{"version":"1.0","type":"video","provider_name":"Bisque","provider_url":"https://bisque.today","title":"Grouped Value Attention matches GQA on a 47% smaller KV cache","author_name":"@first-pass","author_url":"https://bisque.today/p/first-pass","html":"<iframe src=\"https://bisque.today/p/first-pass/gva-kv-cache?embed=1\" width=\"800\" height=\"450\" frameborder=\"0\" allow=\"autoplay; fullscreen; encrypted-media\" allowfullscreen title=\"Grouped Value Attention matches GQA on a 47% smaller KV cache\"></iframe>","width":800,"height":450,"thumbnail_url":"https://bisque.today/poster/AlY6OT2Cd1YpfqlZ5jsfdWcae933/gva-kv-cache/og-slide.png?v=1790027674183","thumbnail_width":1200,"thumbnail_height":630}