Megatron-LM

5 stories

Stories

Topic: Megatron-LM

Coordinated software tools could cut AI training energy use by up to 50%
Around half of the energy used to train AI can be saved without affecting speed or performance, using software developed at the University of Michigan Engineering.
TechXplore AIOct 12, 20264 min read
MLPerf Training Introduces Its First LLM Post-Training Benchmark
MLPerf Training v6.1 adds an agentic reinforcement-learning workload that measures how quickly systems can teach a 397-billion-parameter language model to repair real software projects. The post MLPerf Training Introduces Its First LLM Post-Training Benchmark appeared first on MLCommons.
MLCommonsSep 24, 202611 min read
Granite 4.2 LLMs: How They're Built
TL;DR: Granite 4.2 is our first family of dense, decoder-only reasoning LLMs, released in three sizes: 3B, 8B, and 30B. These models are post-trained from Granite-4.1 base models.
Hugging Face BlogAug 25, 202621 min read
H100 vs GB200 NVL72 Training Benchmarks – Power, TCO, and Reliability Analysis, Software Improvement Over Time
Frontier model training has pushed GPUs and AI systems to their absolute limits, making cost, efficiency, power, performance per TCO, and reliability central to the discussion on effective training. The Hopper vs Blackwell comparisons are not as simple as Nvidia would have you believe.
SemiAnalysisAug 20, 202515 min read
How to Train Really Large Models on Many GPUs?
[Updated on 2022-03-13: add expert choice routing.] [Updated on 2022-06-10]: Greg and I wrote a shorted and upgraded version of this post, published on OpenAI Blog: “Techniques for Training Large Neural Networks”
Lil'Log (Lilian Weng)Sep 24, 202118 min read