specific method · filed under optimization
AdamW for the MoE router
Using AdamW rather than Muon for the MoE router to avoid early-training fluctuations and instability.
Also called we use AdamW for the router.
- source
- 1
- model
- 1
- lab adopt it
- 1
- strongest
- core
How sources treat it
One count per evidence span, weakest treatment to strongest.
core 1
Documented in
Evidence
1 span quoted from the sources, strongest treatment first.
Muon exacerbates early-training fluctuations and destabilizes the router. ... Therefore, we use AdamW for the router.
coreoptimizationin Qwen3.8-NextQwen
Filed alongside
Other methods under optimization :: optimizer.
MuonPer-Head MuonAdamWCategory-specific assignment of Muon and AdamWSplit fused gradients before orthogonalizationNesterov momentumSinkhorn-balanced updateAdam without weight decay for the N-gram embedding tableAdamW for attention and GDN output gatesAdamW for gated-residual low-rank projectionsAdamW for input embeddings and output headAsynchronous Micro-Group pipelineCanzonaCUDA graph capture of the optimizer stepEight-step Newton–Schulz iterationHybrid Muon and AdamW parameter-group optimizer assignmentHybrid Newton–Schulz iterationsHybrid optimization with Muon and AdamMoonlight-style learning-rate scalingMuon orthogonalization accuracy refinementMuon restricted to two-dimensional linear-map weightsMuon SplitMuownPeer-to-peer shard retrieval for Muon orthogonalization