tomaarsen @tomaarsen · 18 Aug 2026
Also faster: multi-column losses now run one forward pass over merged columns, ~1.25x on hard-negative & triplet training, with identical loss trajectories. And fp16 + FlashAttention is now the fastest GPU config for SentenceTransformer at 3.87x over fp32.
405Views
12Likes
0Reposts
2Replies
0Quotes
1Bookmarks
Is that a lot?
0.64×vs this author's median637 views is typical
9Percentile for this authorof 22 recent posts
0.13×vs under 10K median3 020 views is typical
8.41%Reachviews ÷ followers
3.46%Engagement rateof viewers reacted