Update checkpoint for transformers>=4.29

Following the merge of [a PR](https://github.com/huggingface/transformers/pull/24310) in `transformers` it appeared that this model was not properly converted. This PR will fix the inference and was tested using the following script:
```python
>>> from transformers import MarianModel, MarianMTModel
>>> tokenizer = AutoTokenizer.from_pretrained(Helsinki-NLP/opus-mt-tc-big-it-en)
>>> inputs = tokenizer(EHI!Impariamo insieme, return_tensors=pt, padding=True)
>>> model = MarianMTModel.from_pretrained(Helsinki-NLP/opus-mt-tc-big-it-en)
>>> print(tokenizer.batch_decode(model.generate(**inputs)))
["<pad> EHI!Let's Learn Together</s>"]
```

config.json

@@ -7,7 +7,7 @@
   "attention_dropout": 0.0,
   "bad_words_ids": [
     [
-      35697
+      54421
     ]
   ],
   "bos_token_id": 0,
@@ -17,15 +17,15 @@
   "decoder_ffn_dim": 4096,
   "decoder_layerdrop": 0.0,
   "decoder_layers": 6,
-  "decoder_start_token_id": 35697,
-  "decoder_vocab_size": 35698,
+  "decoder_start_token_id": 54421,
+  "decoder_vocab_size": 54422,
   "dropout": 0.1,
   "encoder_attention_heads": 16,
   "encoder_ffn_dim": 4096,
   "encoder_layerdrop": 0.0,
   "encoder_layers": 6,
-  "eos_token_id": 26162,
-  "forced_eos_token_id": 26162,
+  "eos_token_id": 43017,
+  "forced_eos_token_id": 43017,
   "init_std": 0.02,
   "is_encoder_decoder": true,
   "max_length": 512,
@@ -34,12 +34,12 @@
   "normalize_embedding": false,
   "num_beams": 4,
   "num_hidden_layers": 6,
-  "pad_token_id": 35697,
+  "pad_token_id": 54421,
   "scale_embedding": true,
   "share_encoder_decoder_embeddings": true,
   "static_position_embeddings": true,
   "torch_dtype": "float16",
   "transformers_version": "4.34.0.dev0",
   "use_cache": true,
-  "vocab_size": 35698
+  "vocab_size": 54422
 }

generation_config.json

@@ -1,16 +1,16 @@
 {
   "bad_words_ids": [
     [
-      35697
+      54421
     ]
   ],
   "bos_token_id": 0,
-  "decoder_start_token_id": 35697,
-  "eos_token_id": 26162,
-  "forced_eos_token_id": 26162,
+  "decoder_start_token_id": 54421,
+  "eos_token_id": 43017,
+  "forced_eos_token_id": 43017,
   "max_length": 512,
   "num_beams": 4,
-  "pad_token_id": 35697,
+  "pad_token_id": 54421,
   "renormalize_logits": true,
   "transformers_version": "4.34.0.dev0"
 }

model.safetensors

@@ -1,3 +1,3 @@
 version https://git-lfs.github.com/spec/v1
-oid sha256:69ea2cf1906dd5768225f7ef1a778f8eaa87dbc8e2c1e4190ec2ca774975dd69
-size 425925172
+oid sha256:e1a87e60d3e108fb02f3d084707416620059094f3d06f0138444373952ede8b6
+size 464309380

pytorch_model.bin

@@ -1,3 +1,3 @@
 version https://git-lfs.github.com/spec/v1
-oid sha256:ebbeee990e971cb41fd5262b168a044724e2c47dde06de7d10deb49287ad4cfc
-size 425982149
+oid sha256:9150a36c7e89654602c14b7a6714e95a716e56be18d3e37198b425f9b1c51b18
+size 464366341