Update app.py

app.py

@@ -1,10 +1,11 @@
 import streamlit as st
-from transformers import pipeline, AutoModelForSeq2SeqLM, AutoTokenizer
+from transformers import pipeline
 from rembg import remove
 from PIL import Image
 from io import BytesIO
 import base64
 from transformers import GPT2LMHeadModel, GPT2Tokenizer
+from transformers import MarianMTModel, MarianTokenizer
 from transformers import set_seed
 import random
 
@@ -31,6 +32,11 @@ image_classifier = pipeline("image-classification", model="mjsp/sweet")
 recipe_model = GPT2LMHeadModel.from_pretrained("gpt2")
 recipe_tokenizer = GPT2Tokenizer.from_pretrained("gpt2")
 
+# Load the translation model and tokenizer for translation from English to Hindi
+translation_model_name = "Helsinki-NLP/opus-mt-en-hi"
+translator_model = MarianMTModel.from_pretrained(translation_model_name)
+translator_tokenizer = MarianTokenizer.from_pretrained(translation_model_name)
+
 MAX_FILE_SIZE = 5 * 1024 * 1024  # 5MB
 
 def convert_image(img):
@@ -50,6 +56,16 @@ def fix_image(upload):
     st.sidebar.markdown("\n")
     st.sidebar.download_button("Download fixed image", convert_image(fixed), "fixed.png", "image/png")
 
+def translate_to_hindi(text):
+    try:
+        translation_result = translator_tokenizer(text, return_tensors="pt", padding=True)
+        generated_tokens = translator_model.generate(**translation_result)
+        translated_text = translator_tokenizer.batch_decode(generated_tokens, skip_special_tokens=True)[0]
+        return translated_text
+    except Exception as e:
+        st.error(f"Error during translation: {str(e)}")
+        return None
+        
 
 def generate_summary(recipe):
     # Tokenize the text into words and sentences
@@ -1301,24 +1317,6 @@ st.write("## Recipe Generation")
 selected_item = st.selectbox("Select a food item", [entry["title"] for entry in dataset])
 
 
-# Load the translation model and tokenizer for Hindi
-model_name = "t5-base"
-translator_model = AutoModelForSeq2SeqLM.from_pretrained(model_name)
-translator_tokenizer = AutoTokenizer.from_pretrained(model_name)
-
-# Load the translation pipeline
-translator = pipeline(task="translation", model=translator_model, tokenizer=translator_tokenizer)
-
-# Function to translate text to Hindi
-def translate_to_hindi(text):
-    try:
-        translation_result = translator(text, src="en", tgt="hi")
-        return translation_result[0]['translation_text']
-    except Exception as e:
-        st.error(f"Error during translation: {str(e)}")
-        return None
-
-selected_entry = None  # Initialize selected_entry outside the block
 
 if st.button("Generate Recipe"):
     matching_entries = [entry for entry in dataset if entry["title"] == selected_item]
@@ -1351,7 +1349,6 @@ if st.button("Generate Recipe"):
 
 translate_button = st.sidebar.button("Translate to Hindi")
 
-
 if translate_button:
     if selected_entry and "recipe" in selected_entry:
         recipe_to_translate = selected_entry["recipe"]
@@ -1360,8 +1357,7 @@ if translate_button:
             st.write(f"Translated Recipe (Hindi): {translated_recipe}")
     else:
         st.error("Please generate a recipe first.")
-
-
+        
    
 # Add some descriptions and instructions
 st.sidebar.markdown("### Instructions")