fixes for nltk and transformers updates (#6)

- fixes for nltk and transformers updates (d5f8a9a2c51772f299cc8c6fb8a90ca77732c184)
- Update app.py (7798d43d37bfc02159a6ad1b6a805d8cd8c2aea9)

app.py

@@ -6,7 +6,7 @@ import platform
 import torch
 import nltk
 
-nltk.download("punkt")
+nltk.download("punkt_tab")
 
 REMOVED_TARGET_LANGUAGES = {"Ligurian", "Lombard", "Sicilian"}
 
@@ -55,7 +55,7 @@ def translate(text: str, src_lang: str, tgt_lang: str):
             )
             translated_chunk = model.generate(
                 input_ids=torch.tensor([input_tokens]).to(device),
-                forced_bos_token_id=tokenizer.lang_code_to_id[code_mapping[tgt_lang]],
+                forced_bos_token_id=tokenizer.convert_tokens_to_ids(code_mapping[tgt_lang]),
                 max_length=len(input_tokens) + 50,
                 num_return_sequences=1,
             )
@@ -93,4 +93,4 @@ with gr.Blocks() as demo:
         inputs=[input_text, src_lang, target_lang],
         outputs=output,
     )
-demo.launch()
+demo.launch()