On-device perf + memory optimizations: custom SDPA, on-the-fly RoPE, KV cache fix, XNNPACK workspace sharing (#19214)#19214
Open
leixin wants to merge 1 commit intopytorch:mainfrom
Open
On-device perf + memory optimizations: custom SDPA, on-the-fly RoPE, KV cache fix, XNNPACK workspace sharing (#19214)#19214leixin wants to merge 1 commit intopytorch:mainfrom
leixin wants to merge 1 commit intopytorch:mainfrom
