Update app.py

app.py

@@ -3,9 +3,7 @@ import gradio as gr
 import torch
 from transformers import AutoModelForCausalLM, AutoTokenizer
 
-model_id = "lakshraina2/leetcodeAI" 
-
-# Securely grab the token you just saved in the Space settings
+model_id = "lakshraina2/leetcode-coder-1.5B" 
 hf_token = os.environ.get("HF_TOKEN")
 
 print("Loading model securely...")
@@ -18,28 +16,30 @@ model = AutoModelForCausalLM.from_pretrained(
 
 def solve(problem_text):
     if not problem_text or len(problem_text) < 10:
-        return "// Error: Problem text too short or not scraped correctly."
+        return "// Error: Problem text too short."
 
-    prompt = f"Below is a LeetCode problem. Write a complete Python solution.\n\n### Problem:\n{problem_text}\n\n### Solution:\n"
+    # Let's try a simpler, universal prompt format
+    prompt = f"Problem:\n{problem_text}\n\nPython code solution:\n"
     
     inputs = tokenizer(prompt, return_tensors="pt")
     
+    print("Starting generation...") # This will show up in HF Logs
+    
     with torch.no_grad():
         outputs = model.generate(
             input_ids=inputs["input_ids"],
             attention_mask=inputs["attention_mask"],
-            max_new_tokens=1024, 
-            min_new_tokens=50,   
-            temperature=0.1,     
-            do_sample=True,
+            max_new_tokens=512, 
+            do_sample=False, # Force deterministic greedy decoding
             pad_token_id=tokenizer.eos_token_id
         )
     
     full_text = tokenizer.decode(outputs[0], skip_special_tokens=True)
     
-    if "### Solution:" in full_text:
-        return full_text.split("### Solution:")[-1].strip()
-    return full_text.strip()
+    print("RAW MODEL OUTPUT:\n", full_text) # Check HF logs to see exactly what it did
+    
+    # TEMPORARY: Return the whole thing so you can see it in the GUI!
+    return full_text
 
 demo = gr.Interface(fn=solve, inputs="text", outputs="text", api_name="predict")
 demo.launch()