Patch-past-refactor (#21050)

* small patches, forgot a line

* refactor PT

* the actual fix

src/transformers/models/vision_encoder_decoder/modeling_tf_vision_encoder_decoder.py

@@ -729,13 +729,11 @@ class TFVisionEncoderDecoderModel(TFPreTrainedModel, TFCausalLanguageModelingLos
         )
 
     def prepare_inputs_for_generation(
-        self, input_ids, past=None, attention_mask=None, use_cache=None, encoder_outputs=None, **kwargs
+        self, input_ids, past_key_values=None, attention_mask=None, use_cache=None, encoder_outputs=None, **kwargs
     ):
-        decoder_inputs = self.decoder.prepare_inputs_for_generation(input_ids, past=past)
+        decoder_inputs = self.decoder.prepare_inputs_for_generation(input_ids, past_key_values=past_key_values)
         decoder_attention_mask = decoder_inputs["attention_mask"] if "attention_mask" in decoder_inputs else None
         past_key_values = decoder_inputs.get("past_key_values")
-        if past_key_values is None:
-            past_key_values = decoder_inputs.get("past")  # e.g. on TF GPT2
         input_dict = {
             "pixel_values": None,  # needs to be passed to make Keras.layer.__call__ happy
             "attention_mask": attention_mask,

src/transformers/models/vision_encoder_decoder/modeling_vision_encoder_decoder.py

@@ -649,9 +649,9 @@ class VisionEncoderDecoderModel(PreTrainedModel):
         return shift_tokens_right(labels, self.config.pad_token_id, self.config.decoder_start_token_id)
 
     def prepare_inputs_for_generation(
-        self, input_ids, past=None, attention_mask=None, use_cache=None, encoder_outputs=None, **kwargs
+        self, input_ids, past_key_values=None, attention_mask=None, use_cache=None, encoder_outputs=None, **kwargs
     ):
-        decoder_inputs = self.decoder.prepare_inputs_for_generation(input_ids, past=past)
+        decoder_inputs = self.decoder.prepare_inputs_for_generation(input_ids, past_key_values=past_key_values)
         decoder_attention_mask = decoder_inputs["attention_mask"] if "attention_mask" in decoder_inputs else None
         input_dict = {
             "attention_mask": attention_mask,

templates/adding_a_new_model/cookiecutter-template-{{cookiecutter.modelname}}/modeling_{{cookiecutter.lowercase_modelname}}.py

@@ -3333,7 +3333,7 @@ class {{cookiecutter.camelcase_modelname}}ForCausalLM({{cookiecutter.camelcase_m
         if attention_mask is None:
             attention_mask = input_ids.new_ones(input_ids.shape)
 
-        if past:
+        if past_key_values:
             input_ids = input_ids[:, -1:]
         # first step, decoder_cached_states are empty
         return {

tests/pipelines/test_pipelines_image_to_text.py

@@ -55,7 +55,6 @@ class ImageToTextPipelineTests(unittest.TestCase, metaclass=PipelineTestCaseMeta
         )
 
     @require_tf
-    @unittest.skip("Arthur will fix me!")
     def test_small_model_tf(self):
         pipe = pipeline("image-to-text", model="hf-internal-testing/tiny-random-vit-gpt2", framework="tf")
         image = "./tests/fixtures/tests_samples/COCO/000000039769.png"