wassname/pytorch-lightning
1
0
Fork 0
mirror of https://github.com/wassname/pytorch-lightning.git synced 2026-09-11 12:31:23 +08:00
Code Issues Packages Projects Releases Wiki Activity

added single node distdataparallel

Browse Source
This commit is contained in:
William Falcon
2019-07-03 15:16:09 -04:00
parent 96ab78dc41
commit ac57dac235
1 changed files with 2 additions and 2 deletions
Download Patch File Download Diff File Expand all files Collapse all files
pytorch_lightning/models/trainer.py
+2 -2
View File
@@ -282,12 +282,11 @@ class Trainer(TrainerIO):
# when GPU is called, spawn off a single worker for each gpu
if self.on_gpu:
rank = 0
pdb.set_trace()
mp.spawn(self.__dp_train, nprocs=len(self.data_parallel_device_ids), args=(rank, model ))
else:
self.__run_pretrain_routine(model)
def __dp_train(self, gpu_nb, proc_rank, model, cluster_obj):
def __dp_train(self, gpu_nb, proc_rank, model):
"""
Entry point into a DP thread
:param gpu_nb:
@@ -297,6 +296,7 @@ class Trainer(TrainerIO):
"""
# TODO: pass in ip
ip = "127.0.0.1"
print(self.data_parallel_device_ids)
# configure server
rank = proc_rank * len(self.data_parallel_device_ids) + gpu_nb