Update app.py

app.py

@@ -1,41 +1,42 @@
+import os
 import gradio as gr
 import torch
 from transformers import AutoModelForCausalLM, AutoTokenizer
 
-model_id = "lakshraina2/leetcode-coder-1.5B" 
+model_id = "lakshraina2/leetcodeAI" 
 
-print("Loading model...")
-tokenizer = AutoTokenizer.from_pretrained(model_id, token=False)
+# Securely grab the token you just saved in the Space settings
+hf_token = os.environ.get("HF_TOKEN")
+
+print("Loading model securely...")
+tokenizer = AutoTokenizer.from_pretrained(model_id, token=hf_token)
 model = AutoModelForCausalLM.from_pretrained(
     model_id, 
     dtype=torch.float32, 
-    token=False
+    token=hf_token
 )
 
 def solve(problem_text):
     if not problem_text or len(problem_text) < 10:
         return "// Error: Problem text too short or not scraped correctly."
 
-    # Standard Alpaca/Llama prompt format
     prompt = f"Below is a LeetCode problem. Write a complete Python solution.\n\n### Problem:\n{problem_text}\n\n### Solution:\n"
     
     inputs = tokenizer(prompt, return_tensors="pt")
     
-    # Generate with specific constraints to prevent empty output
     with torch.no_grad():
         outputs = model.generate(
             input_ids=inputs["input_ids"],
             attention_mask=inputs["attention_mask"],
-            max_new_tokens=1024, # Increased for complex problems
-            min_new_tokens=50,   # Force the model to talk
-            temperature=0.1,     # Lower temperature = more focused/less random
+            max_new_tokens=1024, 
+            min_new_tokens=50,   
+            temperature=0.1,     
             do_sample=True,
             pad_token_id=tokenizer.eos_token_id
         )
     
     full_text = tokenizer.decode(outputs[0], skip_special_tokens=True)
     
-    # Extract only the part after our '### Solution:' marker
     if "### Solution:" in full_text:
         return full_text.split("### Solution:")[-1].strip()
     return full_text.strip()