Update gradio_app.py

gradio_app.py

@@ -1,20 +1,9 @@
 # Importing libraries
-from transformers import M2M100Tokenizer, M2M100ForConditionalGeneration
 from llama_cpp import Llama
 import gradio as gr
 import psutil
 
 # Initing things                
-print("! DOWNLOADING TOKENIZER AND SETTING ALL UP !")
-translator_tokenizer = M2M100Tokenizer.from_pretrained(            # tokenizer for translator
-    "facebook/m2m100_418M", cache_dir="translator/"
-)
-print("! DOWNLOADING MODEL AND SETTING ALL UP !")
-translator_model = M2M100ForConditionalGeneration.from_pretrained( # translator model
-    "facebook/m2m100_418M", cache_dir="translator/"
-)
-print("! SETTING MODEL IN EVALUATION MODE !")
-translator_model.eval()
 print("! INITING LLAMA MODEL !")
 llm = Llama(model_path="./model.bin")                              # LLaMa model
 llama_model_name = "TheBloke/dolphin-2.2.1-AshhLimaRP-Mistral-7B-GGUF" 
@@ -35,21 +24,6 @@ Or you can once follow steps in Dockerfile and try it on your machine, not in Do
 <br>''' + f"Memory used: {psutil.virtual_memory()[2]}<br>" + '''
 Powered by <a href="https://github.com/abetlen/llama-cpp-python">llama-cpp-python</a> and <a href="https://www.gradio.app/">Gradio</a>.<br><br>'''
 
-'''
-    # Defining languages for translator (i just chose popular on my opinion languages!!!)
-    ru - Russian
-    uk - Ukranian
-    zh - Chinese
-    de - German
-    fr - French
-    hi - Hindi
-    it - Italian
-    ja - Japanese
-    es - Spanish
-    ar - Arabic
-'''
-languages = ["ru", "uk", "zh", "de", "fr", "hi", "it", "ja", "es", "ar"]
-
 # Loading prompt
 with open('system.prompt', 'r', encoding='utf-8') as f:
     prompt = f.read()
@@ -77,21 +51,11 @@ def generate_answer(request: str, max_tokens: int = 256, language: str = "en", c
                 break
             counter += 1
         logs += f"Final attempt: {counter}\n"
-        if len(text.strip()) > 1 and text.strip() not in ['', None, ' ']:
+        if len(text.strip()) <= 1 or text.strip() in ['', None, ' ']:
+            logs += f"Generated and aborted: {text}"
             text = "Sorry, but something went wrong while generating answer. Try again or fix code. If you are maintainer of this space, look into logs."
         
-        if language in languages and language != "en":
-            logs += f"\nTranslating from en to {language}"
-            encoded_input = translator_tokenizer(text, return_tensors="pt")
-            generated_tokens = translator_model.generate(
-                **encoded_input, forced_bos_token_id=translator_tokenizer.get_lang_id(language)
-            )
-            translated_text = translator_tokenizer.batch_decode(
-                generated_tokens, skip_special_tokens=True
-            )[0]
-            logs += f"\nTranslated: {translated_text}\nOriginal: {text}"
-            return translated_text, logs
-        logs += f"\nOriginal: {text}"
+        logs += f"\nFinal: '''{text}'''"
         return text, logs
     except Exception as e:
         print(e)