Spaces:
Running
on
Zero
Running
on
Zero
Update app.py
Browse files
app.py
CHANGED
@@ -6,27 +6,17 @@ from threading import Thread
|
|
6 |
from transformers import AutoModelForCausalLM, AutoTokenizer, TextIteratorStreamer
|
7 |
|
8 |
|
9 |
-
TITLE = "
|
10 |
-
|
11 |
-
DESCRIPTION = "<h3><center>Visit <a href='' target='_blank'>LLaMA Factory</a> for details.</center></h3>"
|
12 |
-
|
13 |
-
CSS = r"""
|
14 |
-
.duplicate-button {
|
15 |
-
margin: auto !important;
|
16 |
-
color: white !important;
|
17 |
-
background: black !important;
|
18 |
-
border-radius: 100vh !important;
|
19 |
-
}
|
20 |
-
"""
|
21 |
|
|
|
22 |
|
23 |
tokenizer = AutoTokenizer.from_pretrained("shenzhi-wang/Llama3-8B-Chinese-Chat")
|
24 |
model = AutoModelForCausalLM.from_pretrained("shenzhi-wang/Llama3-8B-Chinese-Chat", device_map="auto")
|
25 |
|
26 |
|
27 |
-
@spaces.GPU
|
28 |
-
def stream_chat(message: str, history: list, temperature: float, max_new_tokens: int):
|
29 |
-
conversation = []
|
30 |
for prompt, answer in history:
|
31 |
conversation.extend([{"role": "user", "content": prompt}, {"role": "assistant", "content": answer}])
|
32 |
|
@@ -54,16 +44,17 @@ def stream_chat(message: str, history: list, temperature: float, max_new_tokens:
|
|
54 |
yield output
|
55 |
|
56 |
|
57 |
-
with gr.Blocks(fill_height=True
|
58 |
-
gr.HTML(TITLE)
|
59 |
-
gr.HTML(DESCRIPTION)
|
60 |
-
gr.DuplicateButton(value="Duplicate Space for private use", elem_classes="duplicate-button")
|
61 |
-
|
62 |
gr.ChatInterface(
|
63 |
fn=stream_chat,
|
64 |
fill_height=True,
|
65 |
additional_inputs_accordion=gr.Accordion(label="⚙️ Parameters", open=False, render=False),
|
66 |
additional_inputs=[
|
|
|
|
|
|
|
|
|
|
|
67 |
gr.Slider(
|
68 |
minimum=0,
|
69 |
maximum=1,
|
@@ -82,13 +73,12 @@ with gr.Blocks(fill_height=True, css=CSS) as demo:
|
|
82 |
),
|
83 |
],
|
84 |
examples=[
|
85 |
-
[
|
86 |
-
[
|
87 |
-
['What is 9,000 * 9,000?'],
|
88 |
-
['Write a pun-filled happy birthday message to my friend Alex.'],
|
89 |
-
['Justify why a penguin might make a good king of the jungle.']
|
90 |
],
|
91 |
cache_examples=False,
|
|
|
|
|
92 |
)
|
93 |
|
94 |
|
|
|
6 |
from transformers import AutoModelForCausalLM, AutoTokenizer, TextIteratorStreamer
|
7 |
|
8 |
|
9 |
+
TITLE = "Chat with Llama3-8B-Chinese"
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
10 |
|
11 |
+
DESCRIPTION = "Visit <a href='https://huggingface.co/shenzhi-wang/Llama3-8B-Chinese-Chat' target='_blank'>our model page</a> for details."
|
12 |
|
13 |
tokenizer = AutoTokenizer.from_pretrained("shenzhi-wang/Llama3-8B-Chinese-Chat")
|
14 |
model = AutoModelForCausalLM.from_pretrained("shenzhi-wang/Llama3-8B-Chinese-Chat", device_map="auto")
|
15 |
|
16 |
|
17 |
+
@spaces.GPU
|
18 |
+
def stream_chat(message: str, history: list, system: str, temperature: float, max_new_tokens: int):
|
19 |
+
conversation = [{"role": "system", "content": system}]
|
20 |
for prompt, answer in history:
|
21 |
conversation.extend([{"role": "user", "content": prompt}, {"role": "assistant", "content": answer}])
|
22 |
|
|
|
44 |
yield output
|
45 |
|
46 |
|
47 |
+
with gr.Blocks(fill_height=True) as demo:
|
|
|
|
|
|
|
|
|
48 |
gr.ChatInterface(
|
49 |
fn=stream_chat,
|
50 |
fill_height=True,
|
51 |
additional_inputs_accordion=gr.Accordion(label="⚙️ Parameters", open=False, render=False),
|
52 |
additional_inputs=[
|
53 |
+
gr.Text(
|
54 |
+
value="You are a helpful assistant.",
|
55 |
+
label="System",
|
56 |
+
render=False,
|
57 |
+
),
|
58 |
gr.Slider(
|
59 |
minimum=0,
|
60 |
maximum=1,
|
|
|
73 |
),
|
74 |
],
|
75 |
examples=[
|
76 |
+
["我的蓝牙耳机坏了,我该去看牙科还是耳鼻喉科?"],
|
77 |
+
["今日行军进展如何", "扮演诸葛亮和我对话。"],
|
|
|
|
|
|
|
78 |
],
|
79 |
cache_examples=False,
|
80 |
+
title=TITLE,
|
81 |
+
description=DESCRIPTION,
|
82 |
)
|
83 |
|
84 |
|