32c92640f0
Fix pserver checkpoint
...
The pserver checkpoint before failed because the MD5 checksum is
calculated incorrectly. Now changed to CRC32 checksum.
8 years ago
60238a1bfb
Go master, pserver, trainer: switch to log15, away from logrus
8 years ago
33fb8d7abf
fix according to comments
8 years ago
2ee418db78
fix pserver save / load checkpoint
8 years ago
6fab04f4e1
fix vet shadow report
8 years ago
42fe3e88c7
gracefully shutdown pserver, fix gometalinter errors
8 years ago
cb5c7526e5
shutdown master server gracefully
8 years ago
a7e69d949f
do not do log.Errorln when checkpoint is not found (which is normal)
8 years ago
e1e7309789
boring copyright
8 years ago
83f263e6ec
Fix fetch record from master failed ( #2848 )
...
Fix fetch record from master
8 years ago
23b8346072
Fault tolerant distributed training, just work version, with etcd ( #2849 )
...
* using etcd as fault tolerant training
* update
* workable version, ft not tested
* small fix
* update
* remove TODO
8 years ago
46c704ecf0
"fix init error"
8 years ago
e6e2bf45e5
Merge pull request #2832 from helinwang/go_cmake
...
go_binary: remove hardcoded library link path, add pserver client test
8 years ago
19bfb8a1f2
PServer recovery from checkpoint ( #2741 )
...
* Server recovery from checkpoint
8 years ago
2231b92a89
go_binary: remove hardcoded library link path
8 years ago
0ad7053e96
"make parameterCheckpoint exported"
8 years ago
87e7924e4e
"pserver flags type error"
8 years ago
40295b9ed9
"fix pserver saving etcd"
8 years ago
166dfbb085
fix cmake errors
8 years ago
9045063b53
pserver etcd client ( #2559 )
...
* init etcd cclient
* add etcd
* add etcd.go
* fix compile problem
* move code to etcd.go
* add etcd_lister.go for pserver client
* add etcd_client_test.go
* merge etcd_client_test and client_test
* refine client_test.go
* refine code
* format code
* add TODO and use interface instead of struct
* fix typo of initDesiredPservers
* optimize dir structure of go/pserver/client
* add a flag to config index for pserver
* follow comment
* fix path
* optimize code
* remove err in pserver NewEtcd
* restore comment about /ps_desired
8 years ago
9c853c269d
Merge branch 'develop' of https://github.com/PaddlePaddle/Paddle into cmake_go_vendor
8 years ago
4cc9680cc6
Make pserver able to get server index without etcd (decouple pserver with etcd)
...
The pserver need to have server index for saving model on the
distributed file system. The server index comes from etcd if etcd is
used, or user can manually specify them. So we need
pserver.NewService() to take index as an argument. Since index could
come from etcd, it would be strange if pserver takes an index as
argument, at the same time get the index from etcd. so we will need to
decouple pserver with etcd.
8 years ago
155e40ef26
using glide for go package vendor
8 years ago
0824061ea3
set ps_desired when pserver init
8 years ago
7dad02661f
Master server registers itself to etcd.
8 years ago
42313a3c35
rename EtcdStore to Etcd
8 years ago
4422685302
put InMemStore into master package
8 years ago
fd8937556f
Master save and load state from etcd
8 years ago
0b936e9399
update pserver etcd
8 years ago
b7a52bd976
add started info log
8 years ago
badcdfe1e5
pserver etcd registration
8 years ago
54e8263cae
implement master server client, remove unnecessary dummy variable
8 years ago
72a73ab6d2
implement master server client, RPC part.
8 years ago
d0f1890db3
move recordio to github.com/PaddlePaddle/recordio
8 years ago
28b23e6db5
move paddle/go to go, move go/cclient to go/pserver/cclient, move go/crecordio to go/recordio/c
8 years ago
Helin Wang
dongzhihong
Yancey
武毅
gangliao
gongweibao
Qiao Longfei
yi.wu
wuyi05