upd weights: 99% ria, title=36

README.md

@@ -12,7 +12,7 @@ license: MIT
 
 ## Description
 *bert2bert* model, initialized with the `DeepPavlov/rubert-base-cased` pretrained weights and 
-   fine-tuned on the first 90% of ["Rossiya Segodnya" news dataset](https://github.com/RossiyaSegodnya/ria_news_dataset) for 3 epochs.
+   fine-tuned on the first 99% of ["Rossiya Segodnya" news dataset](https://github.com/RossiyaSegodnya/ria_news_dataset) for 2 epochs.
 
 ## Usage example
 
@@ -35,7 +35,7 @@ encoded_batch = tokenizer.prepare_seq2seq_batch(
 
 output_ids = model.generate(
     input_ids=encoded_batch["input_ids"],
-    max_length=32,
+    max_length=36,
     no_repeat_ngram_size=3,
     num_beams=5,
     top_k=0
@@ -80,7 +80,6 @@ python nlp_headline_rus/src/train_seq2seq.py \
 
 ## Validation results
 
-- Using [last 1% of ria](https://drive.google.com/drive/folders/1xtCnkbGNNu5jGQ9H9Mg55Cx7RTcyhQw9) dataset
-- Using [last 10% of ria](https://drive.google.com/drive/folders/1w6rAXhpFUO8I4A7xfHKUjMBPEKBHEO3h) dataset
-- Using [gazeta_ru test](https://drive.google.com/drive/folders/185ALuNVbbT_C1ZHQYn1OlOc9vRVILvHs) split
-- Using [gazeta_ru val](https://drive.google.com/drive/folders/1BLiL3H0n56e8Q9jSuDgaH_3LLpmKxuVG) split
+- Using [last 1% of ria](https://drive.google.com/drive/folders/1ztAeyb1BiLMgXwOgOJS7WMR4PGiI1q92) dataset
+- Using [gazeta_ru test](https://drive.google.com/drive/folders/1CyowuRpecsLTcDbqEfmAvkCWOod58g_e) split
+- Using [gazeta_ru val](https://drive.google.com/drive/folders/1XZFOXHSXLKdhzm61ceVLw3aautrdskIu) split

config.json

@@ -1,5 +1,5 @@
 {
-  "_name_or_path": "/kaggle/input/bert2bertweights-192/ep_2",
+  "_name_or_path": "/kaggle/input/bert2bert-wexp/ep_1_v9",
   "architectures": [
     "EncoderDecoderModel"
   ],
@@ -19,6 +19,7 @@
     "diversity_penalty": 0.0,
     "do_sample": false,
     "early_stopping": false,
+    "encoder_no_repeat_ngram_size": 0,
     "eos_token_id": null,
     "finetuning_task": null,
     "gradient_checkpointing": false,
@@ -74,7 +75,7 @@
     "top_k": 50,
     "top_p": 1.0,
     "torchscript": false,
-    "transformers_version": "4.2.2",
+    "transformers_version": "4.3.2",
     "type_vocab_size": 2,
     "use_bfloat16": false,
     "use_cache": true,
@@ -98,6 +99,7 @@
     "diversity_penalty": 0.0,
     "do_sample": false,
     "early_stopping": false,
+    "encoder_no_repeat_ngram_size": 0,
     "eos_token_id": null,
     "finetuning_task": null,
     "gradient_checkpointing": false,
@@ -153,7 +155,7 @@
     "top_k": 50,
     "top_p": 1.0,
     "torchscript": false,
-    "transformers_version": "4.2.2",
+    "transformers_version": "4.3.2",
     "type_vocab_size": 2,
     "use_bfloat16": false,
     "use_cache": true,

pytorch_model.bin

@@ -1,3 +1,3 @@
 version https://git-lfs.github.com/spec/v1
-oid sha256:e1b2e1c4f765d2c3801c5dcc484ae464b3c0c72d50e9e0aa2589a1db06de504d
+oid sha256:71f249206ee2da240fc75f3b8d228ceee50861ff493ac0b6437e2509ad2754e0
 size 827914439

tokenizer_config.json

@@ -1 +1 @@
-{"do_lower_case": true, "unk_token": "[UNK]", "sep_token": "[SEP]", "pad_token": "[PAD]", "cls_token": "[CLS]", "mask_token": "[MASK]", "tokenize_chinese_chars": true, "strip_accents": null, "special_tokens_map_file": "/kaggle/input/deeppavlov-rubertbasecased/special_tokens_map.json", "name_or_path": "/kaggle/input/bert2bertweights-192/ep_2", "do_basic_tokenize": true, "never_split": null}
+{"do_lower_case": true, "unk_token": "[UNK]", "sep_token": "[SEP]", "pad_token": "[PAD]", "cls_token": "[CLS]", "mask_token": "[MASK]", "tokenize_chinese_chars": true, "strip_accents": null, "special_tokens_map_file": "/kaggle/input/deeppavlov-rubertbasecased/special_tokens_map.json", "name_or_path": "/kaggle/input/bert2bert-wexp/ep_1_v9", "do_basic_tokenize": true, "never_split": null}