Unverified50% confidenceSolutionExact time
ZGCM-1 uses a stable FP8 Muon optimizer, combining the Muon optimizer with FP8 low-precision training while maintaining training stability.
1
Sources
50%
Confidence
Long-term
Relevance
9/15/2026
First Seen
Sources
Related Entities
Related Claims
UnverifiedMUON优化器替代了沿用多年的AdamW优化器，通过引入矩阵正交化步骤使训练曲线更加平滑66% similarUnverifiedWith FP8 mixed precision, expert feed-forward computation in MoE can maintain training stability while reducing memory footprint and bandwidth pressure.64% similarUnverifiedZGCM-1 publicly releases model weights, intermediate checkpoints, complete training code, data and data recipes, and W&B training logs for the pretraining, mid-training, and post-training stages.62% similarUnverifiedThe paper states that ZGCM-1 achieves approximately a 4.2× improvement in time-to-loss efficiency over the baseline during 16K-context pretraining.61% similarUnverified深度学习训练的梯度同步瓶颈使得低耦合任务（如批量推理、超参数搜索、渲染）更适合P2P架构，而分布式训练不适合60% similar
Cite This Claim
Stable URI
https://kongchang.com/claim/921653API
curl https://kongchang.com/api/v1/knowledge/claims/921653MCP
get_claim(id=921653)