changes

app.py

@@ -1,22 +1,22 @@
 import gradio as gr
 from transformers import AutoModelForCausalLM, AutoTokenizer, pipeline
-import os
+import torch
 
 # Gunakan model yang sudah diunggah ke Hugging Face
 model_path = "FadQ/gemma-2b-diary-consultaton-chatbot"
 
-# Load the base model and tokenizer with trust_remote_code=True
-model = AutoModelForCausalLM.from_pretrained(base_model, trust_remote_code=True)
-tokenizer = AutoTokenizer.from_pretrained(base_model, trust_remote_code=True)
+# Load model dan tokenizer dengan `trust_remote_code=True`
+model = AutoModelForCausalLM.from_pretrained(model_path, trust_remote_code=True, torch_dtype=torch.float16, device_map="auto")
+tokenizer = AutoTokenizer.from_pretrained(model_path, trust_remote_code=True)
 
-# Ensure pipeline uses the correct tokenizer and check device
-pipe = pipeline("text-generation", model=model, tokenizer=tokenizer, device=0)  # Ubah ke device=-1 jika pakai CPU
+# Buat pipeline dengan tokenizer yang sesuai
+pipe = pipeline("text-generation", model=model, tokenizer=tokenizer, device=0)
 
 def predict(input_text):
     result = pipe(input_text, max_length=150, num_return_sequences=1)
     return result[0]["generated_text"]
 
-# Create the Gradio interface
+# Buat antarmuka Gradio
 demo = gr.Interface(fn=predict, inputs=gr.Textbox(label="Input Text"), outputs="text")
 
-demo.launch()
+demo.launch()