Update app.py

app.py

@@ -1,5 +1,5 @@
 import gradio as gr
-from transformers import AutoModelForCausalLM, AutoTokenizer
+from transformers import AutoModelForCausalLM, AutoTokenizer, pipeline
 import torch
 import numpy as np
 import cv2
@@ -8,26 +8,28 @@ import pytesseract as tess
 from sentence_transformers import SentenceTransformer, util
 import io
 
-save_directory="spaces/Garvitj/grader"
+# save_directory = "spaces/Garvitj/grader"
 
-# Load the tokenizer from the saved directory
-tokenizer = AutoTokenizer.from_pretrained(save_directory)
+# # Load the tokenizer from the saved directory
+# tokenizer = AutoTokenizer.from_pretrained(save_directory)
 
-# Load the model from the saved directory
-model = AutoModelForCausalLM.from_pretrained(
-    save_directory,
-    torch_dtype=torch.float16 if device == "cuda" else torch.float32,
-    device_map="auto" if device == "cuda" else None
-)
+# # Load the model from the saved directory
+# model = AutoModelForCausalLM.from_pretrained(
+#     save_directory,
+#     torch_dtype=torch.float16 if torch.cuda.is_available() else torch.float32,
+#     device_map="auto" if torch.cuda.is_available() else None
+# )
 
-# Move model to the appropriate device (CPU or CUDA)
-model.to(device)
+# # Move model to the appropriate device (CPU or CUDA)
+# device = "cuda" if torch.cuda.is_available() else "cpu"
+# model.to(device)
 
-print(f"Model and tokenizer loaded from {save_directory}")
+# print(f"Model and tokenizer loaded from {save_directory}")
 
 tess.pytesseract.tesseract_cmd = r"/app/tesseract.exe"
-# Load a smaller version of Sentence-BERT model
-model1 = SentenceTransformer('all-MiniLM-L6-v2')
+
+# Use a pipeline as a high-level helper
+pipe = pipeline("text-generation", model="eachadea/vicuna-7b-1.1")
 
 def get_embedding(text):
     return model1.encode(text, convert_to_tensor=True)
@@ -62,18 +64,13 @@ def evaluate_answer(image):
     model_answer = "The process of photosynthesis helps plants produce glucose using sunlight."
     similarity_score = calculate_similarity(student_answer, model_answer)
     grade = get_grade(similarity_score)
-    feedback = f"Student's answer: {student_answer}\nTeacher's answer: {model_answer}"
+    feedback = generate_response("Student's answer: {student_answer}\nTeacher's answer: {model_answer}")
     return grade, similarity_score * 100, feedback
 
 def generate_response(prompt):
-    inputs = tokenizer(prompt, return_tensors="pt").to(device)
-    
-    # Generate response from the model
-    with torch.no_grad():
-        outputs = model.generate(inputs.input_ids, max_length=150, temperature=0.7)
-    
-    response = tokenizer.decode(outputs[0], skip_special_tokens=True)
-    return response
+    # Generate response from the new model using the pipeline
+    response = pipe(prompt, max_length=150, temperature=0.7)
+    return response[0]['generated_text']
 
 def gradio_interface(image, prompt):
     grade, similarity_score, feedback = evaluate_answer(image)