Loading skill
Install any skill in seconds. Free to start, no credit card required.
Get Started Free →nvidia/nemo-mbridge-perf-moe-comm-overlap
MoE expert-parallel communication overlap in Megatron Bridge. Covers dispatch/combine overlap, flex dispatcher backends, and expert wgrad scheduling.
Passed 17 of 21 test cases with the skill, and 9 of 21 without it.
and 15 more in the full report
11 fixed7 passed either way2 still failing2 broke
Tokens, turns and time describe resource use, not a dollar price.
One of 10 test cases the skill fixed. Same model, same prompt, with and without the skill.
I am running a 16-GPU performance smoke test script (`scripts/performance/run_script.py`) for a Qwen MoE model in Megatron-Bridge to verify EP communication overlap. To prevent issues with Transformer Engine fused permutation during initial bring-up, which specific flag override should I pass on the command line?
| Checked | Without skill | With skill |
|---|---|---|
| model.moe_permute_fusion=false is specified as a command-line override. | Missed | Passed |
| Explains that disabling fused permutation isolates overlap timing from runtime stack failures. | Passed | Passed |
The instructions your model reads when the skill loads.
For the higher-level overview, see:
Use MoE communication overlap when:
EP > 1Avoid 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.
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
expert_model_parallel_size > 1num_moe_experts > 1moe_token_dispatcher_type must be "alltoall" or "flex"virtual_pipeline_model_parallel_size) must be set (non-None)test cases passed, vs 9 / 21 without
Published skills whose descriptions closely match this one, with their headline benchmark lift.
More skills from the same source and nearby in the registry, with their headline benchmark lift.