feat: replace client API with direct model inference

Switched from using the external client to directly loading the model and tokenizer with transformers library. This change improves efficiency by reducing dependency on the client API and allows for more direct control over model inference.

app.py

@@ -1,16 +1,15 @@
 import gradio as gr
-from gradio_client import Client
+from transformers import AutoTokenizer, AutoModelForCausalLM
 
-# Create a client to interact with your Space
-client = Client("umutbozdag/humanizer_model")
+# Load the model and tokenizer
+model = AutoModelForCausalLM.from_pretrained("umutbozdag/autotrain-g39vl-h3lir")
+tokenizer = AutoTokenizer.from_pretrained("umutbozdag/autotrain-g39vl-h3lir")
 
 def generate_text(input_text):
-    # Use the client to predict using your Space
-    result = client.predict(
-        input_text,  # str in 'Input Text' Textbox component
-        api_name="/predict"
-    )
-    return result
+    # Tokenize input and generate text
+    inputs = tokenizer(input_text, return_tensors="pt")
+    outputs = model.generate(**inputs, max_length=100)
+    return tokenizer.decode(outputs[0], skip_special_tokens=True)
 
 iface = gr.Interface(
     fn=generate_text,