Update app.py

app.py

@@ -8,8 +8,9 @@ import spaces
 import markdown
 import requests
 import torch
+import io
 from PIL import Image
-from transformers import MllamaForConditionalGeneration, AutoProcessor
+from transformers import MllamaForConditionalGeneration, AutoProcessor,AutoModelForCausalLM, AutoTokenizer
 
 
 device = 'cuda' if torch.cuda.is_available() else 'cpu'
@@ -27,6 +28,11 @@ model = MllamaForConditionalGeneration.from_pretrained(
 processor = AutoProcessor.from_pretrained(model_id)
 
 
+model_name = "Qwen/Qwen2.5-Coder-7B-Instruct"
+model = AutoModelForCausalLM.from_pretrained(model_name, torch_dtype="auto", device_map="auto")
+tokenizer = AutoTokenizer.from_pretrained(model_name)
+
+
 SYSTEM_INSTRUCTION="You are DailySnap, your job is to anlyse the given image and provide daily journal about the image and use some random time"
 
 def extract_assistant_reply(input_string):
@@ -84,6 +90,28 @@ def generate__image_desc(image):
     html_output = markdown.markdown(markdown_text)
     return html_output
 
+@spaces.GPU
+def generate_journal_infographics(code_input):
+    prompt = f"Generate daily journal inforgraphics using html for the following:\n\n{code_input}"
+    
+    messages = [
+        {"role": "system", "content": "You are DailySnap, a highly efficient and intelligent assistant designed to generate infographics using htmnl bootstrap icon and generate highly appealing daily journal as per the user detail"},
+        {"role": "user", "content": prompt}
+    ]
+    
+    # Prepare inputs for the model
+    text = tokenizer.apply_chat_template(messages, tokenize=False, add_generation_prompt=True)
+    model_inputs = tokenizer([text], return_tensors="pt").to(model.device)
+    
+    # Generate the documentation
+    generated_ids = model.generate(**model_inputs, max_new_tokens=4000)
+    generated_ids = [output_ids[len(input_ids):] for input_ids, output_ids in zip(model_inputs.input_ids, generated_ids)]
+    documentation = tokenizer.batch_decode(generated_ids, skip_special_tokens=True)[0]
+    print(documentation)
+    return documentation
+
+
+
 # Define activity categories based on detected objects
 activity_categories = {
     "Working": ["laptop", "computer", "keyboard", "office chair"],