← All talks

NYU Theory Day Presentation

QJL: 1-Bit Quantized JL Transform for KV Cache Quantization with Zero Overhead

New York, NY, USA

AAAI conference paper · arXiv · Code · ICERM lightning talk

Open original PDF Download PDF
QJL: 1-Bit Quantized JL Transform for KV Cache Quantization with Zero Overhead — slide 1 of 26
Slide 1 of 26 · Select the slide to view it at full size.
QJL: 1-Bit Quantized JL Transform for KV Cache Quantization with Zero Overhead — slide 2 of 26
Slide 2 of 26 · Select the slide to view it at full size.
QJL: 1-Bit Quantized JL Transform for KV Cache Quantization with Zero Overhead — slide 3 of 26
Slide 3 of 26 · Select the slide to view it at full size.
QJL: 1-Bit Quantized JL Transform for KV Cache Quantization with Zero Overhead — slide 4 of 26
Slide 4 of 26 · Select the slide to view it at full size.
QJL: 1-Bit Quantized JL Transform for KV Cache Quantization with Zero Overhead — slide 5 of 26
Slide 5 of 26 · Select the slide to view it at full size.
QJL: 1-Bit Quantized JL Transform for KV Cache Quantization with Zero Overhead — slide 6 of 26
Slide 6 of 26 · Select the slide to view it at full size.
QJL: 1-Bit Quantized JL Transform for KV Cache Quantization with Zero Overhead — slide 7 of 26
Slide 7 of 26 · Select the slide to view it at full size.
QJL: 1-Bit Quantized JL Transform for KV Cache Quantization with Zero Overhead — slide 8 of 26
Slide 8 of 26 · Select the slide to view it at full size.
QJL: 1-Bit Quantized JL Transform for KV Cache Quantization with Zero Overhead — slide 9 of 26
Slide 9 of 26 · Select the slide to view it at full size.
QJL: 1-Bit Quantized JL Transform for KV Cache Quantization with Zero Overhead — slide 10 of 26
Slide 10 of 26 · Select the slide to view it at full size.
QJL: 1-Bit Quantized JL Transform for KV Cache Quantization with Zero Overhead — slide 11 of 26
Slide 11 of 26 · Select the slide to view it at full size.
QJL: 1-Bit Quantized JL Transform for KV Cache Quantization with Zero Overhead — slide 12 of 26
Slide 12 of 26 · Select the slide to view it at full size.
QJL: 1-Bit Quantized JL Transform for KV Cache Quantization with Zero Overhead — slide 13 of 26
Slide 13 of 26 · Select the slide to view it at full size.
QJL: 1-Bit Quantized JL Transform for KV Cache Quantization with Zero Overhead — slide 14 of 26
Slide 14 of 26 · Select the slide to view it at full size.
QJL: 1-Bit Quantized JL Transform for KV Cache Quantization with Zero Overhead — slide 15 of 26
Slide 15 of 26 · Select the slide to view it at full size.
QJL: 1-Bit Quantized JL Transform for KV Cache Quantization with Zero Overhead — slide 16 of 26
Slide 16 of 26 · Select the slide to view it at full size.
QJL: 1-Bit Quantized JL Transform for KV Cache Quantization with Zero Overhead — slide 17 of 26
Slide 17 of 26 · Select the slide to view it at full size.
QJL: 1-Bit Quantized JL Transform for KV Cache Quantization with Zero Overhead — slide 18 of 26
Slide 18 of 26 · Select the slide to view it at full size.
QJL: 1-Bit Quantized JL Transform for KV Cache Quantization with Zero Overhead — slide 19 of 26
Slide 19 of 26 · Select the slide to view it at full size.
QJL: 1-Bit Quantized JL Transform for KV Cache Quantization with Zero Overhead — slide 20 of 26
Slide 20 of 26 · Select the slide to view it at full size.
QJL: 1-Bit Quantized JL Transform for KV Cache Quantization with Zero Overhead — slide 21 of 26
Slide 21 of 26 · Select the slide to view it at full size.
QJL: 1-Bit Quantized JL Transform for KV Cache Quantization with Zero Overhead — slide 22 of 26
Slide 22 of 26 · Select the slide to view it at full size.
QJL: 1-Bit Quantized JL Transform for KV Cache Quantization with Zero Overhead — slide 23 of 26
Slide 23 of 26 · Select the slide to view it at full size.
QJL: 1-Bit Quantized JL Transform for KV Cache Quantization with Zero Overhead — slide 24 of 26
Slide 24 of 26 · Select the slide to view it at full size.
QJL: 1-Bit Quantized JL Transform for KV Cache Quantization with Zero Overhead — slide 25 of 26
Slide 25 of 26 · Select the slide to view it at full size.
QJL: 1-Bit Quantized JL Transform for KV Cache Quantization with Zero Overhead — slide 26 of 26
Slide 26 of 26 · Select the slide to view it at full size.