Hendrik/xgboost
1
0
Fork 0
Code Issues Pull Requests Packages Projects Releases Wiki Activity

skip actions when there is only 1 node

Browse Source
This commit is contained in:
tqchen 2014-12-19 19:21:21 -08:00
parent 6151899ce2
commit 2c0a0671ad
1 changed files with 10 additions and 0 deletions
Show Stats Download Patch File Download Diff File Expand all files Collapse all files

10
src/allreduce_robust.cc
View File

@ -68,6 +68,8 @@ void AllreduceRobust::Allreduce(void *sendrecvbuf_,
ReduceFunction reducer, ReduceFunction reducer,
PreprocFunction prepare_fun, PreprocFunction prepare_fun,
void *prepare_arg) { void *prepare_arg) {
// skip action in single node
if (world_size == 1) return;
bool recovered = RecoverExec(sendrecvbuf_, type_nbytes * count, 0, seq_counter); bool recovered = RecoverExec(sendrecvbuf_, type_nbytes * count, 0, seq_counter);
// now we are free to remove the last result, if any // now we are free to remove the last result, if any
if (resbuf.LastSeqNo() != -1 && if (resbuf.LastSeqNo() != -1 &&
@ -98,6 +100,8 @@ void AllreduceRobust::Allreduce(void *sendrecvbuf_,
* \param root the root worker id to broadcast the data * \param root the root worker id to broadcast the data
*/ */
void AllreduceRobust::Broadcast(void *sendrecvbuf_, size_t total_size, int root) { void AllreduceRobust::Broadcast(void *sendrecvbuf_, size_t total_size, int root) {
// skip action in single node
if (world_size == 1) return;
bool recovered = RecoverExec(sendrecvbuf_, total_size, 0, seq_counter); bool recovered = RecoverExec(sendrecvbuf_, total_size, 0, seq_counter);
// now we are free to remove the last result, if any // now we are free to remove the last result, if any
if (resbuf.LastSeqNo() != -1 && if (resbuf.LastSeqNo() != -1 &&
@ -143,6 +147,8 @@ void AllreduceRobust::Broadcast(void *sendrecvbuf_, size_t total_size, int root)
*/ */
int AllreduceRobust::LoadCheckPoint(ISerializable *global_model, int AllreduceRobust::LoadCheckPoint(ISerializable *global_model,
ISerializable *local_model) { ISerializable *local_model) {
// skip action in single node
if (world_size == 1) return 0;
if (num_local_replica == 0) { if (num_local_replica == 0) {
utils::Check(local_model == NULL, "need to set num_local_replica larger than 1 to checkpoint local_model"); utils::Check(local_model == NULL, "need to set num_local_replica larger than 1 to checkpoint local_model");
} }
@ -200,6 +206,10 @@ int AllreduceRobust::LoadCheckPoint(ISerializable *global_model,
*/ */
void AllreduceRobust::CheckPoint(const ISerializable *global_model, void AllreduceRobust::CheckPoint(const ISerializable *global_model,
const ISerializable *local_model) { const ISerializable *local_model) {
// never do check point in single machine mode
if (world_size == 1) {
version_number += 1; return;
}
if (num_local_replica == 0) { if (num_local_replica == 0) {
utils::Check(local_model == NULL, "need to set num_local_replica larger than 1 to checkpoint local_model"); utils::Check(local_model == NULL, "need to set num_local_replica larger than 1 to checkpoint local_model");
} }