nemo-mbridge-perf-hierarchical-context-parallel

от nvidia

Эксплуатационное руководство по включению иерархического параллелизма контекста в Megatron-Bridge, включая параметры конфигурации, привязки к коду, подводные камни и проверку.

npx skills add https://github.com/nvidia/skills --skill nemo-mbridge-perf-hierarchical-context-parallel

Hierarchical Context Parallel Skill

This skill covers hierarchical context parallelism: nested context-parallel process groups used by cp_comm_type="a2a+p2p" and configured with hierarchical_context_parallel_sizes.

For what hierarchical CP is, when to use it, and the decision tree (a2a+p2p vs pure a2a vs p2p), see:

  • @docs/training/hierarchical-context-parallel.md
  • @skills/nemo-mbridge-perf-hierarchical-context-parallel/card.yaml

Enablement

Minimal Bridge override:

cfg.model.context_parallel_size = 4
cfg.model.cp_comm_type = "a2a+p2p"
cfg.model.hierarchical_context_parallel_sizes = [2, 2]
cfg.dist.use_decentralized_pg = False

Required constraints:

  • prod(hierarchical_context_parallel_sizes) == context_parallel_size
  • seq_length % (2 * context_parallel_size) == 0
  • Transformer Engine >= 1.12.0

Code Anchors

Upstream config and validation:

context_parallel_size: int = 1
"""Splits network input along sequence dimension across GPU ranks."""

hierarchical_context_parallel_sizes: Optional[list[int]] = None
"""Degrees of the hierarchical context parallelism. Users should provide a list to specify 
   the sizes for different levels. Taking the a2a+p2p cp comm type as example, it contains
   groups of two levels, so the first value of the list indicates the group size of the a2a
   communication type, and the second value indicates the group size of the p2p communication
   type.
"""
if args.hierarchical_context_parallel_sizes:
    from numpy import prod
    assert args.context_parallel_size == prod(args.hierarchical_context_parallel_sizes)
if "a2a+p2p" in args.cp_comm_type:
    assert args.hierarchical_context_parallel_sizes is not None, \
    "--hierarchical-context-parallel-sizes must be set when a2a+p2p is used in cp comm"

Bridge MPU path:

parallel_state.initialize_model_parallel(
    ...
    context_parallel_size=model_config.context_parallel_size,
    hierarchical_context_parallel_sizes=model_config.hierarchical_context_parallel_sizes,
    ...
)
...
return ProcessGroupCollection.use_mpu_process_groups()

Bridge decentralized-PG path:

pg_collection = ProcessGroupCollection(
    ...
    cp=cp_pg,
    tp_cp=tp_cp_pg,
    hcp=None,
    ep=ep_pg,
    ...
)

Implementation Map

The code anchors above show the config declarations and argument validation.

Validation (MCore)

TransformerConfig.__post_init__ enforces that a2a+p2p requires HCP sizes and the product matches CP.

Process group creation

parallel_state.initialize_model_parallel creates hierarchical CP sub-groups when HCP sizes are provided via create_hierarchical_groups. Bridge currently gets those groups through the MPU-backed ProcessGroupCollection.

TE integration

TEDotProductAttention passes the hierarchical groups to Transformer Engine when a2a+p2p is used. Requires Transformer Engine >= 1.12.0.

Pitfalls

  1. Bridge HCP is MPU-only today: If use_decentralized_pg=True, Bridge initializes flat CP groups and leaves HCP unset.
  2. No checked-in Bridge recipe currently exercises HCP directly.
  3. Single-GPU load helpers clear hierarchical_context_parallel_sizes.
  4. Silent broken training on old stacks: If you use a2a+p2p without setting hierarchical_context_parallel_sizes, MCore now asserts. Older versions would silently disable CP communication, so each rank attended only to its local chunk and produced artificially high throughput with broken gradients.
  5. Product must match: prod(hierarchical_context_parallel_sizes) must exactly equal context_parallel_size. A mismatch triggers an assertion.
  6. Verify in logs: Look for the process group initialization output. You should see HIERARCHICAL_CONTEXT_PARALLEL_GROUPS being created. If you only see CONTEXT_PARALLEL_GROUP, HCP is not active.

Verification

No dedicated Bridge end-to-end test exists yet for HCP (see @skills/nemo-mbridge-perf-hierarchical-context-parallel/card.yaml follow_up_validation). Use the existing unit tests and log inspection instead.

Run the decentralized-PG unit test to confirm the flat-CP behavior is preserved:

uv run python -m pytest tests/unit_tests/training/test_decentralized_pg.py -q

For a manual smoke check, launch a 4-GPU run with a small recipe and cp_comm_type=a2a+p2p plus hierarchical_context_parallel_sizes=[2,2]:

CUDA_VISIBLE_DEVICES=0,1,2,3 uv run python -m torch.distributed.run --nproc_per_node=4 \
  scripts/training/run_recipe.py \
  --recipe llama32_1b_pretrain_config \
  model.context_parallel_size=4 \
  model.cp_comm_type=a2a+p2p \
  "model.hierarchical_context_parallel_sizes=[2,2]" \
  train.train_iters=2

Success criteria:

  • Logs show HIERARCHICAL_CONTEXT_PARALLEL_GROUPS being created
  • Training completes at least one step without error
  • If you only see CONTEXT_PARALLEL_GROUP, HCP is not active

Больше skills от nvidia

compileiq-debug
nvidia
Используйте, когда что-то не так: Search() зависает, все оценки возвращают INVALID_SCORE, оценки не улучшаются, каждая конфигурация возвращает одно и то же число, ошибки ptxas…
create-github-pr
nvidia
Создание pull request'ов в GitHub с помощью gh CLI. Используйте, когда пользователь хочет создать новый PR, отправить код на ревью или открыть pull request. Ключевые слова для запуска —…
nemoclaw-maintainer-cross-issue-sweep
nvidia
Сканирует другие открытые задачи, чтобы найти те, которые данный PR может исправить или случайно сломать. Выводит возможности смежных исправлений и риски противоречий с указанием файла:строки…
fhir-basics
nvidia
Обучает агентов работе с API FHIR R4, доступным ресурсам, запросам с параметрами поиска и корректному разбору всех форматов ответов…
compileiq-validate-result
nvidia
Используйте ПОСЛЕ завершения поиска и ДО применения ускорения или отправки ACF. Загружает CSV-файл dump_results, извлекает top-K кандидатов (однокритериальный)...
changelog-audit
nvidia
Аудит Warp CHANGELOG.md перед релизом: восстановление потерянных записей, сортировка по влиянию на пользователей, уточнение формулировок, перенос строк и (в режиме релизной ветки) обновление сравнения…
maintain-dynamic-plugins
nvidia
Поддержка загрузчиков динамических плагинов NeMo Relay, манифестов, нативных Rust SDK, протокола gRPC worker, Python worker SDK, документации, тестов и покрытия рабочего процесса релиза
dgx-diagnose
nvidia
Диагностика распространённых проблем DGX Station GB300 — сбои CUDA, ошибочное нацеливание на GPU, ошибки контейнеров vLLM/SGLang, проблемы состояния MIG, ошибки NVLink/Fabric Manager,…