Spaces:
Sleeping
Sleeping
Update app.py
Browse files
app.py
CHANGED
@@ -1,82 +1,82 @@
|
|
1 |
-
import os
|
2 |
-
import gradio as gr
|
3 |
-
import copy
|
4 |
-
from llama_cpp import Llama
|
5 |
-
from huggingface_hub import hf_hub_download
|
6 |
-
|
7 |
-
|
8 |
-
llm = Llama(
|
9 |
-
model_path = "./dpo-internlm2-1_8b.Q4_K_M.gguf",
|
10 |
-
n_ctx=2048,
|
11 |
-
n_threads=2,
|
12 |
-
seed=57199,
|
13 |
-
batch_size=1024,
|
14 |
-
cache=True
|
15 |
-
)
|
16 |
-
|
17 |
-
|
18 |
-
def generate_text(
|
19 |
-
message,
|
20 |
-
history: list[tuple[str, str]],
|
21 |
-
system_message,
|
22 |
-
max_tokens,
|
23 |
-
temperature,
|
24 |
-
top_p,
|
25 |
-
):
|
26 |
-
temp = ""
|
27 |
-
input_prompt = f"
|
28 |
-
for interaction in history:
|
29 |
-
input_prompt = input_prompt + str(interaction[0]) + "
|
30 |
-
|
31 |
-
input_prompt = input_prompt + str(message) + "
|
32 |
-
|
33 |
-
output = llm(
|
34 |
-
input_prompt,
|
35 |
-
temperature=temperature,
|
36 |
-
top_p=top_p,
|
37 |
-
top_k=40,
|
38 |
-
repeat_penalty=1.2,
|
39 |
-
max_tokens=max_tokens,
|
40 |
-
stop=["
|
41 |
-
stream=True,
|
42 |
-
|
43 |
-
)
|
44 |
-
for out in output:
|
45 |
-
stream = copy.deepcopy(out)
|
46 |
-
temp += stream["choices"][0]["text"]
|
47 |
-
yield temp
|
48 |
-
|
49 |
-
|
50 |
-
demo = gr.ChatInterface(
|
51 |
-
generate_text,
|
52 |
-
title="DPO-Internlm2-1_8B",
|
53 |
-
description="Running LLM with llama-cpp-python",
|
54 |
-
theme = gr.themes.Soft(),
|
55 |
-
examples=[
|
56 |
-
['How to setup a human base on Mars? Give short answer.'],
|
57 |
-
['Explain theory of relativity to me like I’m 8 years old.'],
|
58 |
-
['What is 9,000 * 9,000?'],
|
59 |
-
['Write a pun-filled happy birthday message to my friend Alex.'],
|
60 |
-
['Justify why a penguin might make a good king of the jungle.']
|
61 |
-
],
|
62 |
-
cache_examples=False,
|
63 |
-
retry_btn=None,
|
64 |
-
undo_btn="Delete Previous",
|
65 |
-
clear_btn="Clear",
|
66 |
-
additional_inputs=[
|
67 |
-
gr.Textbox(value="You are a friendly Chatbot.", label="System message"),
|
68 |
-
gr.Slider(minimum=1, maximum=2048, value=512, step=1, label="Max new tokens"),
|
69 |
-
gr.Slider(minimum=0.1, maximum=4.0, value=0.3, step=0.1, label="Temperature"),
|
70 |
-
gr.Slider(
|
71 |
-
minimum=0.1,
|
72 |
-
maximum=1.0,
|
73 |
-
value=0.90,
|
74 |
-
step=0.05,
|
75 |
-
label="Top-p (nucleus sampling)",
|
76 |
-
),
|
77 |
-
],
|
78 |
-
)
|
79 |
-
|
80 |
-
|
81 |
-
if __name__ == "__main__":
|
82 |
-
demo.launch()
|
|
|
1 |
+
import os
|
2 |
+
import gradio as gr
|
3 |
+
import copy
|
4 |
+
from llama_cpp import Llama
|
5 |
+
from huggingface_hub import hf_hub_download
|
6 |
+
|
7 |
+
|
8 |
+
llm = Llama(
|
9 |
+
model_path = "./dpo-internlm2-1_8b.Q4_K_M.gguf",
|
10 |
+
n_ctx=2048,
|
11 |
+
n_threads=2,
|
12 |
+
seed=57199,
|
13 |
+
batch_size=1024,
|
14 |
+
cache=True
|
15 |
+
)
|
16 |
+
|
17 |
+
|
18 |
+
def generate_text(
|
19 |
+
message,
|
20 |
+
history: list[tuple[str, str]],
|
21 |
+
system_message,
|
22 |
+
max_tokens,
|
23 |
+
temperature,
|
24 |
+
top_p,
|
25 |
+
):
|
26 |
+
temp = ""
|
27 |
+
input_prompt = f"<s> <|im_start|>system:{system_message}<|im_end|>"
|
28 |
+
for interaction in history:
|
29 |
+
input_prompt = input_prompt + "<|im_start|>user:" + str(interaction[0]) + "<|im_end|><|im_start|>" + "assistant:" + str(interaction[1]) + "<|im_end|>"
|
30 |
+
|
31 |
+
input_prompt = input_prompt + "<|im_start|>user:" + str(message) + "<|im_end|><|im_start|>assistant:"
|
32 |
+
|
33 |
+
output = llm(
|
34 |
+
input_prompt,
|
35 |
+
temperature=temperature,
|
36 |
+
top_p=top_p,
|
37 |
+
top_k=40,
|
38 |
+
repeat_penalty=1.2,
|
39 |
+
max_tokens=max_tokens,
|
40 |
+
stop=["<|im_end|>"],
|
41 |
+
stream=True,
|
42 |
+
|
43 |
+
)
|
44 |
+
for out in output:
|
45 |
+
stream = copy.deepcopy(out)
|
46 |
+
temp += stream["choices"][0]["text"]
|
47 |
+
yield temp
|
48 |
+
|
49 |
+
|
50 |
+
demo = gr.ChatInterface(
|
51 |
+
generate_text,
|
52 |
+
title="DPO-Internlm2-1_8B",
|
53 |
+
description="Running LLM with llama-cpp-python",
|
54 |
+
theme = gr.themes.Soft(),
|
55 |
+
examples=[
|
56 |
+
['How to setup a human base on Mars? Give short answer.'],
|
57 |
+
['Explain theory of relativity to me like I’m 8 years old.'],
|
58 |
+
['What is 9,000 * 9,000?'],
|
59 |
+
['Write a pun-filled happy birthday message to my friend Alex.'],
|
60 |
+
['Justify why a penguin might make a good king of the jungle.']
|
61 |
+
],
|
62 |
+
cache_examples=False,
|
63 |
+
retry_btn=None,
|
64 |
+
undo_btn="Delete Previous",
|
65 |
+
clear_btn="Clear",
|
66 |
+
additional_inputs=[
|
67 |
+
gr.Textbox(value="You are a friendly Chatbot.", label="System message"),
|
68 |
+
gr.Slider(minimum=1, maximum=2048, value=512, step=1, label="Max new tokens"),
|
69 |
+
gr.Slider(minimum=0.1, maximum=4.0, value=0.3, step=0.1, label="Temperature"),
|
70 |
+
gr.Slider(
|
71 |
+
minimum=0.1,
|
72 |
+
maximum=1.0,
|
73 |
+
value=0.90,
|
74 |
+
step=0.05,
|
75 |
+
label="Top-p (nucleus sampling)",
|
76 |
+
),
|
77 |
+
],
|
78 |
+
)
|
79 |
+
|
80 |
+
|
81 |
+
if __name__ == "__main__":
|
82 |
+
demo.launch()
|