Flash-MSA: Accelerating Million-Token Training with Sparse Attention Kernelsnanduruganesh.github.io25 points·by rawsh·há 6 horas·1 comments
Zyphra ZAYA1-base: First large-scale model trained on AMDzyphra.com6 points·by rawsh·há 8 meses·1 comments
Debugging divergence between engine and transformers logprobs for RLgist.github.com2 points·by rawsh·há 10 meses·0 comments