Generate: throw warning when return_dict_in_generate is False but s…

…hould be True (#33146)

src/transformers/generation/configuration_utils.py

@@ -288,7 +288,9 @@ class GenerationConfig(PushToHubMixin):
             Whether or not to return the unprocessed prediction logit scores. See `logits` under returned tensors for
             more details.
         return_dict_in_generate (`bool`, *optional*, defaults to `False`):
-            Whether or not to return a [`~utils.ModelOutput`] instead of a plain tuple.
+            Whether or not to return a [`~utils.ModelOutput`], as opposed to returning exclusively the generated
+            sequence. This flag must be set to `True` to return the generation cache (when `use_cache` is `True`)
+            or optional outputs (see flags starting with `output_`)
 
         > Special tokens that can be used at generation time
 
@@ -334,6 +336,8 @@ class GenerationConfig(PushToHubMixin):
             present in `generate`'s signature will be used in the model forward pass.
     """
 
+    extra_output_flags = ("output_attentions", "output_hidden_states", "output_scores", "output_logits")
+
     def __init__(self, **kwargs):
         # Parameters that control the length of the output
         self.max_length = kwargs.pop("max_length", 20)
@@ -727,7 +731,17 @@ def validate(self, is_init=False):
                 self.watermarking_config = WatermarkingConfig.from_dict(self.watermarking_config)
             self.watermarking_config.validate()
 
-        # 7. check common issue: passing `generate` arguments inside the generation config
+        # 7. other incorrect combinations
+        if self.return_dict_in_generate is not True:
+            for extra_output_flag in self.extra_output_flags:
+                if getattr(self, extra_output_flag) is True:
+                    warnings.warn(
+                        f"`return_dict_in_generate` is NOT set to `True`, but `{extra_output_flag}` is. When "
+                        f"`return_dict_in_generate` is not `True`, `{extra_output_flag}` is ignored.",
+                        UserWarning,
+                    )
+
+        # 8. check common issue: passing `generate` arguments inside the generation config
         generate_arguments = (
             "logits_processor",
             "stopping_criteria",
@@ -786,7 +800,8 @@ def save_pretrained(
 
         if use_auth_token is not None:
             warnings.warn(
-                "The `use_auth_token` argument is deprecated and will be removed in v5 of Transformers. Please use `token` instead.",
+                "The `use_auth_token` argument is deprecated and will be removed in v5 of Transformers. "
+                "Please use `token` instead.",
                 FutureWarning,
             )
             if kwargs.get("token", None) is not None:
@@ -1189,6 +1204,11 @@ def from_model_config(cls, model_config: PretrainedConfig) -> "GenerationConfig"
                     if attr in decoder_config and getattr(config, attr) == getattr(default_generation_config, attr):
                         setattr(config, attr, decoder_config[attr])
 
+        # If any `output_...` flag is set to `True`, we ensure `return_dict_in_generate` is set to `True`.
+        if config.return_dict_in_generate is False:
+            if any(getattr(config, extra_output_flag, False) for extra_output_flag in config.extra_output_flags):
+                config.return_dict_in_generate = True
+
         config._original_object_hash = hash(config)  # Hash to detect whether the instance was modified
         return config
 

tests/generation/test_configuration_utils.py

@@ -136,6 +136,10 @@ def test_validate(self):
             GenerationConfig(do_sample=False, temperature=0.5)
         self.assertEqual(len(captured_warnings), 1)
 
+        with warnings.catch_warnings(record=True) as captured_warnings:
+            GenerationConfig(return_dict_in_generate=False, output_scores=True)
+        self.assertEqual(len(captured_warnings), 1)
+
         # Expanding on the case above, we can update a bad configuration to get rid of the warning. Ideally,
         # that is done by unsetting the parameter (i.e. setting it to None)
         generation_config_bad_temperature = GenerationConfig(do_sample=False, temperature=0.5)