a ‘º”hãã@sVUddlmZgZeeed<eeedœdd„Zeeeee e e e fdœdd„Z d S) é)ÚChunkStorageMetadataÚ__all__)Úshard1Úshard2ÚreturncCsdt|jƒ}t|ƒD]L}|j||j||j|kr:dS|j||j||j|krdSqdS)zCheck if two shards overlap.FT)ÚlenÚoffsetsÚrangeÚsizes)rrZndimsÚi©r úU/var/www/auris/lib/python3.9/site-packages/torch/distributed/checkpoint/resharding.pyÚ"_check_shard_metadata_pair_overlaps  r)Ú saved_shardÚ current_shardrc Csˆg}tt|j|j|j|jƒƒD]d\}\}}}}t||||ƒ}|t||ƒ} ||krdd} ||} n ||} d} | || | | f¡q|S)a[ Return the overlapping region between saved_shard and current_shard. There returned list has the same number of elements as the tensor's dimension. For each element, we produce a tuple with the following contents: (dimension, `saved_shard` offset, `current_shard` offset, length) Offsets are relative to each shard. r)Ú enumerateÚziprr ÚminÚmaxÚappend) rrZnarrowsZdimZsaved_shard_offsetZcurrent_shard_offsetZsaved_shard_sizeZcurrent_shard_sizeZ min_range_endÚlengthZoffset_for_saved_tensorZoffset_for_current_tensorr r r Ú+_shards_get_overlap_region_wrt_saved_tensors8 üÿû  þ  ÿrN) Z%torch.distributed.checkpoint.metadatarrÚlistÚstrÚ__annotations__ÚboolrÚtupleÚintrr r r r Úsþ þ