Help message for --pad_to_max_length

Helsinki-NLP · May 20, 2024 · bda983a · bda983a
1 parent e331a39
commit bda983a
Showing 1 changed file with 7 additions and 1 deletion.
diff --git a/mammoth/opts.py b/mammoth/opts.py
@@ -617,7 +617,13 @@ def _add_train_general_opts(parser):
         choices=["sents", "tokens"],
         help="Batch grouping for batch_size. Standard is sents. Tokens will do dynamic batching",
     )
-    group.add('--pad_to_max_length', '-pad_to_max_length', action='store_true')
+    group.add(
+        '--pad_to_max_length',
+        '-pad_to_max_length',
+        action='store_true',
+        help='Pad all minibatches to max_length instead of to the length of the longest sequence in the minibatch. '
+        'Using this together with batch_type=sents results in tensors of a fixed shape.'
+    )
     group.add('--max_length', '-max_length', type=int, default=None, help='Maximum sequence length.')
     group.add(
         '--task_distribution_strategy',