New version. Staticmethods need to return a IntoPy<PyObject>

which is non trivial for PyClassInitializer. Instead I added a lower
staticmethod that returns raw objects, and the `from_file(s)` methods
are implemented directly in Python.

bindings/python/py_src/tokenizers/implementations/bert_wordpiece.py

@@ -80,6 +80,10 @@ class BertWordPieceTokenizer(BaseTokenizer):
 
         super().__init__(tokenizer, parameters)
 
+    def from_file(vocab: str, **kwargs):
+        vocab = WordPiece.read_file(vocab)
+        return BertWordPieceTokenizer(vocab, **kwargs)
+
     def train(
         self,
         files: Union[str, List[str]],

bindings/python/py_src/tokenizers/implementations/byte_level_bpe.py

@@ -77,6 +77,11 @@ class ByteLevelBPETokenizer(BaseTokenizer):
 
         super().__init__(tokenizer, parameters)
 
+    @staticmethod
+    def from_files(vocab_filename: str, merges_filename: str, **kwargs):
+        vocab, merges = BPE.read_files(vocab_filename, merges_filename)
+        return ByteLevelBPETokenizer(vocab, merges, **kwargs)
+
     def train(
         self,
         files: Union[str, List[str]],

bindings/python/py_src/tokenizers/implementations/char_level_bpe.py

@@ -94,6 +94,11 @@ class CharBPETokenizer(BaseTokenizer):
 
         super().__init__(tokenizer, parameters)
 
+    @staticmethod
+    def from_files(vocab_filename: str, merges_filename: str, **kwargs):
+        vocab, merges = BPE.read_files(vocab_filename, merges_filename)
+        return CharBPETokenizer(vocab, merges, **kwargs)
+
     def train(
         self,
         files: Union[str, List[str]],

bindings/python/py_src/tokenizers/implementations/sentencepiece_bpe.py

@@ -47,6 +47,11 @@ class SentencePieceBPETokenizer(BaseTokenizer):
 
         super().__init__(tokenizer, parameters)
 
+    @staticmethod
+    def from_files(vocab_filename: str, merges_filename: str, **kwargs):
+        vocab, merges = BPE.read_files(vocab_filename, merges_filename)
+        return SentencePieceBPETokenizer(vocab, merges, **kwargs)
+
     def train(
         self,
         files: Union[str, List[str]],

bindings/python/py_src/tokenizers/models/__init__.pyi

@@ -62,8 +62,6 @@ class BPE(Model):
        fuse_unk: (`optional`) bool:
            Multiple unk tokens get fused into only 1
     """
-
-    @staticmethod
     def __init__(
         self,
         vocab: Optional[Union[str, Dict[str, int]]],
@@ -77,6 +75,15 @@ class BPE(Model):
     ):
         pass
 
+    @staticmethod
+    def read_files(vocab_filename: str, merges_filename: str) -> Tuple[Vocab, Merges]:
+        pass
+
+    @staticmethod
+    def from_files(vocab_filename: str, merges_filename: str, **kwargs) -> BPE:
+        vocab, merges = BPE.read_files(vocab_filename, merges_filename) 
+        return BPE(vocab, merges, **kwargs)
+
 class WordPiece(Model):
     """ WordPiece model class
 
@@ -101,6 +108,15 @@ class WordPiece(Model):
     ):
         pass
 
+    @staticmethod
+    def read_file(vocab_filename: str) -> Tuple[Vocab]:
+        pass
+
+    @staticmethod
+    def from_files(vocab_filename: str, **kwargs) -> WordPiece:
+        vocab = WordPiece.read_files(vocab_filename) 
+        return WordPiece(vocab, **kwargs)
+
 class WordLevel(Model):
     """
     Most simple tokenizer model based on mapping token from a vocab file to their corresponding id.
@@ -118,6 +134,15 @@ class WordLevel(Model):
     def __init__(self, vocab: Optional[Union[str, Dict[str, int]]], unk_token: Optional[str]):
         pass
 
+    @staticmethod
+    def read_file(vocab_filename: str) -> Tuple[Vocab]:
+        pass
+
+    @staticmethod
+    def from_files(vocab_filename: str, **kwargs) -> WordLevel:
+        vocab = WordLevel.read_files(vocab_filename) 
+        return WordLevel(vocab, **kwargs)
+
 class Unigram(Model):
     """UnigramEncoding model class