Spaces:
Running
Running
prithivMLmods
commited on
Commit
•
79cade0
1
Parent(s):
240b17d
Create app.py
Browse files
app.py
ADDED
@@ -0,0 +1,75 @@
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
1 |
+
#refer llama recipes for more info https://github.com/huggingface/huggingface-llama-recipes/blob/main/inference-api.ipynb
|
2 |
+
#huggingface-llama-recipes : https://github.com/huggingface/huggingface-llama-recipes/tree/main
|
3 |
+
import gradio as gr
|
4 |
+
from openai import OpenAI
|
5 |
+
import os
|
6 |
+
|
7 |
+
css = '''
|
8 |
+
.gradio-container{max-width: 1000px !important}
|
9 |
+
h1{text-align:center}
|
10 |
+
footer {
|
11 |
+
visibility: hidden
|
12 |
+
}
|
13 |
+
'''
|
14 |
+
|
15 |
+
ACCESS_TOKEN = os.getenv("HF_TOKEN")
|
16 |
+
|
17 |
+
client = OpenAI(
|
18 |
+
base_url="https://api-inference.huggingface.co/v1/",
|
19 |
+
api_key=ACCESS_TOKEN,
|
20 |
+
)
|
21 |
+
|
22 |
+
def respond(
|
23 |
+
message,
|
24 |
+
history: list[tuple[str, str]],
|
25 |
+
system_message,
|
26 |
+
max_tokens,
|
27 |
+
temperature,
|
28 |
+
top_p,
|
29 |
+
):
|
30 |
+
messages = [{"role": "system", "content": system_message}]
|
31 |
+
|
32 |
+
for val in history:
|
33 |
+
if val[0]:
|
34 |
+
messages.append({"role": "user", "content": val[0]})
|
35 |
+
if val[1]:
|
36 |
+
messages.append({"role": "assistant", "content": val[1]})
|
37 |
+
|
38 |
+
messages.append({"role": "user", "content": message})
|
39 |
+
|
40 |
+
response = ""
|
41 |
+
|
42 |
+
for message in client.chat.completions.create(
|
43 |
+
model="meta-llama/Meta-Llama-3.1-8B-Instruct",
|
44 |
+
max_tokens=max_tokens,
|
45 |
+
stream=True,
|
46 |
+
temperature=temperature,
|
47 |
+
top_p=top_p,
|
48 |
+
messages=messages,
|
49 |
+
):
|
50 |
+
token = message.choices[0].delta.content
|
51 |
+
|
52 |
+
response += token
|
53 |
+
yield response
|
54 |
+
|
55 |
+
demo = gr.ChatInterface(
|
56 |
+
respond,
|
57 |
+
additional_inputs=[
|
58 |
+
gr.Textbox(value="", label="System message", visible=False),
|
59 |
+
gr.Slider(minimum=1, maximum=2048, value=512, step=1, label="Max new tokens", visible=False),
|
60 |
+
gr.Slider(minimum=0.1, maximum=4.0, value=0.7, step=0.1, label="Temperature", visible=False),
|
61 |
+
gr.Slider(
|
62 |
+
minimum=0.1,
|
63 |
+
maximum=1.0,
|
64 |
+
value=0.95,
|
65 |
+
step=0.05,
|
66 |
+
label="Top-P",
|
67 |
+
visible=False,
|
68 |
+
),
|
69 |
+
|
70 |
+
],
|
71 |
+
css=css,
|
72 |
+
theme="allenai/gradio-theme",
|
73 |
+
)
|
74 |
+
if __name__ == "__main__":
|
75 |
+
demo.launch()
|