Skip to content

[perf,recipe] Fix Qwen3 30B A3B B200 perf config to use hybridep+flex dispatcher #2499

New issue

Have a question about this project? Sign up for a free GitHub account to open an issue and contact its maintainers and the community.

Sign up for GitHub

By clicking “Sign up for GitHub”, you agree to our terms of service and privacy statement. We’ll occasionally send you account related emails.

Already on GitHub? Sign in to your account

Merged
malay-nagda merged 1 commit into from
Feb 27, 2026
Merged

[perf,recipe] Fix Qwen3 30B A3B B200 perf config to use hybridep+flex dispatcher #2499

Show file tree
Hide file tree
Changes from all commits
Commits
Show all changes
1 commit
Select commit
File filter

Filter by extension

Filter by extension
Conversations
Failed to load comments.
Loading
Jump to
Jump to file
Failed to load files.
Loading
Diff view
Diff view
2 changes: 1 addition & 1 deletion scripts/performance/configs/qwen/qwen3_llm_pretrain.py
View file
Open in desktop
Original file line number Diff line number Diff line change
Expand Up @@ -286,7 +286,7 @@ def qwen3_30b_a3b_pretrain_config_b200(
cfg.mixed_precision = precision_config
cfg.comm_overlap = CommOverlapConfig(tp_comm_overlap=True)
cfg.model.moe_flex_dispatcher_backend = base_cfg.moe_flex_dispatcher_backend
cfg.model.moe_token_dispatcher_type = "alltoall"
cfg.model.moe_token_dispatcher_type = "flex"

set_qwen3_common_configs(cfg)
set_workload_base_configs(cfg, base_cfg)
Expand Down
13 changes: 5 additions & 8 deletions scripts/performance/configs/qwen/qwen3_workload_base_configs.py
View file
Open in desktop
Original file line number Diff line number Diff line change
Expand Up @@ -391,20 +391,17 @@
QWEN3_30B_A3B_PRETRAIN_CONFIG_B200_BF16_V1 = replace(
BASE_QWEN3_30B_A3B_CONFIG,
num_gpus=8,
micro_batch_size=4,
moe_flex_dispatcher_backend="hybridep",
cuda_graph_impl="transformer_engine",
cuda_graph_scope=["moe_router", "moe_preprocess"],
cuda_graph_scope=["attn", "moe_router", "moe_preprocess"],
)


QWEN3_30B_A3B_PRETRAIN_CONFIG_B200_FP8_CS_V1 = replace(
BASE_QWEN3_30B_A3B_CONFIG,
num_gpus=8,
cuda_graph_impl="transformer_engine",
cuda_graph_scope=["moe_router", "moe_preprocess"],
)
QWEN3_30B_A3B_PRETRAIN_CONFIG_B200_FP8_CS_V1 = QWEN3_30B_A3B_PRETRAIN_CONFIG_B200_BF16_V1


QWEN3_30B_A3B_PRETRAIN_CONFIG_B200_FP8_MX_V1 = QWEN3_30B_A3B_PRETRAIN_CONFIG_B200_FP8_CS_V1
QWEN3_30B_A3B_PRETRAIN_CONFIG_B200_FP8_MX_V1 = QWEN3_30B_A3B_PRETRAIN_CONFIG_B200_BF16_V1


QWEN3_30B_A3B_PRETRAIN_CONFIG_H100_BF16_V1 = replace(
Expand Down
Loading