improve saving strategy of sentencepiece tokenizer (#15328)

* add new test

* add a feature to same the sentencepiece tokenizer model when the init file was deleted

* update marian

* update m2m_100

* fix marian

* update speech to text

* override test for layoutxlm

* fix saving bartpho

* remove harcoded values bartpho

* special token string version

* finish bartpho

* override layoutxml test

* add mbart

* move special tokens list

* format

* Revert "format"

This reverts commit 37a40df37903a932c2f951cbd33acb684246bae7.

* simplify list of string of special tokens

* Re-write `self.fairseq_tokens_to_ids ` initialization logic with special tokens
Co-authored-by: Sylvain Gugger <sylvain.gugger@gmail.com>
Co-authored-by: Sylvain Gugger <sylvain.gugger@gmail.com>

src/transformers/models/albert/tokenization_albert.py

@@ -343,7 +343,11 @@ class AlbertTokenizer(PreTrainedTokenizer):
             save_directory, (filename_prefix + "-" if filename_prefix else "") + VOCAB_FILES_NAMES["vocab_file"]
         )
 
-        if os.path.abspath(self.vocab_file) != os.path.abspath(out_vocab_file):
+        if os.path.abspath(self.vocab_file) != os.path.abspath(out_vocab_file) and os.path.isfile(self.vocab_file):
             copyfile(self.vocab_file, out_vocab_file)
+        elif not os.path.isfile(self.vocab_file):
+            with open(out_vocab_file, "wb") as fi:
+                content_spiece_model = self.sp_model.serialized_model_proto()
+                fi.write(content_spiece_model)
 
         return (out_vocab_file,)

src/transformers/models/bartpho/tokenization_bartpho.py

@@ -157,12 +157,20 @@ class BartphoTokenizer(PreTrainedTokenizer):
         self.sp_model.Load(str(vocab_file))
 
         # Load the reduced vocab
-        self.fairseq_tokens_to_ids = {"<s>": 0, "<pad>": 1, "</s>": 2, "<unk>": 3}
+
+        # Keep order of special tokens for backward compatibility
+        self.fairseq_tokens_to_ids = {}
+        cnt = 0
+        for token in [bos_token, pad_token, eos_token, unk_token, sep_token, cls_token]:
+            if str(token) not in self.fairseq_tokens_to_ids:
+                self.fairseq_tokens_to_ids[str(token)] = cnt
+                cnt += 1
         with open(monolingual_vocab_file, "r", encoding="utf-8") as f:
             for line in f.readlines():
                 token = line.strip().split()[0]
                 self.fairseq_tokens_to_ids[token] = len(self.fairseq_tokens_to_ids)
-        self.fairseq_tokens_to_ids["<mask>"] = len(self.fairseq_tokens_to_ids)
+        if str(mask_token) not in self.fairseq_tokens_to_ids:
+            self.fairseq_tokens_to_ids[str(mask_token)] = len(self.fairseq_tokens_to_ids)
 
         self.fairseq_ids_to_tokens = {v: k for k, v in self.fairseq_tokens_to_ids.items()}
 
@@ -278,7 +286,7 @@ class BartphoTokenizer(PreTrainedTokenizer):
         if token in self.fairseq_tokens_to_ids:
             return self.fairseq_tokens_to_ids[token]
         else:
-            return self.fairseq_tokens_to_ids["<unk>"]
+            return self.unk_token_id
 
     def _convert_id_to_token(self, index):
         """Converts an index (integer) in a token (str) using the vocab."""
@@ -301,10 +309,21 @@ class BartphoTokenizer(PreTrainedTokenizer):
             (filename_prefix + "-" if filename_prefix else "") + VOCAB_FILES_NAMES["monolingual_vocab_file"],
         )
 
-        if os.path.abspath(self.vocab_file) != os.path.abspath(out_vocab_file):
+        if os.path.abspath(self.vocab_file) != os.path.abspath(out_vocab_file) and os.path.isfile(self.vocab_file):
             copyfile(self.vocab_file, out_vocab_file)
-
-        if os.path.abspath(self.monolingual_vocab_file) != os.path.abspath(out_monolingual_vocab_file):
+        elif not os.path.isfile(self.vocab_file):
+            with open(out_vocab_file, "wb") as fi:
+                content_spiece_model = self.sp_model.serialized_model_proto()
+                fi.write(content_spiece_model)
+
+        if os.path.abspath(self.monolingual_vocab_file) != os.path.abspath(
+            out_monolingual_vocab_file
+        ) and os.path.isfile(self.monolingual_vocab_file):
             copyfile(self.monolingual_vocab_file, out_monolingual_vocab_file)
+        elif not os.path.isfile(self.monolingual_vocab_file):
+            with open(out_monolingual_vocab_file, "w", encoding="utf-8") as fp:
+                for token in self.fairseq_tokens_to_ids:
+                    if token not in self.all_special_tokens:
+                        fp.write(f"{str(token)} \n")
 
         return out_vocab_file, out_monolingual_vocab_file

src/transformers/models/bert_generation/tokenization_bert_generation.py

@@ -160,7 +160,11 @@ class BertGenerationTokenizer(PreTrainedTokenizer):
             save_directory, (filename_prefix + "-" if filename_prefix else "") + VOCAB_FILES_NAMES["vocab_file"]
         )
 
-        if os.path.abspath(self.vocab_file) != os.path.abspath(out_vocab_file):
+        if os.path.abspath(self.vocab_file) != os.path.abspath(out_vocab_file) and os.path.isfile(self.vocab_file):
             copyfile(self.vocab_file, out_vocab_file)
+        elif not os.path.isfile(self.vocab_file):
+            with open(out_vocab_file, "wb") as fi:
+                content_spiece_model = self.sp_model.serialized_model_proto()
+                fi.write(content_spiece_model)
 
         return (out_vocab_file,)

src/transformers/models/big_bird/tokenization_big_bird.py

@@ -189,8 +189,12 @@ class BigBirdTokenizer(PreTrainedTokenizer):
             save_directory, (filename_prefix + "-" if filename_prefix else "") + VOCAB_FILES_NAMES["vocab_file"]
         )
 
-        if os.path.abspath(self.vocab_file) != os.path.abspath(out_vocab_file):
+        if os.path.abspath(self.vocab_file) != os.path.abspath(out_vocab_file) and os.path.isfile(self.vocab_file):
             copyfile(self.vocab_file, out_vocab_file)
+        elif not os.path.isfile(self.vocab_file):
+            with open(out_vocab_file, "wb") as fi:
+                content_spiece_model = self.sp_model.serialized_model_proto()
+                fi.write(content_spiece_model)
 
         return (out_vocab_file,)
 

src/transformers/models/camembert/tokenization_camembert.py

@@ -288,7 +288,11 @@ class CamembertTokenizer(PreTrainedTokenizer):
             save_directory, (filename_prefix + "-" if filename_prefix else "") + VOCAB_FILES_NAMES["vocab_file"]
         )
 
-        if os.path.abspath(self.vocab_file) != os.path.abspath(out_vocab_file):
+        if os.path.abspath(self.vocab_file) != os.path.abspath(out_vocab_file) and os.path.isfile(self.vocab_file):
             copyfile(self.vocab_file, out_vocab_file)
+        elif not os.path.isfile(self.vocab_file):
+            with open(out_vocab_file, "wb") as fi:
+                content_spiece_model = self.sp_model.serialized_model_proto()
+                fi.write(content_spiece_model)
 
         return (out_vocab_file,)

src/transformers/models/fnet/tokenization_fnet.py

@@ -305,7 +305,11 @@ class FNetTokenizer(PreTrainedTokenizer):
             save_directory, (filename_prefix + "-" if filename_prefix else "") + VOCAB_FILES_NAMES["vocab_file"]
         )
 
-        if os.path.abspath(self.vocab_file) != os.path.abspath(out_vocab_file):
+        if os.path.abspath(self.vocab_file) != os.path.abspath(out_vocab_file) and os.path.isfile(self.vocab_file):
             copyfile(self.vocab_file, out_vocab_file)
+        elif not os.path.isfile(self.vocab_file):
+            with open(out_vocab_file, "wb") as fi:
+                content_spiece_model = self.sp_model.serialized_model_proto()
+                fi.write(content_spiece_model)
 
         return (out_vocab_file,)

src/transformers/models/layoutxlm/tokenization_layoutxlm.py

@@ -331,8 +331,12 @@ class LayoutXLMTokenizer(PreTrainedTokenizer):
             save_directory, (filename_prefix + "-" if filename_prefix else "") + VOCAB_FILES_NAMES["vocab_file"]
         )
 
-        if os.path.abspath(self.vocab_file) != os.path.abspath(out_vocab_file):
+        if os.path.abspath(self.vocab_file) != os.path.abspath(out_vocab_file) and os.path.isfile(self.vocab_file):
             copyfile(self.vocab_file, out_vocab_file)
+        elif not os.path.isfile(self.vocab_file):
+            with open(out_vocab_file, "wb") as fi:
+                content_spiece_model = self.sp_model.serialized_model_proto()
+                fi.write(content_spiece_model)
 
         return (out_vocab_file,)
 

src/transformers/models/m2m_100/tokenization_m2m_100.py

@@ -13,6 +13,7 @@
 # limitations under the License.
 """Tokenization classes for M2M100."""
 import json
+import os
 from contextlib import contextmanager
 from pathlib import Path
 from shutil import copyfile
@@ -312,8 +313,12 @@ class M2M100Tokenizer(PreTrainedTokenizer):
 
         save_json(self.encoder, vocab_save_path)
 
-        if not spm_save_path.exists():
+        if os.path.abspath(self.spm_file) != os.path.abspath(spm_save_path) and os.path.isfile(self.spm_file):
             copyfile(self.spm_file, spm_save_path)
+        elif not os.path.isfile(self.spm_file):
+            with open(spm_save_path, "wb") as fi:
+                content_spiece_model = self.sp_model.serialized_model_proto()
+                fi.write(content_spiece_model)
 
         return (str(vocab_save_path), str(spm_save_path))
 

src/transformers/models/marian/tokenization_marian.py

@@ -11,8 +11,8 @@
 # WITHOUT WARRANTIES OR CONDITIONS OF ANY KIND, either express or implied.
 # See the License for the specific language governing permissions and
 # limitations under the License.
-
 import json
+import os
 import re
 import warnings
 from contextlib import contextmanager
@@ -23,7 +23,10 @@ from typing import Any, Dict, List, Optional, Tuple, Union
 import sentencepiece
 
 from ...tokenization_utils import PreTrainedTokenizer
+from ...utils import logging
+
 
+logger = logging.get_logger(__name__)
 
 VOCAB_FILES_NAMES = {
     "source_spm": "source.spm",
@@ -277,21 +280,35 @@ class MarianTokenizer(PreTrainedTokenizer):
         return len(self.encoder)
 
     def save_vocabulary(self, save_directory: str, filename_prefix: Optional[str] = None) -> Tuple[str]:
-        save_dir = Path(save_directory)
-        assert save_dir.is_dir(), f"{save_directory} should be a directory"
-        save_json(
-            self.encoder,
-            save_dir / ((filename_prefix + "-" if filename_prefix else "") + self.vocab_files_names["vocab"]),
+        if not os.path.isdir(save_directory):
+            logger.error(f"Vocabulary path ({save_directory}) should be a directory")
+            return
+        saved_files = []
+        out_vocab_file = os.path.join(
+            save_directory, (filename_prefix + "-" if filename_prefix else "") + VOCAB_FILES_NAMES["vocab"]
         )
 
-        for orig, f in zip(["source.spm", "target.spm"], self.spm_files):
-            dest_path = save_dir / ((filename_prefix + "-" if filename_prefix else "") + Path(f).name)
-            if not dest_path.exists():
-                copyfile(f, save_dir / orig)
+        save_json(self.encoder, out_vocab_file)
+        saved_files.append(out_vocab_file)
 
-        return tuple(
-            save_dir / ((filename_prefix + "-" if filename_prefix else "") + f) for f in self.vocab_files_names
+        for spm_save_filename, spm_orig_path, spm_model in zip(
+            [VOCAB_FILES_NAMES["source_spm"], VOCAB_FILES_NAMES["target_spm"]],
+            self.spm_files,
+            [self.spm_source, self.spm_target],
+        ):
+            spm_save_path = os.path.join(
+                save_directory, (filename_prefix + "-" if filename_prefix else "") + spm_save_filename
             )
+            if os.path.abspath(spm_orig_path) != os.path.abspath(spm_save_path) and os.path.isfile(spm_orig_path):
+                copyfile(spm_orig_path, spm_save_path)
+                saved_files.append(spm_save_path)
+            elif not os.path.isfile(spm_orig_path):
+                with open(spm_save_path, "wb") as fi:
+                    content_spiece_model = spm_model.serialized_model_proto()
+                    fi.write(content_spiece_model)
+                saved_files.append(spm_save_path)
+
+        return tuple(saved_files)
 
     def get_vocab(self) -> Dict:
         vocab = self.encoder.copy()

src/transformers/models/mbart/tokenization_mbart.py

@@ -315,8 +315,12 @@ class MBartTokenizer(PreTrainedTokenizer):
             save_directory, (filename_prefix + "-" if filename_prefix else "") + VOCAB_FILES_NAMES["vocab_file"]
         )
 
-        if os.path.abspath(self.vocab_file) != os.path.abspath(out_vocab_file):
+        if os.path.abspath(self.vocab_file) != os.path.abspath(out_vocab_file) and os.path.isfile(self.vocab_file):
             copyfile(self.vocab_file, out_vocab_file)
+        elif not os.path.isfile(self.vocab_file):
+            with open(out_vocab_file, "wb") as fi:
+                content_spiece_model = self.sp_model.serialized_model_proto()
+                fi.write(content_spiece_model)
 
         return (out_vocab_file,)
 

src/transformers/models/mbart50/tokenization_mbart50.py

@@ -245,8 +245,12 @@ class MBart50Tokenizer(PreTrainedTokenizer):
             save_directory, (filename_prefix + "-" if filename_prefix else "") + VOCAB_FILES_NAMES["vocab_file"]
         )
 
-        if os.path.abspath(self.vocab_file) != os.path.abspath(out_vocab_file):
+        if os.path.abspath(self.vocab_file) != os.path.abspath(out_vocab_file) and os.path.isfile(self.vocab_file):
             copyfile(self.vocab_file, out_vocab_file)
+        elif not os.path.isfile(self.vocab_file):
+            with open(out_vocab_file, "wb") as fi:
+                content_spiece_model = self.sp_model.serialized_model_proto()
+                fi.write(content_spiece_model)
 
         return (out_vocab_file,)
 

src/transformers/models/pegasus/tokenization_pegasus.py

@@ -285,7 +285,11 @@ class PegasusTokenizer(PreTrainedTokenizer):
             save_directory, (filename_prefix + "-" if filename_prefix else "") + VOCAB_FILES_NAMES["vocab_file"]
         )
 
-        if os.path.abspath(self.vocab_file) != os.path.abspath(out_vocab_file):
+        if os.path.abspath(self.vocab_file) != os.path.abspath(out_vocab_file) and os.path.isfile(self.vocab_file):
             copyfile(self.vocab_file, out_vocab_file)
+        elif not os.path.isfile(self.vocab_file):
+            with open(out_vocab_file, "wb") as fi:
+                content_spiece_model = self.sp_model.serialized_model_proto()
+                fi.write(content_spiece_model)
 
         return (out_vocab_file,)

src/transformers/models/reformer/tokenization_reformer.py

@@ -167,7 +167,11 @@ class ReformerTokenizer(PreTrainedTokenizer):
             save_directory, (filename_prefix + "-" if filename_prefix else "") + VOCAB_FILES_NAMES["vocab_file"]
         )
 
-        if os.path.abspath(self.vocab_file) != os.path.abspath(out_vocab_file):
+        if os.path.abspath(self.vocab_file) != os.path.abspath(out_vocab_file) and os.path.isfile(self.vocab_file):
             copyfile(self.vocab_file, out_vocab_file)
+        elif not os.path.isfile(self.vocab_file):
+            with open(out_vocab_file, "wb") as fi:
+                content_spiece_model = self.sp_model.serialized_model_proto()
+                fi.write(content_spiece_model)
 
         return (out_vocab_file,)

src/transformers/models/speech_to_text/tokenization_speech_to_text.py

@@ -13,8 +13,8 @@
 # See the License for the specific language governing permissions and
 # limitations under the License.
 """Tokenization classes for Speech2Text."""
-
 import json
+import os
 from pathlib import Path
 from shutil import copyfile
 from typing import Any, Dict, List, Optional, Tuple, Union
@@ -260,8 +260,12 @@ class Speech2TextTokenizer(PreTrainedTokenizer):
 
         save_json(self.encoder, vocab_save_path)
 
-        if not spm_save_path.exists():
+        if os.path.abspath(self.spm_file) != os.path.abspath(spm_save_path) and os.path.isfile(self.spm_file):
             copyfile(self.spm_file, spm_save_path)
+        elif not os.path.isfile(self.spm_file):
+            with open(spm_save_path, "wb") as fi:
+                content_spiece_model = self.sp_model.serialized_model_proto()
+                fi.write(content_spiece_model)
 
         return (str(vocab_save_path), str(spm_save_path))
 

src/transformers/models/t5/tokenization_t5.py

@@ -303,8 +303,11 @@ class T5Tokenizer(PreTrainedTokenizer):
             save_directory, (filename_prefix + "-" if filename_prefix else "") + VOCAB_FILES_NAMES["vocab_file"]
         )
 
-        if os.path.abspath(self.vocab_file) != os.path.abspath(out_vocab_file):
+        if os.path.abspath(self.vocab_file) != os.path.abspath(out_vocab_file) and os.path.isfile(self.vocab_file):
             copyfile(self.vocab_file, out_vocab_file)
-            logger.info(f"Copy vocab file to {out_vocab_file}")
+        elif not os.path.isfile(self.vocab_file):
+            with open(out_vocab_file, "wb") as fi:
+                content_spiece_model = self.sp_model.serialized_model_proto()
+                fi.write(content_spiece_model)
 
         return (out_vocab_file,)

src/transformers/models/xlm_prophetnet/tokenization_xlm_prophetnet.py

@@ -302,8 +302,12 @@ class XLMProphetNetTokenizer(PreTrainedTokenizer):
             save_directory, (filename_prefix + "-" if filename_prefix else "") + VOCAB_FILES_NAMES["vocab_file"]
         )
 
-        if os.path.abspath(self.vocab_file) != os.path.abspath(out_vocab_file):
+        if os.path.abspath(self.vocab_file) != os.path.abspath(out_vocab_file) and os.path.isfile(self.vocab_file):
             copyfile(self.vocab_file, out_vocab_file)
+        elif not os.path.isfile(self.vocab_file):
+            with open(out_vocab_file, "wb") as fi:
+                content_spiece_model = self.sp_model.serialized_model_proto()
+                fi.write(content_spiece_model)
 
         return (out_vocab_file,)
 

src/transformers/models/xlm_roberta/tokenization_xlm_roberta.py

@@ -310,7 +310,11 @@ class XLMRobertaTokenizer(PreTrainedTokenizer):
             save_directory, (filename_prefix + "-" if filename_prefix else "") + VOCAB_FILES_NAMES["vocab_file"]
         )
 
-        if os.path.abspath(self.vocab_file) != os.path.abspath(out_vocab_file):
+        if os.path.abspath(self.vocab_file) != os.path.abspath(out_vocab_file) and os.path.isfile(self.vocab_file):
             copyfile(self.vocab_file, out_vocab_file)
+        elif not os.path.isfile(self.vocab_file):
+            with open(out_vocab_file, "wb") as fi:
+                content_spiece_model = self.sp_model.serialized_model_proto()
+                fi.write(content_spiece_model)
 
         return (out_vocab_file,)

src/transformers/models/xlnet/tokenization_xlnet.py

@@ -342,7 +342,11 @@ class XLNetTokenizer(PreTrainedTokenizer):
             save_directory, (filename_prefix + "-" if filename_prefix else "") + VOCAB_FILES_NAMES["vocab_file"]
         )
 
-        if os.path.abspath(self.vocab_file) != os.path.abspath(out_vocab_file):
+        if os.path.abspath(self.vocab_file) != os.path.abspath(out_vocab_file) and os.path.isfile(self.vocab_file):
             copyfile(self.vocab_file, out_vocab_file)
+        elif not os.path.isfile(self.vocab_file):
+            with open(out_vocab_file, "wb") as fi:
+                content_spiece_model = self.sp_model.serialized_model_proto()
+                fi.write(content_spiece_model)
 
         return (out_vocab_file,)

tests/test_tokenization_common.py

@@ -394,6 +394,33 @@ class TokenizerTesterMixin:
         self.assertEqual(tokenizer_new.sp_model_kwargs, sp_model_kwargs)
         self.check_subword_sampling(tokenizer_new)
 
+    def test_save_sentencepiece_tokenizer(self) -> None:
+        if not self.test_sentencepiece or not self.test_slow_tokenizer:
+            return
+        # We want to verify that we will be able to save the tokenizer even if the original files that were used to
+        # build the tokenizer have been deleted in the meantime.
+        text = "This is text to test the tokenizer."
+
+        tokenizer_slow_1 = self.get_tokenizer()
+        encoding_tokenizer_slow_1 = tokenizer_slow_1(text)
+
+        tmpdirname_1 = tempfile.mkdtemp()
+        tmpdirname_2 = tempfile.mkdtemp()
+
+        tokenizer_slow_1.save_pretrained(tmpdirname_1)
+        tokenizer_slow_2 = self.tokenizer_class.from_pretrained(tmpdirname_1)
+        encoding_tokenizer_slow_2 = tokenizer_slow_2(text)
+
+        shutil.rmtree(tmpdirname_1)
+        tokenizer_slow_2.save_pretrained(tmpdirname_2)
+
+        tokenizer_slow_3 = self.tokenizer_class.from_pretrained(tmpdirname_2)
+        encoding_tokenizer_slow_3 = tokenizer_slow_3(text)
+        shutil.rmtree(tmpdirname_2)
+
+        self.assertEqual(encoding_tokenizer_slow_1, encoding_tokenizer_slow_2)
+        self.assertEqual(encoding_tokenizer_slow_1, encoding_tokenizer_slow_3)
+
     def test_model_input_names_signature(self):
         accepted_model_main_input_names = [
             "input_ids",  # nlp models

tests/test_tokenization_layoutxlm.py

@@ -99,6 +99,44 @@ class LayoutXLMTokenizationTest(TokenizerTesterMixin, unittest.TestCase):
         output_text = "unwanted, running"
         return input_text, output_text
 
+    # override test in `test_tokenization_common.py` because of the required input format of the `__call__`` method of
+    # this tokenizer
+    def test_save_sentencepiece_tokenizer(self) -> None:
+        if not self.test_sentencepiece or not self.test_slow_tokenizer:
+            return
+        # We want to verify that we will be able to save the tokenizer even if the original files that were used to
+        # build the tokenizer have been deleted in the meantime.
+        words, boxes = self.get_words_and_boxes()
+
+        tokenizer_slow_1 = self.get_tokenizer()
+        encoding_tokenizer_slow_1 = tokenizer_slow_1(
+            words,
+            boxes=boxes,
+        )
+
+        tmpdirname_1 = tempfile.mkdtemp()
+        tmpdirname_2 = tempfile.mkdtemp()
+
+        tokenizer_slow_1.save_pretrained(tmpdirname_1)
+        tokenizer_slow_2 = self.tokenizer_class.from_pretrained(tmpdirname_1)
+        encoding_tokenizer_slow_2 = tokenizer_slow_2(
+            words,
+            boxes=boxes,
+        )
+
+        shutil.rmtree(tmpdirname_1)
+        tokenizer_slow_2.save_pretrained(tmpdirname_2)
+
+        tokenizer_slow_3 = self.tokenizer_class.from_pretrained(tmpdirname_2)
+        encoding_tokenizer_slow_3 = tokenizer_slow_3(
+            words,
+            boxes=boxes,
+        )
+        shutil.rmtree(tmpdirname_2)
+
+        self.assertEqual(encoding_tokenizer_slow_1, encoding_tokenizer_slow_2)
+        self.assertEqual(encoding_tokenizer_slow_1, encoding_tokenizer_slow_3)
+
     @slow
     def test_sequence_builders(self):
         tokenizer = self.tokenizer_class.from_pretrained("microsoft/layoutxlm-base")