Sign up to save your podcasts
Or
Share When Precision Meets Position: BFloat16 Breaks Down RoPE in Long-Context Training
Share to email
Share to Facebook
Share to X
Share to email
Share to Facebook
Share to X
Or
When Precision Meets Position: BFloat16 Breaks Down RoPE in Long-Context Training
AnchorAttention addresses numerical issues in BFloat16 with Rotary Positional Embedding, enhancing long-context performance and training efficiency in large language models while maintaining task capabilities.
https://arxiv.org/abs//2411.13476
YouTube: https://www.youtube.com/@ArxivPapers
TikTok: https://www.tiktok.com/@arxiv_papers
Apple Podcasts: https://podcasts.apple.com/us/podcast/arxiv-papers/id1692476016
Spotify: https://podcasters.spotify.com/pod/show/arxiv-papers
View all episodes
By Igor Melnyk
5
33 ratings
AnchorAttention addresses numerical issues in BFloat16 with Rotary Positional Embedding, enhancing long-context performance and training efficiency in large language models while maintaining task capabilities.
https://arxiv.org/abs//2411.13476
YouTube: https://www.youtube.com/@ArxivPapers
TikTok: https://www.tiktok.com/@arxiv_papers
Apple Podcasts: https://podcasts.apple.com/us/podcast/arxiv-papers/id1692476016
Spotify: https://podcasters.spotify.com/pod/show/arxiv-papers
More shows like Arxiv Papers
View all
Exchanges
977 Listeners
Odd Lots
1,993 Listeners
The TWIML AI Podcast (formerly This Week in Machine Learning & Artificial Intelligence)
443 Listeners
The Daily
113,121 Listeners
All-In with Chamath, Jason, Sacks & Friedberg
10,254 Listeners
Hard Fork
5,576 Listeners
UnHerd with Freddie Sayers
221 Listeners
Unsupervised Learning with Jacob Effron
51 Listeners
Latent Space: The AI Engineer Podcast
101 Listeners
BG2Pod with Brad Gerstner and Bill Gurley
475 Listeners