ciflow/trunk/201842: [symm_mem] Free reserved virtual address ranges on teardown
- PyTorch 近 90 天出现 2736 次
- PyTorch 近 90 天第 2693 次发版
- 上一次:同一天稍早 · ciflow/torchtitan/201769: Update
发生了什么
map_block() reserves a virtual address (VA) range with cuMemAddressReserve. No code frees this range. ~AllocationRef unmaps and releases the memory, but it does not call cuMemAddressFree. The multicast cleanup in MulticastSetup::release() has the same problem. Each alloc and free leaks one VA range of block_size. Each peer mapping and each multicast mapping from rendezvous also leaks one range. The physical memory i…
摘要按规则整理自下方来源原文