Freeze to Transformers 4.0

Remove warnings + errors caused by deprecations

.gitignore

@@ -1,3 +1,4 @@
+
 # secrets detector
 .secrets.baseline
 
@@ -123,6 +124,7 @@ examples/data.xlsx
 # custom folders
 docs_source/_build/
 docs_source/_modules/
+notebooks/ 
 
 .vscode
 _archive

requirements.txt

@@ -1,4 +1,4 @@
-transformers>=2.5.0
-torch>=1.3.0
+transformers==4.0.0
+torch==1.5.0
 scipy==1.4.1
-pandas==1.0.3
+pandas==1.0.3

rxnmapper/core.py

@@ -114,13 +114,11 @@ def convert_batch_to_attns(
 
         encoded_ids = self.tokenizer.batch_encode_plus(
             rxn_smiles_list,
-            pad_to_max_length=True,
+            padding=True,
             return_tensors="pt",
-            return_special_tokens_masks=True,
         )
-        parsed_input = {
-            k: encoded_ids[k].to(self.device)
-            for k in ["input_ids", "token_type_ids", "attention_mask"]
+        parsed_input = { 
+            k: v.to(self.device) for k, v in encoded_ids.items()
         }
         with torch.no_grad():
             output = self.model(**parsed_input)

rxnmapper/tokenization_smiles.py

@@ -41,9 +41,6 @@ def __init__(
             **vocab_file**: Path to a SMILES character per line vocabulary file
         """
         super().__init__(vocab_file, **kwargs)
-        # take into account special tokens in max length
-        self.max_len_single_sentence = self.max_len - 2
-        self.max_len_sentences_pair = self.max_len - 3
 
         if not os.path.isfile(vocab_file):
             raise ValueError(
@@ -60,7 +57,7 @@ def __init__(
             [(ids, tok) for tok, ids in self.vocab.items()]
         )
         self.basic_tokenizer = BasicSmilesTokenizer()
-        self.init_kwargs["max_len"] = self.max_len
+        self.init_kwargs["model_max_length"] = self.model_max_length
 
     @property
     def vocab_size(self):