Spaces:
Runtime error
Runtime error
ehristoforu
commited on
Commit
•
1e11dd2
1
Parent(s):
8bdbddd
Upload 3 files
Browse files- model (1).py +57 -0
- requirements (14).txt +9 -0
- style (2).css +16 -0
model (1).py
ADDED
@@ -0,0 +1,57 @@
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
1 |
+
import os
|
2 |
+
from typing import Iterator
|
3 |
+
|
4 |
+
from text_generation import Client
|
5 |
+
|
6 |
+
model_id = 'codellama/CodeLlama-34b-Instruct-hf'
|
7 |
+
|
8 |
+
API_URL = "https://api-inference.huggingface.co/models/" + model_id
|
9 |
+
HF_TOKEN = os.environ.get("HF_READ_TOKEN", None)
|
10 |
+
|
11 |
+
client = Client(
|
12 |
+
API_URL,
|
13 |
+
headers={"Authorization": f"Bearer {HF_TOKEN}"},
|
14 |
+
)
|
15 |
+
EOS_STRING = "</s>"
|
16 |
+
EOT_STRING = "<EOT>"
|
17 |
+
|
18 |
+
|
19 |
+
def get_prompt(message: str, chat_history: list[tuple[str, str]],
|
20 |
+
system_prompt: str) -> str:
|
21 |
+
texts = [f'<s>[INST] <<SYS>>\n{system_prompt}\n<</SYS>>\n\n']
|
22 |
+
# The first user input is _not_ stripped
|
23 |
+
do_strip = False
|
24 |
+
for user_input, response in chat_history:
|
25 |
+
user_input = user_input.strip() if do_strip else user_input
|
26 |
+
do_strip = True
|
27 |
+
texts.append(f'{user_input} [/INST] {response.strip()} </s><s>[INST] ')
|
28 |
+
message = message.strip() if do_strip else message
|
29 |
+
texts.append(f'{message} [/INST]')
|
30 |
+
return ''.join(texts)
|
31 |
+
|
32 |
+
|
33 |
+
def run(message: str,
|
34 |
+
chat_history: list[tuple[str, str]],
|
35 |
+
system_prompt: str,
|
36 |
+
max_new_tokens: int = 1024,
|
37 |
+
temperature: float = 0.1,
|
38 |
+
top_p: float = 0.9,
|
39 |
+
top_k: int = 50) -> Iterator[str]:
|
40 |
+
prompt = get_prompt(message, chat_history, system_prompt)
|
41 |
+
|
42 |
+
generate_kwargs = dict(
|
43 |
+
max_new_tokens=max_new_tokens,
|
44 |
+
do_sample=True,
|
45 |
+
top_p=top_p,
|
46 |
+
top_k=top_k,
|
47 |
+
temperature=temperature,
|
48 |
+
)
|
49 |
+
stream = client.generate_stream(prompt, **generate_kwargs)
|
50 |
+
output = ""
|
51 |
+
for response in stream:
|
52 |
+
if any([end_token in response.token.text for end_token in [EOS_STRING, EOT_STRING]]):
|
53 |
+
return output
|
54 |
+
else:
|
55 |
+
output += response.token.text
|
56 |
+
yield output
|
57 |
+
return output
|
requirements (14).txt
ADDED
@@ -0,0 +1,9 @@
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
1 |
+
accelerate
|
2 |
+
bitsandbytes
|
3 |
+
gradio
|
4 |
+
protobuf
|
5 |
+
scipy
|
6 |
+
sentencepiece
|
7 |
+
torch
|
8 |
+
text_generation
|
9 |
+
git+https://github.com/huggingface/transformers@main
|
style (2).css
ADDED
@@ -0,0 +1,16 @@
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
1 |
+
h1 {
|
2 |
+
text-align: center;
|
3 |
+
}
|
4 |
+
|
5 |
+
#duplicate-button {
|
6 |
+
margin: auto;
|
7 |
+
color: white;
|
8 |
+
background: #1565c0;
|
9 |
+
border-radius: 100vh;
|
10 |
+
}
|
11 |
+
|
12 |
+
#component-0 {
|
13 |
+
max-width: 900px;
|
14 |
+
margin: auto;
|
15 |
+
padding-top: 1.5rem;
|
16 |
+
}
|