Side effect of blocking Chinese firms from buying the best NVIDIA cards: top models are now explicitly being trained to work well on older/cheaper GPUs. The new SoTA model from @Kimi_Moonshot uses plain old BF16 ops (after dequant from INT4); no need for expensive FP4 support.
🚀 "Quantization is not a compromise — it's the next paradigm." After K2-Thinking's release, many developers have been curious about its native INT4 quantization format. 刘少伟, infra engineer at @Kimi_Moonshot and Zhihu contributor, shares an insider's view on why this choice…
Finally wrapped up the hardware half of Nand2Tetris, took me a week. It had been on my to-do list for way longer than I’d like to admit.
Deepseek engineers so cracked they bypassed cuda
United States เทรนด์
- 1. Florida 104K posts
- 2. Texas 177K posts
- 3. Ohio State 28.5K posts
- 4. Ohio State 28.5K posts
- 5. #SmallBusinessSaturday 2,444 posts
- 6. Kentucky 14.9K posts
- 7. Kentucky 14.9K posts
- 8. Go Blue 6,933 posts
- 9. Buckeyes 5,267 posts
- 10. Go Bucks 2,363 posts
- 11. Leeds 26.5K posts
- 12. Saban 5,781 posts
- 13. Sunderland 22.5K posts
- 14. Good Saturday 36.5K posts
- 15. Grade 3 2,930 posts
- 16. Tyler Adams 3,154 posts
- 17. The Game 1.04M posts
- 18. #SaturdayVibes 4,156 posts
- 19. #MeAndTheeSeriesEP3 1.53M posts
- 20. Georgia 52.1K posts
Something went wrong.
Something went wrong.