Validate on all sets based on --save-interval-updates

Summary: Pull Request resolved: https://github.com/pytorch/fairseq/pull/693

Differential Revision: D15174831

fbshipit-source-id: 98688b1269ead5694e5116659ff64507d3c0d1c0

train.py

@@ -134,7 +134,7 @@ def train(args, trainer, task, epoch_itr):
     )
 
     extra_meters = collections.defaultdict(lambda: AverageMeter())
-    first_valid = args.valid_subset.split(',')[0]
+    valid_subsets = args.valid_subset.split(',')
     max_update = args.max_update or math.inf
     for i, samples in enumerate(progress, start=epoch_itr.iterations_in_epoch):
         log_output = trainer.train_step(samples)
@@ -159,7 +159,7 @@ def train(args, trainer, task, epoch_itr):
 
         num_updates = trainer.get_num_updates()
         if args.save_interval_updates > 0 and num_updates % args.save_interval_updates == 0 and num_updates > 0:
-            valid_losses = validate(args, trainer, task, epoch_itr, [first_valid])
+            valid_losses = validate(args, trainer, task, epoch_itr, valid_subsets)
             save_checkpoint(args, trainer, epoch_itr, valid_losses[0])
 
         if num_updates >= max_update: