feat(pytorch): add last-only prefill checkpoint saving for hybrid attention - #5049

feat(pytorch): add last-only prefill checkpoint saving for hybrid attention#5049

RunningLeon merged 1 commit into

InternLM:mainInternLM/lmdeploy:mainfrom

caikun-pjlab:optimize/optimize_state_kvcache_savingcaikun-pjlab/lmdeploy:optimize/optimize_state_kvcache_savingCopy head branch name to clipboard