Spaces:
Sleeping
Sleeping
Update app.py
Browse files
app.py
CHANGED
|
@@ -33,64 +33,57 @@ try:
|
|
| 33 |
else:
|
| 34 |
print("dato default")
|
| 35 |
except:
|
| 36 |
-
|
| 37 |
-
|
| 38 |
-
|
| 39 |
-
|
| 40 |
-
|
| 41 |
-
text_splitter = RecursiveCharacterTextSplitter(chunk_size=1000, chunk_overlap=150)
|
| 42 |
-
#Transformado a tipo de dato especifico para esto
|
| 43 |
-
docs = text_splitter.split_documents(data) # 'data' holds the text you want to split, split the text into documents using the text splitter.
|
| 44 |
-
|
| 45 |
-
#Modelo QA sentence similarity
|
| 46 |
-
modelPath = 'sentence-transformers/paraphrase-multilingual-MiniLM-L12-v2' #español
|
| 47 |
-
model_kwargs = {'device':'cpu'} # o cuda
|
| 48 |
-
encode_kwargs = {'normalize_embeddings': False}
|
| 49 |
-
|
| 50 |
-
#Embeddings que transforman a vectores densos multidimensionales las preguntas del SII
|
| 51 |
-
embeddings = HuggingFaceEmbeddings(
|
| 52 |
-
model_name=modelPath, # Ruta a modelo Pre entrenado
|
| 53 |
-
model_kwargs=model_kwargs, # Opciones de configuracion del modelo
|
| 54 |
-
encode_kwargs=encode_kwargs # Opciones de Encoding
|
| 55 |
-
)
|
| 56 |
-
|
| 57 |
-
#DB y retriever
|
| 58 |
-
db = FAISS.from_documents(docs, embeddings) # Create a retriever object from the 'db' with a search configuration where it retrieves up to 4 relevant splits/documents.
|
| 59 |
-
retriever = db.as_retriever(search_kwargs={"k": 3})
|
| 60 |
-
|
| 61 |
-
template = """Responde la pregunta basado unicamente en el siguiente contexto
|
| 62 |
-
|
| 63 |
-
{contexto}
|
| 64 |
-
|
| 65 |
-
Pregunta: {pregunta}
|
| 66 |
-
|
| 67 |
-
"""
|
| 68 |
-
|
| 69 |
-
#LLM
|
| 70 |
-
prompt = ChatPromptTemplate.from_template(template)
|
| 71 |
-
model = ChatOpenAI(model="gpt-3.5-turbo-1106", temperature=0)
|
| 72 |
-
|
| 73 |
-
def format_docs(docs):
|
| 74 |
-
return "\n\n".join([d.page_content for d in docs])
|
| 75 |
-
|
| 76 |
-
chain = (
|
| 77 |
-
{"contexto": retriever | format_docs, "pregunta": RunnablePassthrough()}
|
| 78 |
-
| prompt
|
| 79 |
-
| model
|
| 80 |
-
| StrOutputParser()
|
| 81 |
-
)
|
| 82 |
-
print(chain)
|
| 83 |
-
add_data(chain)
|
| 84 |
|
| 85 |
-
|
| 86 |
-
|
| 87 |
-
|
| 88 |
-
|
| 89 |
-
|
| 90 |
-
|
| 91 |
-
|
| 92 |
-
|
| 93 |
-
|
| 94 |
-
|
| 95 |
-
|
| 96 |
-
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
| 33 |
else:
|
| 34 |
print("dato default")
|
| 35 |
except:
|
| 36 |
+
loader = PyPDFLoader("https://www.sii.cl/normativa_legislacion/circulares/2024/circu3.pdf")
|
| 37 |
+
data = loader.load()
|
| 38 |
+
text_splitter = RecursiveCharacterTextSplitter(chunk_size=1000, chunk_overlap=150)
|
| 39 |
+
#Transformado a tipo de dato especifico para esto
|
| 40 |
+
docs = text_splitter.split_documents(data) # 'data' holds the text you want to split, split the text into documents using the text splitter.
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
| 41 |
|
| 42 |
+
#Modelo QA sentence similarity
|
| 43 |
+
modelPath = 'sentence-transformers/paraphrase-multilingual-MiniLM-L12-v2' #español
|
| 44 |
+
model_kwargs = {'device':'cpu'} # o cuda
|
| 45 |
+
encode_kwargs = {'normalize_embeddings': False}
|
| 46 |
+
|
| 47 |
+
#Embeddings que transforman a vectores densos multidimensionales las preguntas del SII
|
| 48 |
+
embeddings = HuggingFaceEmbeddings(
|
| 49 |
+
model_name=modelPath, # Ruta a modelo Pre entrenado
|
| 50 |
+
model_kwargs=model_kwargs, # Opciones de configuracion del modelo
|
| 51 |
+
encode_kwargs=encode_kwargs # Opciones de Encoding
|
| 52 |
+
)
|
| 53 |
+
|
| 54 |
+
#DB y retriever
|
| 55 |
+
db = FAISS.from_documents(docs, embeddings) # Create a retriever object from the 'db' with a search configuration where it retrieves up to 4 relevant splits/documents.
|
| 56 |
+
retriever = db.as_retriever(search_kwargs={"k": 3})
|
| 57 |
+
|
| 58 |
+
template = """Responde la pregunta basado unicamente en el siguiente contexto
|
| 59 |
+
|
| 60 |
+
{contexto}
|
| 61 |
+
|
| 62 |
+
Pregunta: {pregunta}
|
| 63 |
+
|
| 64 |
+
"""
|
| 65 |
+
|
| 66 |
+
#LLM
|
| 67 |
+
prompt = ChatPromptTemplate.from_template(template)
|
| 68 |
+
model = ChatOpenAI(model="gpt-3.5-turbo-1106", temperature=0)
|
| 69 |
+
|
| 70 |
+
def format_docs(docs):
|
| 71 |
+
return "\n\n".join([d.page_content for d in docs])
|
| 72 |
+
|
| 73 |
+
chain = (
|
| 74 |
+
{"contexto": retriever | format_docs, "pregunta": RunnablePassthrough()}
|
| 75 |
+
| prompt
|
| 76 |
+
| model
|
| 77 |
+
| StrOutputParser()
|
| 78 |
+
)
|
| 79 |
+
add_data(chain)
|
| 80 |
+
|
| 81 |
+
pregunta = st.text_area('Ingresa algun texto:')
|
| 82 |
+
tmp_button = st.button("CLICK")
|
| 83 |
+
if tmp_button: #Esperar al boton
|
| 84 |
+
out = chain.invoke(pregunta)
|
| 85 |
+
st.write(out)
|
| 86 |
+
#st.rerun() #Restart app
|
| 87 |
+
else:
|
| 88 |
+
st.stop()
|
| 89 |
+
st.write(st.session_state["chain"])
|