wiki checkpoint best

config.json

@@ -1,4 +1,5 @@
 {
+  "_name_or_path": "ru-longformer-tiny-16384-new_texts-20epoch-2",
   "architectures": [
     "LongformerForMaskedLM"
   ],
@@ -25,7 +26,7 @@
   "position_embedding_type": "absolute",
   "sep_token_id": 2,
   "torch_dtype": "float32",
-  "transformers_version": "4.29.2",
+  "transformers_version": "4.31.0",
   "type_vocab_size": 2,
   "vocab_size": 83830
 }

pytorch_model.bin

@@ -1,3 +1,3 @@
 version https://git-lfs.github.com/spec/v1
-oid sha256:1ea56eb49abad0501a5f7fd33f2c6e822f025d65a891f62c7cb9583d7e451c0e
+oid sha256:62971f83771ea2494667046f749fd57156e0c30ccc2cc333c06c7240b1a36a31
 size 138548629

special_tokens_map.json

@@ -2,7 +2,13 @@
   "bos_token": "<s>",
   "cls_token": "[CLS]",
   "eos_token": "</s>",
-  "mask_token": "[MASK]",
+  "mask_token": {
+    "content": "[MASK]",
+    "lstrip": true,
+    "normalized": false,
+    "rstrip": false,
+    "single_word": false
+  },
   "pad_token": "[PAD]",
   "sep_token": "[SEP]",
   "unk_token": "[UNK]"

tokenizer.json

@@ -2,12 +2,14 @@
   "version": "1.0",
   "truncation": {
     "direction": "Right",
-    "max_length": 512,
+    "max_length": 16384,
     "strategy": "LongestFirst",
     "stride": 0
   },
   "padding": {
-    "strategy": "BatchLongest",
+    "strategy": {
+      "Fixed": 16384
+    },
     "direction": "Right",
     "pad_to_multiple_of": null,
     "pad_id": 0,
@@ -55,7 +57,7 @@
       "id": 4,
       "content": "[MASK]",
       "single_word": false,
-      "lstrip": false,
+      "lstrip": true,
       "rstrip": false,
       "normalized": false,
       "special": true

tokenizer_config.json

@@ -1,8 +1,12 @@
 {
+  "add_prefix_space": false,
+  "bos_token": "<s>",
   "clean_up_tokenization_spaces": true,
   "cls_token": "[CLS]",
   "do_basic_tokenize": true,
   "do_lower_case": false,
+  "eos_token": "</s>",
+  "errors": "replace",
   "mask_token": "[MASK]",
   "model_max_length": 16384,
   "never_split": null,
@@ -10,6 +14,7 @@
   "sep_token": "[SEP]",
   "strip_accents": null,
   "tokenize_chinese_chars": true,
-  "tokenizer_class": "BertTokenizer",
+  "tokenizer_class": "LongformerTokenizer",
+  "trim_offsets": true,
   "unk_token": "[UNK]"
 }