params should not be reinitialized for loaded models

Author: bmccann

OpenNMT/train.py

@@ -151,8 +151,6 @@ def eval(model, criterion, data):
 def trainModel(model, trainData, validData, dataset, optim):
  print(model)
  model.train()
- for p in model.parameters():
- p.data.uniform_(-opt.param_init, opt.param_init)
 
  # define criterion of each GPU
  criterion = NMTCriterion(dataset['dicts']['tgt'].size())
@@ -258,6 +256,8 @@ def main():
  lr_decay=opt.learning_rate_decay,
  start_decay_at=opt.start_decay_at
  )
+ for p in model.parameters():
+ p.data.uniform_(-opt.param_init, opt.param_init)
  else:
  print('Loading from checkpoint at %s' % opt.train_from)
  checkpoint = torch.load(opt.train_from)