Fix weight loading issue (#14016)

Co-authored-by: ydshieh <ydshieh@users.noreply.github.com>

src/transformers/models/encoder_decoder/modeling_tf_encoder_decoder.py

@@ -15,6 +15,7 @@
 """ Classes to support TF Encoder-Decoder architectures """
 
 
+import tempfile
 from typing import Optional
 
 import tensorflow as tf
@@ -254,6 +255,11 @@ def from_pretrained(cls, pretrained_model_name_or_path, *model_args, **kwargs):
             >>> # This is only for copying some specific attributes of this particular model.
             >>> model.config = _model.config
 
+        Example::
+
+            >>> from transformers import TFEncoderDecoderModel
+            >>> model = TFEncoderDecoderModel.from_pretrained("ydshieh/bert2bert-cnn_dailymail-fp16")
+
         """
 
         from_pt = kwargs.pop("from_pt", False)
@@ -369,6 +375,14 @@ def from_encoder_decoder_pretrained(
             kwargs_encoder["load_weight_prefix"] = cls.load_weight_prefix
             encoder = TFAutoModel.from_pretrained(encoder_pretrained_model_name_or_path, *model_args, **kwargs_encoder)
 
+            # This is necessary to make `from_pretrained` following `save_pretrained` work correctly
+            if kwargs_encoder.get("from_pt", None):
+                del kwargs_encoder["from_pt"]
+                with tempfile.TemporaryDirectory() as tmp_dirname:
+                    encoder.save_pretrained(tmp_dirname)
+                    del encoder
+                    encoder = TFAutoModel.from_pretrained(tmp_dirname, *model_args, **kwargs_encoder)
+
         decoder = kwargs_decoder.pop("model", None)
         if decoder is None:
             if decoder_pretrained_model_name_or_path is None:
@@ -397,6 +411,14 @@ def from_encoder_decoder_pretrained(
             kwargs_decoder["load_weight_prefix"] = cls.load_weight_prefix
             decoder = TFAutoModelForCausalLM.from_pretrained(decoder_pretrained_model_name_or_path, **kwargs_decoder)
 
+            # This is necessary to make `from_pretrained` following `save_pretrained` work correctly
+            if kwargs_decoder.get("from_pt", None):
+                del kwargs_decoder["from_pt"]
+                with tempfile.TemporaryDirectory() as tmp_dirname:
+                    decoder.save_pretrained(tmp_dirname)
+                    del decoder
+                    decoder = TFAutoModelForCausalLM.from_pretrained(tmp_dirname, **kwargs_decoder)
+
         # Make sure these 2 `tf.keras.Model` have fixed names so `from_pretrained` could load model weights correctly.
         if encoder.name != "encoder":
             raise ValueError("encoder model must be created with the name `encoder`.")

tests/test_modeling_tf_encoder_decoder.py

@@ -457,6 +457,14 @@ def test_bert2bert_summarization(self):
 
         self.assertEqual(summary, [EXPECTED_SUMMARY_STUDENTS])
 
+        # Test with the TF checkpoint
+        model = TFEncoderDecoderModel.from_pretrained("ydshieh/bert2bert-cnn_dailymail-fp16")
+
+        output_ids = model.generate(input_ids=input_dict["input_ids"], max_length=None).numpy().tolist()
+        summary = tokenizer.batch_decode(output_ids, skip_special_tokens=True)
+
+        self.assertEqual(summary, [EXPECTED_SUMMARY_STUDENTS])
+
 
 @require_tf
 class TFGPT2EncoderDecoderModelTest(TFEncoderDecoderMixin, unittest.TestCase):
@@ -785,6 +793,16 @@ def test_encoder_decoder_save_load_from_encoder_decoder_from_pt(self):
         max_diff = np.max(np.abs(logits_pt.detach().cpu().numpy() - logits_tf.numpy()))
         self.assertAlmostEqual(max_diff, 0.0, places=3)
 
+        # Make sure `from_pretrained` following `save_pretrained` work and give the same result
+        with tempfile.TemporaryDirectory() as tmp_dirname:
+            encoder_decoder_tf.save_pretrained(tmp_dirname)
+            encoder_decoder_tf = TFEncoderDecoderModel.from_pretrained(tmp_dirname)
+
+            logits_tf_2 = encoder_decoder_tf(input_ids=input_ids, decoder_input_ids=decoder_input_ids).logits
+
+            max_diff = np.max(np.abs(logits_tf_2.numpy() - logits_tf.numpy()))
+            self.assertAlmostEqual(max_diff, 0.0, places=3)
+
         # TensorFlow => PyTorch
         with tempfile.TemporaryDirectory() as tmp_dirname:
             encoder_decoder_tf.save_pretrained(tmp_dirname)