fix: another approach

app.py

@@ -1,101 +1,34 @@
-import json
 import gradio as gr
-import torch
-from transformers import AutoModelForCausalLM, AutoTokenizer
 import spaces
+from transformers import AutoTokenizer, AutoModelForCausalLM
 import logging
 
 # Configure logging
 logging.basicConfig(level=logging.INFO)
 
-# Configuration
-MODEL_NAME = "speakleash/Bielik-11B-v2.3-Instruct"
-# DEVICE = "cuda" if torch.cuda.is_available() else "cpu"
-DEVICE = "cuda"
-TORCH_DTYPE = torch.bfloat16 if torch.cuda.is_available() else torch.float32
-MAX_TOKENS = 1000
-
 # Load model and tokenizer
 logging.info("Loading model and tokenizer...")
-tokenizer = AutoTokenizer.from_pretrained(MODEL_NAME)
-model = AutoModelForCausalLM.from_pretrained(MODEL_NAME, torch_dtype=TORCH_DTYPE).to(
-    DEVICE
-)
-
-logging.info("Model and tokenizer loaded successfully.")
-
-# Load prompts
-logging.info("Loading prompts from prompts.json...")
-with open("prompts.json") as f:
-    prompts = json.load(f)
+model_name = "speakleash/Bielik-11B-v2.3-Instruct"
+model = AutoModelForCausalLM.from_pretrained(model_name)
+tokenizer = AutoTokenizer.from_pretrained(model_name)
 
 
 @spaces.GPU
-def transform_text(prompt_name, user_input):
-    """Transform text using selected prompt and Bielik model"""
-    try:
-        logging.info(f"Transforming text with prompt: {prompt_name}")
-
-        if not prompt_name:
-            logging.error("No prompt selected.")
-            return "Error: No prompt selected."
-
-        # Get selected prompt
-        selected_prompt = next((p for p in prompts if p["name"] == prompt_name), None)
-        if selected_prompt is None:
-            logging.error(f"Prompt '{prompt_name}' not found.")
-            return f"Error: Prompt '{prompt_name}' not found."
-
-        # Create messages structure
-        messages = [
-            {"role": "system", "content": selected_prompt["system_message"]},
-            {"role": "user", "content": user_input},
-        ]
-
-        logging.info("Tokenizing input and generating output...")
-        input_ids = tokenizer.apply_chat_template(messages, return_tensors="pt").to(
-            DEVICE
-        )
-
-        generated_ids = model.generate(
-            input_ids, max_new_tokens=MAX_TOKENS, do_sample=True
-        )
-
-        result = tokenizer.batch_decode(generated_ids)[0]
-        logging.info("Text transformation successful.")
-        return result
-
-    except Exception as e:
-        logging.error("An error occurred during text transformation", exc_info=True)
-        return f"Error: {str(e)}"
-
-
-# Create Gradio interface
-with gr.Blocks(title="Bielik Goblin") as interface:
-    gr.Markdown("# Bielik Goblin")
-
-    prompt_select = gr.Dropdown(
-        choices=[p["name"] for p in prompts],
-        label="Wybierz prompt",
-        interactive=True,
-        value="Parafraza",  # Set "Parafraza" as the default value
-    )
-
-    user_input = gr.Textbox(
-        label="Twój tekst", placeholder="Wpisz tutaj swój tekst...", lines=5
-    )
+def process_text(input_text):
+    inputs = tokenizer(input_text, return_tensors="pt")
+    outputs = model(**inputs)
+    # Process outputs as needed
+    return outputs
 
-    transform_btn = gr.Button("Przekształć tekst", variant="primary")
 
-    with gr.Column():
-        output = gr.Textbox(label="Wynik", interactive=False)
+def generate(text):
+    hardcoded_prompt = "Stwórz zwięzłe podsumowanie tekstu, zachowując kluczowe punkty. Maksymalnie 3 zdania"
+    combined_text = hardcoded_prompt + text
+    return process_text(combined_text)
 
-    transform_btn.click(
-        fn=transform_text,
-        inputs=[prompt_select, user_input],
-        outputs=output,
-    )
 
-logging.info("Launching Gradio interface...")
-interface.queue().launch(debug=True)
-logging.info("Gradio interface launched successfully and ready to accept requests.")
+gr.Interface(
+    fn=generate,
+    inputs=gr.Text(),
+    outputs=gr.Text(),
+).launch()