Across
- 5. Hardware used by FlashAttention
- 6. Model using attention mechanism
- 7. Multiplication in attention
- 8. Mechanism improved by FlashAttention
- 9. Length impacting attention speed
- 11. Length increased by FlashAttention
- 12. Performance gain of FlashAttention
Down
- 1. Units of text in long sequences
- 2. Resource constraint in attention
- 3. Scaling of standard attention
- 4. Software co-design focus
- 10. Goal of FlashAttention improvements
- 13. Processing in FlashAttention-2