cuda handling
Browse files
app.py
CHANGED
|
@@ -1,4 +1,5 @@
|
|
| 1 |
import gradio as gr
|
|
|
|
| 2 |
import random
|
| 3 |
import time
|
| 4 |
from transformers import pipeline
|
|
@@ -7,7 +8,7 @@ from transformers import pipeline
|
|
| 7 |
generator = pipeline(
|
| 8 |
'text-generation',
|
| 9 |
model="heegyu/gorani-v0",
|
| 10 |
-
device="cuda:0"
|
| 11 |
)
|
| 12 |
|
| 13 |
def query(message, chat_history, max_turn=4):
|
|
@@ -15,7 +16,7 @@ def query(message, chat_history, max_turn=4):
|
|
| 15 |
if len(chat_history) > max_turn:
|
| 16 |
chat_history = chat_history[-max_turn:]
|
| 17 |
for user, bot in chat_history:
|
| 18 |
-
prompt.append(f"<usr> {user}")
|
| 19 |
prompt.append(f"<bot> {bot}")
|
| 20 |
prompt.append(f"<usr> {message}")
|
| 21 |
prompt = "\n".join(prompt) + "\n<bot>"
|
|
|
|
| 1 |
import gradio as gr
|
| 2 |
+
import torch
|
| 3 |
import random
|
| 4 |
import time
|
| 5 |
from transformers import pipeline
|
|
|
|
| 8 |
generator = pipeline(
|
| 9 |
'text-generation',
|
| 10 |
model="heegyu/gorani-v0",
|
| 11 |
+
device="cuda:0" if torch.cuda.is_available() else 'cpu'
|
| 12 |
)
|
| 13 |
|
| 14 |
def query(message, chat_history, max_turn=4):
|
|
|
|
| 16 |
if len(chat_history) > max_turn:
|
| 17 |
chat_history = chat_history[-max_turn:]
|
| 18 |
for user, bot in chat_history:
|
| 19 |
+
prompt.append(f"<usr> 대답해: {user}")
|
| 20 |
prompt.append(f"<bot> {bot}")
|
| 21 |
prompt.append(f"<usr> {message}")
|
| 22 |
prompt = "\n".join(prompt) + "\n<bot>"
|