Correct eos_token_id settings in generate (#15403)

* Correct eos_token_id set in generate

* Set eos_token_id in test

* Correct eos_token_id set in generate

* Set eos_token_id in test

src/transformers/generation_utils.py

@@ -1049,6 +1049,8 @@ class GenerationMixin:
 
         pad_token_id = pad_token_id if pad_token_id is not None else self.config.pad_token_id
         eos_token_id = eos_token_id if eos_token_id is not None else self.config.eos_token_id
+        if eos_token_id is None and hasattr(self.config, "decoder"):
+            eos_token_id = self.config.decoder.eos_token_id
 
         output_scores = output_scores if output_scores is not None else self.config.output_scores
         output_attentions = output_attentions if output_attentions is not None else self.config.output_attentions

tests/test_modeling_encoder_decoder.py

@@ -395,6 +395,9 @@ class EncoderDecoderMixin:
     def check_encoder_decoder_model_generate(self, input_ids, config, decoder_config, **kwargs):
         encoder_model, decoder_model = self.get_encoder_decoder_model(config, decoder_config)
         enc_dec_model = EncoderDecoderModel(encoder=encoder_model, decoder=decoder_model)
+
+        # Generate until max length
+        enc_dec_model.config.decoder.eos_token_id = None
         enc_dec_model.to(torch_device)
 
         # Bert does not have a bos token id, so use pad_token_id instead