Spaces:
Sleeping
Sleeping
Update app.py
Browse files
app.py
CHANGED
|
@@ -2,8 +2,7 @@ import gradio as gr
|
|
| 2 |
from huggingface_hub import InferenceClient
|
| 3 |
import os
|
| 4 |
|
| 5 |
-
#
|
| 6 |
-
# Zorg dat je 'Request Access' hebt gedaan op de HF pagina van dit model!
|
| 7 |
client = InferenceClient(
|
| 8 |
"meta-llama/Llama-3.1-8B-Instruct",
|
| 9 |
token=os.getenv("HF_TOKEN")
|
|
@@ -11,27 +10,27 @@ client = InferenceClient(
|
|
| 11 |
|
| 12 |
def respond(
|
| 13 |
message,
|
| 14 |
-
history,
|
| 15 |
system_message,
|
| 16 |
max_tokens,
|
| 17 |
temperature,
|
| 18 |
top_p,
|
| 19 |
):
|
| 20 |
-
# We beginnen met de system prompt
|
| 21 |
messages = [{"role": "system", "content": system_message}]
|
| 22 |
|
| 23 |
-
#
|
| 24 |
-
|
| 25 |
-
|
| 26 |
-
|
|
|
|
|
|
|
|
|
|
| 27 |
|
| 28 |
-
# Voeg de nieuwste vraag van de gebruiker toe
|
| 29 |
messages.append({"role": "user", "content": message})
|
| 30 |
|
| 31 |
response = ""
|
| 32 |
|
| 33 |
try:
|
| 34 |
-
# Roep de AI aan
|
| 35 |
for msg in client.chat_completion(
|
| 36 |
messages,
|
| 37 |
max_tokens=max_tokens,
|
|
@@ -43,12 +42,11 @@ def respond(
|
|
| 43 |
response += token
|
| 44 |
yield response
|
| 45 |
except Exception as e:
|
| 46 |
-
yield f"Verbindingsfout: {str(e)}.
|
| 47 |
|
| 48 |
-
#
|
| 49 |
demo = gr.ChatInterface(
|
| 50 |
respond,
|
| 51 |
-
type="messages", # DIT IS DE BELANGRIJKE FIX: Vertelt Gradio het nieuwe format te gebruiken
|
| 52 |
additional_inputs=[
|
| 53 |
gr.Textbox(
|
| 54 |
value="Bismillah. You are Noor AI, a wise Islamic scholar. Answer accurately based on Quran and Sunnah. End with Wa Allahu a'la wa a'lam.",
|
|
|
|
| 2 |
from huggingface_hub import InferenceClient
|
| 3 |
import os
|
| 4 |
|
| 5 |
+
# Gebruik Llama-3.1-8B-Instruct (Zorg dat je 'Request Access' hebt gedaan op HF!)
|
|
|
|
| 6 |
client = InferenceClient(
|
| 7 |
"meta-llama/Llama-3.1-8B-Instruct",
|
| 8 |
token=os.getenv("HF_TOKEN")
|
|
|
|
| 10 |
|
| 11 |
def respond(
|
| 12 |
message,
|
| 13 |
+
history,
|
| 14 |
system_message,
|
| 15 |
max_tokens,
|
| 16 |
temperature,
|
| 17 |
top_p,
|
| 18 |
):
|
|
|
|
| 19 |
messages = [{"role": "system", "content": system_message}]
|
| 20 |
|
| 21 |
+
# Veilige manier om geschiedenis te verwerken (werkt in Gradio 4, 5 en 6)
|
| 22 |
+
for val in history:
|
| 23 |
+
if isinstance(val, dict): # Moderne Gradio stijl
|
| 24 |
+
messages.append(val)
|
| 25 |
+
elif isinstance(val, (list, tuple)): # Klassieke Gradio stijl
|
| 26 |
+
if val[0]: messages.append({"role": "user", "content": val[0]})
|
| 27 |
+
if val[1]: messages.append({"role": "assistant", "content": val[1]})
|
| 28 |
|
|
|
|
| 29 |
messages.append({"role": "user", "content": message})
|
| 30 |
|
| 31 |
response = ""
|
| 32 |
|
| 33 |
try:
|
|
|
|
| 34 |
for msg in client.chat_completion(
|
| 35 |
messages,
|
| 36 |
max_tokens=max_tokens,
|
|
|
|
| 42 |
response += token
|
| 43 |
yield response
|
| 44 |
except Exception as e:
|
| 45 |
+
yield f"Verbindingsfout: {str(e)}. Controleer of je 'Request Access' hebt gedaan voor Llama-3.1 op Hugging Face."
|
| 46 |
|
| 47 |
+
# GEEN 'type' argument meer, zodat het nooit meer crasht op TypeError
|
| 48 |
demo = gr.ChatInterface(
|
| 49 |
respond,
|
|
|
|
| 50 |
additional_inputs=[
|
| 51 |
gr.Textbox(
|
| 52 |
value="Bismillah. You are Noor AI, a wise Islamic scholar. Answer accurately based on Quran and Sunnah. End with Wa Allahu a'la wa a'lam.",
|