_backend None \
  --moe_a2a_overlap false \
  --tokenizer_type NullTokenizer \
  comm_overlap.overlap_moe_expert_parallel_comm=true \
  comm_overlap.delay_wgrad_compute=false \
  model.moe_shared_expert_overlap=false

If fused MoE permutation fails during bring-up, add model.moe_permute_fusion=false to separate overlap timing from runtime-stack validation, then retest with the matched production container.

For performance validation, use an unprofiled steady window as the acceptance metric. Use a matched Nsight A/B to establish causality:

  1. Keep dispatcher, routing, CUDA graphs, batch shape, parallelism, and runtime fixed.
  2. Toggle only overlap_moe_expert_parallel_comm; keep delay_wgrad_compute=false for the first isolation.
  3. Compare communication and compute interval unions and their intersection, not only summed kernel durations.
  4. Report steady step time, model TFLOPS/GPU, loss finiteness, skipped/NaN iterations, and peak allocated memory.

Last signature refresh: 2026-08-03.

Thêm skills từ nvidia

compileiq-debug
nvidia
Sử dụng khi có điều gì đó không ổn: Search() bị treo, tất cả các đánh giá đều trả về INVALID_SCORE, điểm số không cải thiện, mọi cấu hình đều trả về cùng một số, lỗi ptxas…
create-github-pr
nvidia
Tạo pull request GitHub bằng cách sử dụng gh CLI. Sử dụng khi người dùng muốn tạo PR mới, gửi mã để xem xét, hoặc mở pull request. Từ khóa kích hoạt -…
nemoclaw-maintainer-cross-issue-sweep
nvidia
Quét các vấn đề đang mở khác để tìm những vấn đề mà một PR nhất định có thể sửa hoặc vô tình làm hỏng. Đưa ra các cơ hội sửa lỗi liền kề và rủi ro mâu thuẫn với file:dòng…
fhir-basics
nvidia
Dạy các tác nhân cách hoạt động của API FHIR R4, những tài nguyên có sẵn, cách truy vấn chúng với tham số tìm kiếm, và cách phân tích chính xác tất cả các định dạng phản hồi…
compileiq-validate-result
nvidia
Sử dụng SAU KHI tìm kiếm hoàn tất và TRƯỚC KHI yêu cầu tăng tốc hoặc gửi ACF. Tải tệp CSV dump_results, trích xuất các ứng viên top-K (đơn mục tiêu)…
changelog-audit
nvidia
Kiểm tra Warp CHANGELOG.md trước khi phát hành: khôi phục các mục bị mất, sắp xếp theo tác động người dùng, tinh chỉnh ngôn ngữ mục, xuống dòng và (chế độ nhánh phát hành) so sánh bump…
maintain-dynamic-plugins
nvidia
Duy trì các bộ nạp plugin động NeMo Relay, tệp kê khai, SDK gốc Rust, giao thức worker gRPC, SDK worker Python, tài liệu, kiểm thử và phạm vi quy trình phát hành
dgx-diagnose
nvidia
Chẩn đoán các sự cố thường gặp của DGX Station GB300 — lỗi CUDA, nhắm sai GPU, lỗi container vLLM/SGLang, vấn đề trạng thái MIG, lỗi NVLink/Fabric Manager,…