Update app.py

app.py

@@ -2,7 +2,6 @@ from transformers import AutoModelForCausalLM, AutoTokenizer, TextIteratorStream
 import streamlit as st
 from huggingface_hub import login
 import pandas as pd
-from threading import Thread
 
 # Token Secret de Hugging Face
 huggingface_token = st.secrets["HUGGINGFACEHUB_API_TOKEN"]
@@ -33,22 +32,14 @@ def generate_response(input_text, temperature=0.7, max_new_tokens=20):
         top_p=0.9,
         temperature=temperature,
         num_return_sequences=3,
-        eos_token_id=[tokenizer.eos_token_id]
+        eos_token_id=tokenizer.eos_token_id  # Cambiado a un entero
     )
 
     try:
-        t = Thread(target=model.generate, kwargs=generate_kwargs)
-        t.start()
-        t.join()  # Asegura que la generación haya terminado
+        outputs = model.generate(**generate_kwargs)
 
-        outputs = []
-        for text in streamer:
-            outputs.append(text)
-        if not outputs:
-            raise ValueError("No se generó ninguna respuesta.")
-        
-        response = "".join(outputs).strip().split("\n")[0]
-        return response
+        response = tokenizer.decode(outputs[0], skip_special_tokens=True).strip()
+        return response.split("\n")[0]
     except Exception as e:
         st.error(f"Error durante la generación: {e}")
         return "Error en la generación de texto."
@@ -61,28 +52,20 @@ def main():
     if uploaded_file is not None:
         df = pd.read_csv(uploaded_file)
         query = 'aspiring human resources specialist'
-        value = 0.00
+        
         if 'job_title' in df.columns:
-            job_titles = df['job_title']
+            job_titles = df['job_title'].tolist()
 
             # Definir el prompt con in-context learning
             initial_prompt = (
-                "Step 1: Extract the first record from the dataframe df.\n"
-                f" {df.iloc[0]['job_title']}\n"
-                #f"List: {job_titles}\n"
-                #"First job title: \n"
-                #"\n"
-                "Step 2: Calculate the cosine similarity score between the job_title of the extracted record {df.iloc[0]['job_title']} and the given {query} and assign it to {value}.\n"
-                f"Query: '{query}'\n"
-                "Cosine similarity score: \n"
-                "Step 3: Print the value of the calculated cosine similarity"
-                f"Result: {value}"
+                f"Extract the first record from the dataframe df.\n"
+                f"First job title: '{df.iloc[0]['job_title']}'\n"
+                f"Calculate the cosine similarity between this job title and the query: '{query}'.\n"
+                "Print the cosine similarity score."
             )
 
-            
             st.write("Prompt inicial con In-context Learning:\n")
             st.write(initial_prompt)
-            st.write(query)
 
             if st.button("Generar respuesta"):
                 with st.spinner("Generando respuesta..."):