Commit Graph
255 Commits
Author SHA1 Message Date
William Falcon 470f3e6d29 added training router 2019-07-18 11:08:48 -04:00
William Falcon c12a0b57da added dp and ddp flag 2019-07-18 11:03:16 -04:00
William Falcon e7ecfa15f8 added option and flag 2019-07-18 10:56:45 -04:00
William Falcon a41abad5b2 Update trainer.py 2019-07-16 17:02:21 -04:00
William Falcon a83588b14e Update trainer.py 2019-07-16 13:12:56 -04:00
Cinjon Resnick fbd3873a0f add a hook for on_tng_metrics so that users get access to the grad_norm and mem_map dicts. 2019-07-16 12:51:48 -04:00
William Falcon 28cfddbe65 accept dist sampler classes 2019-07-16 12:44:58 -04:00
William Falcon 967e57f071 early stop starts counting once min epochs met 2019-07-16 10:00:03 -04:00
William Falcon d12f6b7dd8 added summary flag 2019-07-15 21:11:29 -04:00
William Falcon 182c025c88 removed validation call 2019-07-15 20:48:46 -04:00
William Falcon 58e6199ce8 removed validation call 2019-07-15 14:56:56 -04:00
William Falcon 6a33f0d483 made early stop checkpoint optional 2019-07-15 14:54:38 -04:00
William Falcon dd230a93e8 made early stop checkpoint optional 2019-07-15 14:53:37 -04:00
William Falcon 3aa9cfc18e made checkpoint callback optional 2019-07-15 13:18:56 -04:00
William Falcon e57f461323 made checkpoint callback optional 2019-07-15 13:17:38 -04:00
William Falcon b4b8a3dfde fixed none bug 2019-07-15 13:01:08 -04:00
William Falcon ad24bef1c9 removed print statements 2019-07-14 18:12:41 -04:00
William Falcon 50246a5066 working on single gpu init speed 2019-07-14 17:33:48 -04:00
William Falcon 21914cb1c1 working on single gpu init speed 2019-07-14 17:15:20 -04:00
William Falcon 904935cf98 working on single gpu init speed 2019-07-14 17:11:52 -04:00
William Falcon 468e75c180 working on single gpu init speed 2019-07-14 17:10:13 -04:00
William Falcon 849f52b7a6 modified single gpu init 2019-07-14 17:01:18 -04:00
William Falcon e520297781 modified single gpu init 2019-07-14 16:57:15 -04:00
William Falcon cefc27112d ddp flag change 2019-07-13 22:28:08 -04:00
William Falcon 6876f60098 merge 2019-07-13 22:21:17 -04:00
William Falcon e9f5913dac enabling gpu size = 1 to run without data parallel 2019-07-13 22:16:10 -04:00
William Falcon 7da82c2560 added fallback local init 2019-07-13 22:16:10 -04:00
William Falcon a2639c6894 added fallback local init 2019-07-13 22:16:10 -04:00
William Falcon eb05fa316f added fallback local init 2019-07-13 22:16:10 -04:00
William Falcon 6d55adb0d8 fixed nccl init 2019-07-13 22:16:10 -04:00
William Falcon cff0500a63 fixed nccl init 2019-07-13 22:16:10 -04:00
William Falcon f3ca184fb6 fixed nccl init 2019-07-13 22:16:10 -04:00
William Falcon 3239c9fdf8 fixed nccl init 2019-07-13 22:16:10 -04:00
William Falcon 7e37f68a5b fixed nccl init 2019-07-13 22:16:10 -04:00
William Falcon 960937ebe9 fixed nccl init 2019-07-13 22:16:10 -04:00
William Falcon a87784b4c5 fixed nccl init 2019-07-13 22:16:10 -04:00
William Falcon 5812efcf24 fixed nccl init 2019-07-13 22:16:10 -04:00
William Falcon e82014ec6c fixed nccl init 2019-07-13 22:16:10 -04:00
William Falcon dc87a4fc91 fixed nccl init 2019-07-13 22:16:10 -04:00
William Falcon 4f5eef2e78 fixed nccl init 2019-07-13 22:16:10 -04:00
William Falcon 6c02afefca fixed nccl init 2019-07-13 22:16:10 -04:00
William Falcon 4696e12641 fixed nccl init 2019-07-13 22:16:10 -04:00
William Falcon c1b21fb1e4 Merge pull request #11 from cinjon/modulefix
trainer: module fix.
2019-07-12 15:28:48 -04:00
William Falcon 8451bb7745 fixed nccl init 2019-07-12 15:25:34 -04:00
William Falcon 1a1771cfd8 fixed nccl init 2019-07-12 15:24:42 -04:00
William Falcon 1952e9be49 fixed nccl init 2019-07-12 15:11:32 -04:00
William Falcon 19391b1df1 fixed nccl init 2019-07-12 15:04:20 -04:00
William Falcon 369174c4d3 fixed nccl init 2019-07-12 14:36:00 -04:00
William Falcon ba38037917 fixed nccl init 2019-07-12 13:39:58 -04:00
William Falcon a7bb731a1d testing env init 2019-07-12 13:19:10 -04:00