Spaces:
Runtime error
Runtime error
Update app.py
Browse files
app.py
CHANGED
@@ -1,6 +1,7 @@
|
|
1 |
import os
|
2 |
import requests
|
3 |
import json
|
|
|
4 |
|
5 |
import gradio as gr
|
6 |
from transformers import AutoTokenizer
|
@@ -44,6 +45,8 @@ HEADERS = {
|
|
44 |
PRESENCE_PENALTY=0
|
45 |
FREQUENCY_PENALTY=1.1
|
46 |
|
|
|
|
|
47 |
tokenizer = AutoTokenizer.from_pretrained("MediaTek-Research/Breeze-7B-Instruct-v0_1")
|
48 |
|
49 |
def refusal_condition(query):
|
@@ -153,7 +156,8 @@ with gr.Blocks() as demo:
|
|
153 |
"frequency_penalty": FREQUENCY_PENALTY,
|
154 |
"stream": True,
|
155 |
}
|
156 |
-
|
|
|
157 |
with requests.post(API_URL, headers=HEADERS, data=json.dumps(data), stream=True) as r:
|
158 |
for response in r.iter_lines():
|
159 |
if len(response) > 0:
|
@@ -169,6 +173,8 @@ with gr.Blocks() as demo:
|
|
169 |
history[-1][1] += delta
|
170 |
|
171 |
yield history
|
|
|
|
|
172 |
|
173 |
if history[-1][1].endswith('</s>'):
|
174 |
history[-1][1] = history[-1][1][:-4]
|
@@ -262,5 +268,5 @@ with gr.Blocks() as demo:
|
|
262 |
|
263 |
gr.Markdown(LICENSE)
|
264 |
|
265 |
-
demo.queue(concurrency_count=1, max_size=
|
266 |
demo.launch()
|
|
|
1 |
import os
|
2 |
import requests
|
3 |
import json
|
4 |
+
import time
|
5 |
|
6 |
import gradio as gr
|
7 |
from transformers import AutoTokenizer
|
|
|
45 |
PRESENCE_PENALTY=0
|
46 |
FREQUENCY_PENALTY=1.1
|
47 |
|
48 |
+
MAX_SEC = 30
|
49 |
+
|
50 |
tokenizer = AutoTokenizer.from_pretrained("MediaTek-Research/Breeze-7B-Instruct-v0_1")
|
51 |
|
52 |
def refusal_condition(query):
|
|
|
156 |
"frequency_penalty": FREQUENCY_PENALTY,
|
157 |
"stream": True,
|
158 |
}
|
159 |
+
|
160 |
+
start_time = time.time()
|
161 |
with requests.post(API_URL, headers=HEADERS, data=json.dumps(data), stream=True) as r:
|
162 |
for response in r.iter_lines():
|
163 |
if len(response) > 0:
|
|
|
173 |
history[-1][1] += delta
|
174 |
|
175 |
yield history
|
176 |
+
if time.time() - start_time > MAX_SEC:
|
177 |
+
break
|
178 |
|
179 |
if history[-1][1].endswith('</s>'):
|
180 |
history[-1][1] = history[-1][1][:-4]
|
|
|
268 |
|
269 |
gr.Markdown(LICENSE)
|
270 |
|
271 |
+
demo.queue(concurrency_count=1, max_size=128)
|
272 |
demo.launch()
|