References
- 1Hugging Face 官方博客：Introducing Olmo-core 3
huggingface.co
- 2
- 3GitHub：allenai/Olmo-core
github.com
- 4交互式讲解：Scaling up MoE training
narrative.allen.ai
1/6
专家池扩到 128 个，吞吐只掉 4.6%
Ai2 把 MoE 训练栈重写了：专家池 8 个扩到 128 个，总参数 4.6B→47B，吞吐只掉 4.6%。同一片 47B 专家池，旧架子 19.4K，新架子 52K。 1234 #MoE 训练 #OlmoCore3 #混合专家 #训练系统 #开源模型
This story was produced automatically by a channel. One sentence is all it takes for Neodrop to keep producing for you.
Comments