Update app.py

app.py

@@ -1,4 +1,4 @@
-from transformers import AutoModelForCausalLM, AutoTokenizer, AutoConfig
+from transformers import AutoModelForCausalLM, AutoTokenizer
 import torch
 import pandas as pd
 import streamlit as st
@@ -6,53 +6,50 @@ import streamlit as st
 # Tu token secreto de Hugging Face
 huggingface_token = st.secrets["HUGGINGFACEHUB_API_TOKEN"]
 
-# Cargar el modelo y tokenizer de un modelo más pequeño de LLaMA, usando el token secreto
-model_name = "meta-llama/Llama-3.1-8B"  # Cambiado a un modelo más pequeño
+# Cargar el modelo y tokenizer
+model_name = "meta-llama/Llama-3.1-8B"
 tokenizer = AutoTokenizer.from_pretrained(model_name, use_auth_token=huggingface_token)
-
-
-# Cargar el modelo en formato FP16 para ahorrar memoria
 model = AutoModelForCausalLM.from_pretrained(model_name, use_auth_token=huggingface_token, torch_dtype=torch.float16)
 
-
-
 # Usar Streamlit para cargar el archivo CSV
 uploaded_file = st.file_uploader("Por favor sube un archivo CSV:", type="csv")
 
 if uploaded_file is not None:
-    # Cargar el CSV y extraer la columna 'job_title'
     df = pd.read_csv(uploaded_file)
-    job_title = df['job_title'].tolist()
-
-    # Crear la lista de job titles en formato de texto para el prompt
-    job_title_text = "\n".join(f"- {title}" for title in job_title)
-
-    # Usar Streamlit para recoger la query del usuario
-    user_query = st.text_input("Introduce tu query:")
-
-    if user_query:
-        # Crear el prompt usando los job titles del CSV y la query del usuario
-        prompt = f"""
-        You are an advanced AI assistant trained to process job titles and user queries. I will provide you with a list of job titles, and a user query. Your task is to:
-        1. Calculate the cosine similarity score between the query and each job title.
-        2. Rank the job titles from the most similar to the least similar based on their semantic meaning.
-        3. Return the top 5 job titles with their cosine similarity scores.
-        Here is the list of job titles from the CSV:
-        {job_title_text}
-        The user's query is: "{user_query}"
-        Now, compute the similarity scores, rank the job titles, and return the top 5.
-        """
-
-        # Tokenizar y generar respuesta
-        inputs = tokenizer(prompt, return_tensors="pt")
-
-        # Desactivar el cálculo de gradientes para ahorrar memoria
-        with torch.no_grad():
-            outputs = model.generate(**inputs, max_new_tokens=200)
-
-        # Decodificar y mostrar resultados
-        response = tokenizer.decode(outputs[0], skip_special_tokens=True)
-        st.write(response)
+    if 'job_title' not in df.columns:
+        st.error("El archivo CSV debe contener una columna llamada 'job_title'.")
+    else:
+        job_title = df['job_title'].tolist()
+        job_title_text = "\n".join(f"- {title}" for title in job_title)
+
+        user_query = st.text_input("Introduce tu query:")
+
+        if user_query:
+            prompt = f"""
+            You are an advanced AI assistant trained to process job titles and user queries. I will provide you with a list of job titles, and a user query. Your task is to:
+            1. Calculate the cosine similarity score between the query and each job title.
+            2. Rank the job titles from the most similar to the least similar based on their semantic meaning.
+            3. Return the top 5 job titles with their cosine similarity scores.
+            Here is the list of job titles from the CSV:
+            {job_title_text}
+            The user's query is: "{user_query}"
+            Now, compute the similarity scores, rank the job titles, and return the top 5.
+            """
+
+            inputs = tokenizer(prompt, return_tensors="pt")
+
+            try:
+                with torch.no_grad():
+                    outputs = model.generate(**inputs, max_new_tokens=200)
+
+                response = tokenizer.decode(outputs[0], skip_special_tokens=True)
+                
+                if response:
+                    st.write(response)
+                else:
+                    st.error("No se generó ninguna respuesta.")
+            except Exception as e:
+                st.error(f"Error al generar la respuesta: {e}")
 
         # Limpiar memoria después de la inferencia
-        torch.cuda.empty_cache()  # Si estás usando una GPU
+        torch.cuda.empty_cache()  # Si estás usando una GPU