Spaces:
Sleeping
Sleeping
Commit 路
939a60b
1
Parent(s): 1d90bff
arreglos 5
Browse files
app.py
CHANGED
|
@@ -1,50 +1,49 @@
|
|
| 1 |
import gradio as gr
|
| 2 |
from huggingface_hub import InferenceClient
|
| 3 |
|
| 4 |
-
|
| 5 |
-
For more information on `huggingface_hub` Inference API support, please check the docs: https://huggingface.co/docs/huggingface_hub/v0.22.2/en/guides/inference
|
| 6 |
-
"""
|
| 7 |
client = InferenceClient("AuriLab/gpt-bi-instruct-cesar")
|
| 8 |
|
|
|
|
|
|
|
| 9 |
|
| 10 |
-
def respond(
|
| 11 |
-
|
| 12 |
-
history: list[tuple[str, str]],
|
| 13 |
-
):
|
| 14 |
messages = [{"role": "system", "content": "Gpt-Bi zara, AuriLabsek sortutako assitente digitala."}]
|
| 15 |
-
|
| 16 |
for val in history:
|
| 17 |
if val[0]:
|
| 18 |
messages.append({"role": "user", "content": val[0]})
|
| 19 |
if val[1]:
|
| 20 |
messages.append({"role": "assistant", "content": val[1]})
|
| 21 |
-
|
| 22 |
messages.append({"role": "user", "content": message})
|
| 23 |
|
| 24 |
response = ""
|
| 25 |
-
|
| 26 |
-
for
|
| 27 |
messages,
|
| 28 |
-
max_tokens=60,
|
| 29 |
stream=True,
|
| 30 |
temperature=0.7,
|
| 31 |
presence_penalty=1.5,
|
| 32 |
top_p=0.85,
|
| 33 |
):
|
| 34 |
-
token =
|
| 35 |
-
|
| 36 |
response += token
|
| 37 |
-
yield response
|
| 38 |
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
| 39 |
|
| 40 |
-
|
| 41 |
-
For information on how to customize the ChatInterface, peruse the gradio docs: https://www.gradio.app/docs/chatinterface
|
| 42 |
-
"""
|
| 43 |
demo = gr.ChatInterface(
|
| 44 |
respond,
|
| 45 |
title="Demo GPT-BI Instruct",
|
| 46 |
)
|
| 47 |
|
| 48 |
-
|
| 49 |
if __name__ == "__main__":
|
| 50 |
demo.launch()
|
|
|
|
| 1 |
import gradio as gr
|
| 2 |
from huggingface_hub import InferenceClient
|
| 3 |
|
| 4 |
+
# Inicializa el cliente con el modelo deseado
|
|
|
|
|
|
|
| 5 |
client = InferenceClient("AuriLab/gpt-bi-instruct-cesar")
|
| 6 |
|
| 7 |
+
# Define las secuencias de parada que detendr谩n la generaci贸n
|
| 8 |
+
stop_sequences = [".", "?", ".\n", "\n\n"]
|
| 9 |
|
| 10 |
+
def respond(message, history: list[tuple[str, str]]):
|
| 11 |
+
# Construir el historial de mensajes para la conversaci贸n
|
|
|
|
|
|
|
| 12 |
messages = [{"role": "system", "content": "Gpt-Bi zara, AuriLabsek sortutako assitente digitala."}]
|
|
|
|
| 13 |
for val in history:
|
| 14 |
if val[0]:
|
| 15 |
messages.append({"role": "user", "content": val[0]})
|
| 16 |
if val[1]:
|
| 17 |
messages.append({"role": "assistant", "content": val[1]})
|
|
|
|
| 18 |
messages.append({"role": "user", "content": message})
|
| 19 |
|
| 20 |
response = ""
|
| 21 |
+
# Inicia la generaci贸n en modo streaming
|
| 22 |
+
for token_msg in client.chat_completion(
|
| 23 |
messages,
|
|
|
|
| 24 |
stream=True,
|
| 25 |
temperature=0.7,
|
| 26 |
presence_penalty=1.5,
|
| 27 |
top_p=0.85,
|
| 28 |
):
|
| 29 |
+
token = token_msg.choices[0].delta.content
|
|
|
|
| 30 |
response += token
|
|
|
|
| 31 |
|
| 32 |
+
# Comprueba si la respuesta termina con alguna de las secuencias de parada
|
| 33 |
+
for stop_seq in stop_sequences:
|
| 34 |
+
if response.endswith(stop_seq):
|
| 35 |
+
# Opcionalmente, se puede remover la secuencia de parada final
|
| 36 |
+
response = response[:-len(stop_seq)]
|
| 37 |
+
yield response
|
| 38 |
+
return # Detener la generaci贸n
|
| 39 |
+
|
| 40 |
+
yield response
|
| 41 |
|
| 42 |
+
# Configuraci贸n de la interfaz de chat con Gradio
|
|
|
|
|
|
|
| 43 |
demo = gr.ChatInterface(
|
| 44 |
respond,
|
| 45 |
title="Demo GPT-BI Instruct",
|
| 46 |
)
|
| 47 |
|
|
|
|
| 48 |
if __name__ == "__main__":
|
| 49 |
demo.launch()
|