Update app.py

app.py

@@ -3,7 +3,6 @@ import torch
 from transformers import AutoModelForCausalLM, AutoTokenizer
 import spaces
 
-# Your exact repository and GGUF path
 REPO_ID = "NiftyOracleAI/NiftyOracle-Coder-30B"
 GGUF_FILE = "gguf/codesymb_1b_chat_FP16.gguf"
 
@@ -18,36 +17,37 @@ model = AutoModelForCausalLM.from_pretrained(
 )
 print("✅ Model loaded into memory!")
 
-# ZeroGPU decorator: dynamically attaches the GPU when a prompt is sent
 @spaces.GPU
 def chat_with_codesymb(message, history):
-    # ChatML prompt formatting
-    prompt = (
-        f"<|system|>\nYou are CodeSymb, a world-class coding assistant by NiftyOracleAI.\n"
-        f"<|user|>\n{message}\n"
-        f"<|assistant|>\n"
-    )
+    # Standard format without unmapped special tag conflicts
+    prompt = f"User: {message}\nAssistant:"
     
     inputs = tokenizer(prompt, return_tensors="pt").to(model.device)
     
     with torch.no_grad():
         outputs = model.generate(
             **inputs,
-            max_new_tokens=512,
-            temperature=0.3,
+            max_new_tokens=256,
+            temperature=0.6,
+            top_p=0.9,
+            repetition_penalty=1.25,        # 🛑 Breaks the infinite repetition loop!
+            no_repeat_ngram_size=3,          # 🛑 Prevents repeating 3-word loops
             do_sample=True,
-            pad_token_id=tokenizer.eos_token_id
+            pad_token_id=tokenizer.eos_token_id if tokenizer.eos_token_id is not None else 2
         )
     
-    # Extract only the generated reply
     response = tokenizer.decode(outputs[0][inputs.input_ids.shape[1]:], skip_special_tokens=True)
-    return response
+    
+    # Strip any stray prompt echoes
+    if "User:" in response:
+        response = response.split("User:")[0]
+        
+    return response.strip()
 
-# Clean Gradio chat interface
 demo = gr.ChatInterface(
     fn=chat_with_codesymb,
     title="CodeSymb-1B Chat (ZeroGPU)",
-    description="Live inference for CodeSymb-1B FP16 GGUF running on Hugging Face ZeroGPU."
+    description="Live inference for CodeSymb-1B FP16 GGUF."
 )
 
 if __name__ == "__main__":