Use --train-subset and --valid-subset properly

Summary: Pull Request resolved: https://github.com/pytorch/fairseq/pull/614

Differential Revision: D14712321

Pulled By: myleott

fbshipit-source-id: 8ef973c5d30ebccf0df0f1cabdddd590248a8f8d

train.py

@@ -39,7 +39,7 @@ def main(args, init_distributed=False):
     task = tasks.setup_task(args)
 
     # Load dataset splits
-    load_dataset_splits(task, ['train', 'valid'])
+    load_dataset_splits(args, task)
 
     # Initialize distributed training (after data loading)
     if init_distributed:
@@ -64,8 +64,8 @@ def main(args, init_distributed=False):
         task.max_positions(),
         model.max_positions(),
     )
-    dummy_batch = task.dataset('train').get_dummy_batch(args.max_tokens, max_positions)
-    oom_batch = task.dataset('train').get_dummy_batch(1, max_positions)
+    dummy_batch = task.dataset(args.train_subset).get_dummy_batch(args.max_tokens, max_positions)
+    oom_batch = task.dataset(args.train_subset).get_dummy_batch(1, max_positions)
 
     # Build trainer
     trainer = Trainer(args, task, model, criterion, dummy_batch, oom_batch)
@@ -358,11 +358,9 @@ def load_checkpoint(args, trainer, epoch_itr):
     return False
 
 
-def load_dataset_splits(task, splits):
-    for split in splits:
-        if split == 'train':
-            task.load_dataset(split, combine=True)
-        else:
+def load_dataset_splits(args, task):
+    task.load_dataset(args.train_subset, combine=True)
+    for split in args.valid_subset.split(','):
         for k in itertools.count():
             split_k = split + (str(k) if k > 0 else '')
             try: