Cover Image

MOSS: Efficient and Accurate FP8 LLM Training with Microscaling and Automatic Scaling

Fuente: arXiv

Saved in:

Bibliographic Details
Main Authors:	Zhang, Yu, Zhen, Hui-Ling, Yuan, Mingxuan, Yu, Bei
Format:	Preprint
Published:	2025
Subjects:	Machine Learning Artificial Intelligence
Online Access:	Acceder al recurso
Tags:	Add Tag No Tags, Be the first to tag this record!

Similar Items

MixPE: Quantization and Hardware Co-design for Efficient LLM Inference
by: Zhang, Yu, et al.
Published: (2024)

SageAttention3: Microscaling FP4 Attention for Inference and An Exploration of 8-Bit Training
by: Zhang, Jintao, et al.
Published: (2025)

From Pruning to Grafting: Dynamic Knowledge Redistribution via Learnable Layer Fusion
by: Pei, Zehua, et al.
Published: (2024)

TrimR: Verifier-based Training-Free Thinking Compression for Efficient Test-Time Scaling
by: Lin, Weizhe, et al.
Published: (2025)

Analytical FFN-to-MoE Restructuring via Activation Pattern Analysis
by: Pei, Zehua, et al.
Published: (2025)

The Curse and Blessing of Mean Bias in FP4-Quantized LLM Training
by: Cao, Hengjie, et al.
Published: (2026)

COAT: Compressing Optimizer states and Activation for Memory-Efficient FP8 Training
by: Xi, Haocheng, et al.
Published: (2024)

To FP8 and Back Again: Quantifying Reduced Precision Effects on LLM Training Stability
by: Lee, Joonhyung, et al.
Published: (2024)

Post Training Quantization of Large Language Models with Microscaling Formats
by: Sharify, Sayeh, et al.
Published: (2024)

Scaling FP8 training to trillion-token LLMs
by: Fishman, Maxim, et al.
Published: (2024)

Certifying Language Model Robustness with Fuzzed Randomized Smoothing: An Efficient Defense Against Backdoor Attacks
by: He, Bowei, et al.
Published: (2025)

SVDq: 1.25-bit and 410x Key Cache Compression for LLM Attention
by: Yankun, Hong, et al.
Published: (2025)

MicroMix: Efficient Mixed-Precision Quantization with Microscaling Formats for Large Language Models
by: Liu, Wenyuan, et al.
Published: (2025)

Practical FP4 Training for Large-Scale MoE Models on Hopper GPUs
by: Zhang, Wuyue, et al.
Published: (2026)

FP6-LLM: Efficiently Serving Large Language Models Through FP6-Centric Algorithm-System Co-Design
by: Xia, Haojun, et al.
Published: (2024)

Efficient Post-training Quantization with FP8 Formats
by: Shen, Haihao, et al.
Published: (2023)

FP8-Flow-MoE: A Casting-Free FP8 Recipe without Double Quantization Error
by: Wang, Fengjuan, et al.
Published: (2025)

KVTuner: Sensitivity-Aware Layer-Wise Mixed-Precision KV Cache Quantization for Efficient and Nearly Lossless LLM Inference
by: Li, Xing, et al.
Published: (2025)

Efficient Training of Large-Scale AI Models Through Federated Mixture-of-Experts: A System-Level Approach
by: Chen, Xiaobing, et al.
Published: (2025)

Automatic Reward Shaping from Confounded Offline Data
by: Li, Mingxuan, et al.
Published: (2025)

GNNavigator: Towards Adaptive Training of Graph Neural Networks via Automatic Guideline Exploration
by: Qiao, Tong, et al.
Published: (2024)

FP4 All the Way: Fully Quantized Training of LLMs
by: Chmiel, Brian, et al.
Published: (2025)

Beyond Speedup -- Utilizing KV Cache for Sampling and Reasoning
by: Xing, Zeyu, et al.
Published: (2026)

Rel-MOSS: Towards Imbalanced Relational Deep Learning on Relational Databases
by: Yin, Jun, et al.
Published: (2026)

Automatic Configuration of LLM Post-Training Pipelines
by: Chwa, Channe, et al.
Published: (2026)

Learning Multi-graph Structure for Temporal Knowledge Graph Reasoning
by: Zhang, Jinchuan, et al.
Published: (2023)

StreamFP: Learnable Fingerprint-guided Data Selection for Efficient Stream Learning
by: Shi, Tongjun, et al.
Published: (2024)

FP4 Explore, BF16 Train: Diffusion Reinforcement Learning via Efficient Rollout Scaling
by: Li, Yitong, et al.
Published: (2026)

Scaling with Collapse: Efficient and Predictable Training of LLM Families
by: Bergsma, Shane, et al.
Published: (2025)

One Train for Two Tasks: An Encrypted Traffic Classification Framework Using Supervised Contrastive Learning
by: Zhang, Haozhen, et al.
Published: (2024)

Efficient Network Automatic Relevance Determination
by: Zhang, Hongwei, et al.
Published: (2025)

The Graph's Apprentice: Teaching an LLM Low Level Knowledge for Circuit Quality Estimation
by: Moravej, Reza, et al.
Published: (2024)

Defeating the Training-Inference Mismatch via FP16
by: Qi, Penghui, et al.
Published: (2025)

Precision-Scalable Microscaling Datapaths with Optimized Reduction Tree for Efficient NPU Integration
by: Cuyckens, Stef, et al.
Published: (2025)

Memory-Efficient LLM Training with Dynamic Sparsity: From Stability to Practical Scaling
by: Xiao, Qiao, et al.
Published: (2026)

GraSS: Combining Graph Neural Networks with Expert Knowledge for SAT Solver Selection
by: Zhang, Zhanguang, et al.
Published: (2024)

DiLA: Enhancing LLM Tool Learning with Differential Logic Layer
by: Zhang, Yu, et al.
Published: (2024)

Confounding Robust Continuous Control via Automatic Reward Shaping
by: Juliani, Mateo, et al.
Published: (2026)

HardSATGEN: Understanding the Difficulty of Hard SAT Formula Generation and A Strong Structure-Hardness-Aware Baseline
by: Li, Yang, et al.
Published: (2023)

INT v.s. FP: A Comprehensive Study of Fine-Grained Low-bit Quantization Formats
by: Chen, Mengzhao, et al.
Published: (2025)