Paddle

Commit Graph

Author	SHA1	Message	Date
WangXi	572c466d19	[Prepare for MultiProcess xpu] unified gen nccl id, refine imperative reducer (#30455 )	4 years ago
ShenLiang	01e2874a0e	Support multi-stream communication for dynamic graph distributed (#29525 ) * fix fleet for multi-stream * fix memcpy for ncclid * use sync to solve move operation	4 years ago
ShenLiang	e2d01eb650	Support dynamic graph distributed (#28997 ) * add reducer * refine envent for memorycopy * add concat&split for allreduce * apply concat & split for fuse tensor * fix nccl dep * fix the untest, compile problem and ddp initialize problem * fix untest for mac & add some comments & solve the repeated param in sublayers * fix untest for windows & fix document	4 years ago
danleifeng	a24d186814	fix nccl init failed in parallel dygraph mode (#28497 )	4 years ago
Chen Weihang	c42e656179	Add retry for dygraph parallel socket bind (#28404 ) * add retry for dygraph parallel socket bind * change to loop always * fix writing error	4 years ago
danleifeng	f29fb396df	dygraph nccl init support host domain name (#28107 ) * nccl init support hostname and ip; test=develop	4 years ago
Leo Chen	a5b3263782	Refine error msg in paddle/fluid/imperative (#27521 ) * refine err msg * follow comments	4 years ago
Chen Weihang	aa0f254fbe	Add macro BOOST_GET to enrich the error information of boost :: get (#24175 ) * add new macro BOOST_GET_SAFELY & unittests, test=develop * add different macro type, test=develop * fix get macro type in executor, test=develop * four macro part change backup * using one macro for all case, test=develop * revert attribute change, test=develop * change to three func to solve gcc4.8 bug, test=develop * polish some details, test=develop	5 years ago
Yi Liu	2169e6fb58	Initialize global nccl_comm in PE (#23275 )	5 years ago
Yi Liu	121b2aed4d	initialize global nccl context in dygraph (#23037 ) initialize global nccl context in dygraph test=develop	5 years ago
Wilber	7bc4b09500	add WITH_NCCL option for cmake. (#22384 ) cmake选项中添加了WITH_NCCL，显示指定是否编译NCCL的部分代码，WITH_NCCL默认打开，但如果WITH_GPU为OFF，则关闭WITH_NCCL 添加了PADDLE_WITH_NCCL定义单机单卡能够关闭NCCL编译，多卡的话需要默认打开NCCL，如果关闭NCCL，则只能使用单卡 Co-authored-by: 石晓伟 <39303645+Shixiaowei02@users.noreply.github.com>	5 years ago
chengduo	cca26f5c42	polish multi process warning info (#19961 ) test=develop	5 years ago
chengduo	5436d66667	close socket connect (#17862 ) test=develop	6 years ago
Yan Xu	55e3c6949b	disable reuse port test=develop (#16704 )	6 years ago
Yan Xu	b4c3a6aa0b	[Imperative] implement imperative NCCLParallelContext (#16477 ) add NCCLParallelContext for parallel dygraph	6 years ago

15 Commits (f8da5536edaa004fd42988539508f6810a2fe958)