Skip to main content
zenodoopen

Training performance comparison includes Oases and selective recomputation

<p>Training performance on LLaMA-7B model. "Megatron SS" represents the Megatron with selective recomputation. Compared to Megatron SS, Oases can accelerate training by 1.15&ndash;1.38&times;.</p>

ShareScore

32/100

Overall dataset sharing score

Score breakdown

These five areas show where the dataset supports — or may limit — practical reuse.

Stewardship
4
Harmonization
4
Access
16
Reuse readiness
8
Engagement
0