code
Browse files
app.py
CHANGED
|
@@ -5,7 +5,7 @@ import gradio as gr
|
|
| 5 |
import os
|
| 6 |
from threading import Thread
|
| 7 |
|
| 8 |
-
|
| 9 |
|
| 10 |
os.environ['PYTORCH_CUDA_ALLOC_CONF'] = 'expandable_segments:True'
|
| 11 |
|
|
@@ -50,6 +50,7 @@ def respond(
|
|
| 50 |
temperature=temperature,
|
| 51 |
streamer=streamer,
|
| 52 |
top_p=top_p,
|
|
|
|
| 53 |
do_sample=True,
|
| 54 |
)
|
| 55 |
|
|
|
|
| 5 |
import os
|
| 6 |
from threading import Thread
|
| 7 |
|
| 8 |
+
os.system("rm -rf /data-nvme/zerogpu-offload/*")
|
| 9 |
|
| 10 |
os.environ['PYTORCH_CUDA_ALLOC_CONF'] = 'expandable_segments:True'
|
| 11 |
|
|
|
|
| 50 |
temperature=temperature,
|
| 51 |
streamer=streamer,
|
| 52 |
top_p=top_p,
|
| 53 |
+
return_full_text=False,
|
| 54 |
do_sample=True,
|
| 55 |
)
|
| 56 |
|