Fix generation docstrings regarding input_ids=None (#12823)

src/transformers/generation_flax_utils.py

@@ -190,7 +190,7 @@ def generate(
 
         Parameters:
 
-            input_ids (:obj:`jnp.ndarray` of shape :obj:`(batch_size, sequence_length)`, `optional`):
+            input_ids (:obj:`jnp.ndarray` of shape :obj:`(batch_size, sequence_length)`):
                 The sequence used as a prompt for the generation.
             max_length (:obj:`int`, `optional`, defaults to 20):
                 The maximum length of the sequence to be generated.

src/transformers/generation_tf_utils.py

@@ -400,8 +400,8 @@ def generate(
         Parameters:
 
             input_ids (:obj:`tf.Tensor` of :obj:`dtype=tf.int32` and shape :obj:`(batch_size, sequence_length)`, `optional`):
-                The sequence used as a prompt for the generation. If :obj:`None` the method initializes it as an empty
-                :obj:`tf.Tensor` of shape :obj:`(1,)`.
+                The sequence used as a prompt for the generation. If :obj:`None` the method initializes it with
+                :obj:`bos_token_id` and a batch size of 1.
             max_length (:obj:`int`, `optional`, defaults to 20):
                 The maximum length of the sequence to be generated.
             min_length (:obj:`int`, `optional`, defaults to 10):

src/transformers/generation_utils.py

@@ -692,8 +692,8 @@ def generate(
         Parameters:
 
             input_ids (:obj:`torch.LongTensor` of shape :obj:`(batch_size, sequence_length)`, `optional`):
-                The sequence used as a prompt for the generation. If :obj:`None` the method initializes it as an empty
-                :obj:`torch.LongTensor` of shape :obj:`(1,)`.
+                The sequence used as a prompt for the generation. If :obj:`None` the method initializes it with
+                :obj:`bos_token_id` and a batch size of 1.
             max_length (:obj:`int`, `optional`, defaults to :obj:`model.config.max_length`):
                 The maximum length of the sequence to be generated.
             max_new_tokens (:obj:`int`, `optional`, defaults to None):
@@ -1165,9 +1165,8 @@ def greedy_search(
 
         Parameters:
 
-            input_ids (:obj:`torch.LongTensor` of shape :obj:`(batch_size, sequence_length)`, `optional`):
-                The sequence used as a prompt for the generation. If :obj:`None` the method initializes it as an empty
-                :obj:`torch.LongTensor` of shape :obj:`(1,)`.
+            input_ids (:obj:`torch.LongTensor` of shape :obj:`(batch_size, sequence_length)`):
+                The sequence used as a prompt for the generation.
             logits_processor (:obj:`LogitsProcessorList`, `optional`):
                 An instance of :class:`~transformers.LogitsProcessorList`. List of instances of class derived from
                 :class:`~transformers.LogitsProcessor` used to modify the prediction scores of the language modeling
@@ -1391,9 +1390,8 @@ def sample(
 
         Parameters:
 
-            input_ids (:obj:`torch.LongTensor` of shape :obj:`(batch_size, sequence_length)`, `optional`):
-                The sequence used as a prompt for the generation. If :obj:`None` the method initializes it as an empty
-                :obj:`torch.LongTensor` of shape :obj:`(1,)`.
+            input_ids (:obj:`torch.LongTensor` of shape :obj:`(batch_size, sequence_length)`):
+                The sequence used as a prompt for the generation.
             logits_processor (:obj:`LogitsProcessorList`, `optional`):
                 An instance of :class:`~transformers.LogitsProcessorList`. List of instances of class derived from
                 :class:`~transformers.LogitsProcessor` used to modify the prediction scores of the language modeling
@@ -1632,9 +1630,8 @@ def beam_search(
 
         Parameters:
 
-            input_ids (:obj:`torch.LongTensor` of shape :obj:`(batch_size, sequence_length)`, `optional`):
-                The sequence used as a prompt for the generation. If :obj:`None` the method initializes it as an empty
-                :obj:`torch.LongTensor` of shape :obj:`(1,)`.
+            input_ids (:obj:`torch.LongTensor` of shape :obj:`(batch_size, sequence_length)`):
+                The sequence used as a prompt for the generation.
             beam_scorer (:obj:`BeamScorer`):
                 An derived instance of :class:`~transformers.BeamScorer` that defines how beam hypotheses are
                 constructed, stored and sorted during generation. For more information, the documentation of
@@ -1925,9 +1922,8 @@ def beam_sample(
 
         Parameters:
 
-            input_ids (:obj:`torch.LongTensor` of shape :obj:`(batch_size, sequence_length)`, `optional`):
-                The sequence used as a prompt for the generation. If :obj:`None` the method initializes it as an empty
-                :obj:`torch.LongTensor` of shape :obj:`(1,)`.
+            input_ids (:obj:`torch.LongTensor` of shape :obj:`(batch_size, sequence_length)`):
+                The sequence used as a prompt for the generation.
             beam_scorer (:obj:`BeamScorer`):
                 A derived instance of :class:`~transformers.BeamScorer` that defines how beam hypotheses are
                 constructed, stored and sorted during generation. For more information, the documentation of
@@ -2228,9 +2224,8 @@ def group_beam_search(
 
         Parameters:
 
-            input_ids (:obj:`torch.LongTensor` of shape :obj:`(batch_size, sequence_length)`, `optional`):
-                The sequence used as a prompt for the generation. If :obj:`None` the method initializes it as an empty
-                :obj:`torch.LongTensor` of shape :obj:`(1,)`.
+            input_ids (:obj:`torch.LongTensor` of shape :obj:`(batch_size, sequence_length)`):
+                The sequence used as a prompt for the generation.
             beam_scorer (:obj:`BeamScorer`):
                 An derived instance of :class:`~transformers.BeamScorer` that defines how beam hypotheses are
                 constructed, stored and sorted during generation. For more information, the documentation of