Fixing a hard to trigger bug for `text-generation` pipeline. (#18131)

* Fixing a bug where attention mask was not passed to generate.

* Fixing zero-size prompts.

* Comment on top.

src/transformers/pipelines/text_generation.py

@@ -205,14 +205,17 @@ class TextGenerationPipeline(Pipeline):
 
     def _forward(self, model_inputs, **generate_kwargs):
         input_ids = model_inputs["input_ids"]
+        attention_mask = model_inputs.get("attention_mask", None)
         # Allow empty prompts
         if input_ids.shape[1] == 0:
             input_ids = None
+            attention_mask = None
             in_b = 1
         else:
             in_b = input_ids.shape[0]
         prompt_text = model_inputs.pop("prompt_text")
-        generated_sequence = self.model.generate(input_ids=input_ids, **generate_kwargs)  # BS x SL
+        # BS x SL
+        generated_sequence = self.model.generate(input_ids=input_ids, attention_mask=attention_mask, **generate_kwargs)
         out_b = generated_sequence.shape[0]
         if self.framework == "pt":
             generated_sequence = generated_sequence.reshape(in_b, out_b // in_b, *generated_sequence.shape[1:])