‹ projects

cluster-rnn

a distributed Torch7 RNN cluster over MPI
Log | Files | Refs | README

DateCommit messageAuthorFiles+-
2017-02-18 23:58adjust default parametersumhau2+4-4
2017-02-16 18:37diagumhau1+4-2
2017-02-16 18:36diagnosticsumhau1+1-0
2017-02-16 18:14enable opt field in checkpointumhau1+1-1
2017-02-16 18:13include test run commandumhau2+5-1
2017-02-16 17:24troubleshooting uncreated initial filesumhau2+9-4
2017-02-15 22:59allow fractional epochsumhau1+1-1
2017-02-15 22:53asdumhau1+1-1
2017-02-15 22:51asdumhau1+1-1
2017-02-15 22:24asdumhau1+2-0
2017-02-15 21:46khjbumhau1+8-8
2017-02-15 21:02sddsffdumhau1+7-7
2017-02-15 20:51hide opt classumhau1+1-1
2017-02-15 19:55hide all mentions of pclientumhau1+1-1
2017-02-15 19:29shorten training againumhau1+1-1
2017-02-15 19:27shorten training timeumhau1+1-1
2017-02-15 19:19dittoumhau1+1-1
2017-02-15 19:18removed optimm_state from checkpoint. this allows saves, but prevents me from picking up where I left off. Necessary, because the pclient settings don't like being saved.umhau1+1-1
2017-02-15 19:10mention rank in status updatesumhau1+1-1
2017-02-15 19:08pclient data should not be part of the optim_state variableumhau1+2-2
2017-02-15 06:48pumhau1+1-1
2017-02-15 06:46pumhau1+1-1
2017-02-15 06:44pumhau1+4-4
2017-02-15 06:37pumhau1+1-1
2017-02-15 06:28pumhau1+2-2
2017-02-15 06:17pumhau1+1-1
2017-02-15 06:16parametersumhau1+1-1
2017-02-15 05:59shorten communication periodumhau1+1-1
2017-02-15 05:56make more mpi parameters availableumhau1+2-0
2017-02-15 05:52make more parameters accessible for adjustmentumhau1+17-4
2017-02-15 05:40fix learning rateumhau1+1-1
2017-02-15 05:21change default optimizerumhau1+1-1
2017-02-15 02:30shorten training time for testingumhau1+1-1
2017-02-14 23:29show hostname in statusumhau1+2-1
2017-02-14 23:16change name of mpi options objectumhau2+21-20
2017-02-14 23:08fix nil rank valueumhau1+1-1
2017-02-14 23:04don't need the msgd or downpour algorithmsumhau1+0-0
2017-02-14 23:03don't need the msgd or downpour algorithmsumhau1+2-2
2017-02-14 23:03Don't need the msgd or downpour algorithmsumhau1+0-0
2017-02-14 23:00readme includes example command, mlaunch runs train.luaumhau2+10-3
2017-02-14 22:53implement EAMSGD in word-rnnumhau10+245-193
2017-02-14 21:21better explanatory commentsumhau1+60-40
2017-02-14 21:18added explanation to readme.umhau1+6-0
2017-02-11 08:07initial compilationumhau30+42876-0
2017-02-11 07:49Initial commitumhau2+43-0