update

app.py

@@ -4,16 +4,17 @@ import torch
 
 model_name = "bharatgenai/AgriParam"
 
-print("Loading AgriParam model... This may take 5-15 minutes on free CPU")
+print("Loading AgriParam model... This may take some time on free CPU")
 
-tokenizer = AutoTokenizer.from_pretrained(model_name, trust_remote_code=True)
+tokenizer = AutoTokenizer.from_pretrained(model_name, trust_remote_code=False)
 
 model = AutoModelForCausalLM.from_pretrained(
     model_name,
-    torch_dtype=torch.float16,
-    device_map="cpu",
     trust_remote_code=True,
-    low_cpu_mem_usage=True
+    torch_dtype=torch.float16,          # Changed to float16 (more compatible on CPU)
+    device_map="cpu",
+    low_cpu_mem_usage=True,
+    attn_implementation="eager"         # This helps avoid rope_scaling issues
 )
 
 def chat(message, history):
@@ -26,7 +27,8 @@ def chat(message, history):
         temperature=0.7,
         do_sample=True,
         pad_token_id=tokenizer.eos_token_id,
-        eos_token_id=tokenizer.eos_token_id
+        eos_token_id=tokenizer.eos_token_id,
+        use_cache=False                    # Important for stability on CPU
     )
     
     response = tokenizer.decode(outputs[0], skip_special_tokens=True)
@@ -36,7 +38,7 @@ def chat(message, history):
 demo = gr.ChatInterface(
     fn=chat,
     title="🌾 AgriParam - Agriculture Assistant",
-    description="Test version running on free CPU • English, Hindi & Marathi",
+    description="Test version on free CPU • English, Hindi & Marathi",
     examples=[
         ["महाराष्ट्रात कापूस पिकावर लाल किडीचा उपाय काय आहे?"],
         ["Best practices for organic wheat farming in Maharashtra?"],