Update app.py

app.py

@@ -1,5 +1,6 @@
 import gradio as gr
 from transformers import pipeline, AutoTokenizer, AutoModelForCausalLM
+import torch  # Import torch
 
 # --- Model Loading (Do this only once, outside the function) ---
 
@@ -7,15 +8,13 @@ from transformers import pipeline, AutoTokenizer, AutoModelForCausalLM
 use_pipeline = True  # Set to False to use the manual method
 
 if use_pipeline:
-    pipe = pipeline("text-generation", model="kakaocorp/kanana-nano-2.1b-base")
+    pipe = pipeline("text-generation", model="kakaocorp/kanana-nano-2.1b-base", device="cpu") #Explicitly on CPU
 else:
     # Option 2: Manual Tokenizer and Model (More Control)
     tokenizer = AutoTokenizer.from_pretrained("kakaocorp/kanana-nano-2.1b-base")
     model = AutoModelForCausalLM.from_pretrained("kakaocorp/kanana-nano-2.1b-base")
-    # Move model to GPU if available
-    if model.device.type != 'cuda' and torch.cuda.is_available():
-        model = model.to("cuda")
-        print("Model moved to CUDA")
+    #  No need to move to GPU. It will default to CPU.
+    print("Model loaded on CPU")
 
 # --- Generation Function ---
 
@@ -45,9 +44,7 @@ def generate_text(prompt, max_length=50, temperature=1.0, top_k=50, top_p=1.0, n
     else:  # Manual method
         try:
             inputs = tokenizer(prompt, return_tensors="pt")
-              # Move input tensors to the same device as the model
-            inputs = {k: v.to(model.device) for k, v in inputs.items()}
-            
+            # No need to move to GPU.  Inputs will default to CPU.
             outputs = model.generate(
                 **inputs,
                 max_length=max_length,
@@ -59,12 +56,12 @@ def generate_text(prompt, max_length=50, temperature=1.0, top_k=50, top_p=1.0, n
                 pad_token_id=tokenizer.eos_token_id,  # Ensure padding is correct
                 do_sample=True # Ensure sampling happens.
             )
-            
+
             generated_texts = []
             for i in range(outputs.shape[0]):
                  generated_text = tokenizer.decode(outputs[i], skip_special_tokens=True)
                  generated_texts.append(generated_text)
-           
+
             return "\n\n".join(generated_texts)
         except Exception as e:
             return f"Error during generation: {e}"
@@ -90,7 +87,7 @@ with gr.Blocks() as demo:
             generate_button = gr.Button("Generate")
 
         with gr.Column():
-            output_text = gr.Textbox(label="Generated Text", readonly=True)
+            output_text = gr.Textbox(label="Generated Text", interactive=False) # Use interactive=False
 
     generate_button.click(
         generate_text,
@@ -106,4 +103,4 @@ with gr.Blocks() as demo:
         outputs=output_text,
     )
 
-demo.launch(share=True)
+demo.launch() # Remove share=True for local testing, add it back for deployment