Update app.py

app.py

@@ -5,36 +5,36 @@ from transformers import AutoTokenizer, AutoModelForCausalLM
 from diffusers import StableDiffusionPipeline
 from rouge_score import rouge_scorer
 from PIL import Image
+import clip
 import tempfile
 import os
+import math
 import time
-import clip
-import torchvision.transforms as transforms
 
-# Use CUDA if available
+# Device setup
 device = "cuda" if torch.cuda.is_available() else "cpu"
 
-# Load translation model (Tamil to English)
+# Translation model
 translator_model = MBartForConditionalGeneration.from_pretrained("facebook/mbart-large-50-many-to-many-mmt").to(device)
 translator_tokenizer = MBart50TokenizerFast.from_pretrained("facebook/mbart-large-50-many-to-many-mmt")
 translator_tokenizer.src_lang = "ta_IN"
 
-# Load GPT-2 for creative text generation
+# GPT-2 for creative text
 gen_model = AutoModelForCausalLM.from_pretrained("gpt2").to(device)
 gen_tokenizer = AutoTokenizer.from_pretrained("gpt2")
 
-# Load a lightweight image generation model
+# Stable Diffusion v1.4
 pipe = StableDiffusionPipeline.from_pretrained(
-    "OFA-Sys/small-stable-diffusion-v0",
+    "stabilityai/stable-diffusion-1-4",
     torch_dtype=torch.float32,
-    use_auth_token=os.getenv("HF_TOKEN")  # Set in Hugging Face Space secrets
+    use_auth_token=os.getenv("HF_TOKEN")  # set this on Hugging Face Spaces
 ).to(device)
-pipe.safety_checker = None  # Optional: disable for speed
+pipe.safety_checker = None  # Optional
 
-# Load CLIP model for image-text similarity
+# Load CLIP for image-text similarity
 clip_model, clip_preprocess = clip.load("ViT-B/32", device=device)
 
-# Translation Function
+# Translation function
 def translate_tamil_to_english(text, reference=None):
     start = time.time()
     inputs = translator_tokenizer(text, return_tensors="pt").to(device)
@@ -53,56 +53,58 @@ def translate_tamil_to_english(text, reference=None):
 
     return translated, duration, rouge_l
 
-# Creative Text Generator with Perplexity
+# Text generation with repetition & perplexity
 def generate_creative_text(prompt, max_length=100):
     start = time.time()
     input_ids = gen_tokenizer.encode(prompt, return_tensors="pt").to(device)
-    output = gen_model.generate(input_ids, max_length=max_length, do_sample=True, top_k=50, temperature=0.9)
+    output = gen_model.generate(
+        input_ids, max_length=max_length,
+        do_sample=True, top_k=50, temperature=0.9
+    )
     text = gen_tokenizer.decode(output[0], skip_special_tokens=True)
     duration = round(time.time() - start, 2)
 
     tokens = text.split()
     repetition_rate = sum(t1 == t2 for t1, t2 in zip(tokens, tokens[1:])) / len(tokens)
 
-    # Perplexity calculation
+    # Perplexity
     with torch.no_grad():
-        input_ids = gen_tokenizer.encode(text, return_tensors="pt").to(device)
         outputs = gen_model(input_ids, labels=input_ids)
         loss = outputs.loss
-        perplexity = torch.exp(loss).item()
+        perplexity = math.exp(loss.item())
 
-    return text, duration, len(tokens), round(repetition_rate, 4), round(perplexity, 4)
+    return text, duration, len(tokens), round(repetition_rate, 4), round(perplexity, 3)
 
-# AI Image Generator with CLIP Similarity
+# Image generation + CLIP similarity
 def generate_image(prompt):
     try:
         start = time.time()
         result = pipe(prompt)
         image = result.images[0].resize((256, 256))
+        duration = round(time.time() - start, 2)
 
+        # Save image
         tmp_file = tempfile.NamedTemporaryFile(delete=False, suffix=".png")
         image.save(tmp_file.name)
 
         # CLIP similarity
         image_input = clip_preprocess(image).unsqueeze(0).to(device)
-        text_input = clip.tokenize([prompt]).to(device)
-
+        text_input = clip.tokenize(prompt).to(device)
         with torch.no_grad():
             image_features = clip_model.encode_image(image_input)
             text_features = clip_model.encode_text(text_input)
             similarity = torch.cosine_similarity(image_features, text_features).item()
 
-        return tmp_file.name, round(time.time() - start, 2), round(similarity, 4)
-
+        return tmp_file.name, duration, round(similarity, 4)
     except Exception as e:
-        return None, f"Image generation failed: {str(e)}", None
+        return None, 0, f"Image generation failed: {str(e)}"
 
 # Streamlit UI
 st.set_page_config(page_title="Tamil → English + AI Art", layout="centered")
 st.title("🧠 Tamil → English + 🎨 Creative Text + AI Image")
 
 tamil_input = st.text_area("✍️ Enter Tamil text here", height=150)
-reference_input = st.text_input("📘 Optional: Reference English translation for ROUGE-L")
+reference_input = st.text_input("📘 Optional: Reference English translation for ROUGE")
 
 if st.button("🚀 Generate Output"):
     if not tamil_input.strip():
@@ -115,27 +117,23 @@ if st.button("🚀 Generate Output"):
         st.markdown(f"**📝 English Translation:** `{english_text}`")
         if rouge_l is not None:
             st.markdown(f"📊 **ROUGE-L Score:** `{rouge_l}`")
-        else:
-            st.info("ℹ️ ROUGE-L not calculated. Reference not provided.")
 
-        with st.spinner("🎨 Generating image..."):
-            image_path, img_time, clip_score = generate_image(english_text)
+        with st.spinner("🖼️ Generating image..."):
+            image_path, img_time, similarity = generate_image(english_text)
 
-        if image_path:
+        if isinstance(similarity, float):
             st.success(f"🖼️ Image generated in {img_time} seconds")
             st.image(Image.open(image_path), caption="AI-Generated Image", use_column_width=True)
-            st.markdown(f"🔍 **CLIP Text-Image Similarity:** `{clip_score}`")
+            st.markdown(f"🎯 **CLIP Text-Image Similarity:** `{similarity}`")
         else:
-            st.error(image_path)
+            st.error(similarity)
 
         with st.spinner("💡 Generating creative text..."):
             creative, c_time, tokens, rep_rate, perplexity = generate_creative_text(english_text)
 
         st.success(f"✨ Creative text generated in {c_time} seconds")
-        st.markdown("**🧠 Creative Output:**")
-        st.text(creative)
-        st.markdown(f"📌 Tokens: `{tokens}`")
-        st.markdown(f"🔁 Repetition Rate: `{rep_rate}`")
+        st.markdown(f"**🧠 Creative Output:** `{creative}`")
+        st.markdown(f"📌 Tokens: `{tokens}`, 🔁 Repetition Rate: `{rep_rate}`")
         st.markdown(f"📉 Perplexity: `{perplexity}`")
 
 st.markdown("---")