Update app.py

app.py

@@ -11,10 +11,8 @@ import spacy
 import io
 from spacy.cli import download
 
-# Download the model if it is not already present
+# Download and load the spaCy model
 download("en_core_web_sm")
-
-# Load the model
 nlp = spacy.load("en_core_web_sm")
 
 # Download NLTK WordNet data
@@ -22,40 +20,24 @@ import nltk
 nltk.download('wordnet')
 nltk.download('omw-1.4')
 
-# Load spaCy model
-nlp = spacy.load("en_core_web_sm")
-
 # Load the pre-trained model for image captioning
 model_name = "NourFakih/Vit-GPT2-COCO2017Flickr-85k-11"
 model = VisionEncoderDecoderModel.from_pretrained(model_name)
 feature_extractor = ViTImageProcessor.from_pretrained(model_name)
 tokenizer = AutoTokenizer.from_pretrained(model_name)
-# GPT2 only has bos/eos tokens but not decoder_start/pad tokens
 tokenizer.pad_token = tokenizer.eos_token
 
-# update the model config
+# Update the model config
 model.config.eos_token_id = tokenizer.eos_token_id
 model.config.decoder_start_token_id = tokenizer.bos_token_id
 model.config.pad_token_id = tokenizer.pad_token_id
 
-def preprocess_query(query):
-    doc = nlp(query)
-    tokens = set()
-    for token in doc:
-        tokens.add(token.text)
-        tokens.add(token.lemma_)
-        tokens.update(get_synonyms(token.text))
-    st.write(f"Query tokens: {tokens}")  # Debugging line
-    return tokens
-
 def generate_caption(image):
     pixel_values = feature_extractor(images=image, return_tensors="pt").pixel_values
     output_ids = model.generate(pixel_values)
     caption = tokenizer.decode(output_ids[0], skip_special_tokens=True)
-    st.write(f"Generated caption: {caption}")  # Debugging line
     return caption
 
-
 def get_synonyms(word):
     synonyms = set()
     for syn in wordnet.synsets(word):
@@ -63,6 +45,14 @@ def get_synonyms(word):
             synonyms.add(lemma.name())
     return synonyms
 
+def preprocess_query(query):
+    doc = nlp(query)
+    tokens = set()
+    for token in doc:
+        tokens.add(token.text)
+        tokens.add(token.lemma_)
+        tokens.update(get_synonyms(token.text))
+    return tokens
 
 def search_captions(query, captions):
     query_tokens = preprocess_query(query)