Update app.py
Browse files
app.py
CHANGED
@@ -1,63 +1,151 @@
|
|
1 |
import gradio as gr
|
|
|
2 |
from huggingface_hub import InferenceClient
|
3 |
|
4 |
-
|
5 |
-
|
6 |
-
|
7 |
-
|
|
|
|
|
|
|
8 |
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
9 |
|
10 |
def respond(
|
|
|
11 |
message,
|
|
|
12 |
history: list[tuple[str, str]],
|
|
|
13 |
system_message,
|
|
|
14 |
max_tokens,
|
|
|
15 |
temperature,
|
|
|
16 |
top_p,
|
|
|
|
|
|
|
17 |
):
|
|
|
18 |
messages = [{"role": "system", "content": system_message}]
|
19 |
|
20 |
-
for
|
21 |
-
|
22 |
-
|
23 |
-
|
24 |
-
messages.append({"role": "
|
|
|
|
|
|
|
|
|
25 |
|
26 |
messages.append({"role": "user", "content": message})
|
27 |
|
28 |
-
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
29 |
|
30 |
-
|
31 |
-
|
32 |
-
|
33 |
-
|
34 |
-
|
35 |
-
top_p=top_p,
|
36 |
-
):
|
37 |
-
token = message.choices[0].delta.content
|
38 |
|
39 |
-
response += token
|
40 |
yield response
|
41 |
|
42 |
-
|
43 |
-
|
44 |
-
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
45 |
demo = gr.ChatInterface(
|
|
|
46 |
respond,
|
|
|
47 |
additional_inputs=[
|
48 |
-
|
49 |
-
gr.
|
|
|
|
|
|
|
50 |
gr.Slider(minimum=0.1, maximum=4.0, value=0.7, step=0.1, label="Temperature"),
|
|
|
51 |
gr.Slider(
|
|
|
52 |
minimum=0.1,
|
|
|
53 |
maximum=1.0,
|
|
|
54 |
value=0.95,
|
|
|
55 |
step=0.05,
|
|
|
56 |
label="Top-p (nucleus sampling)",
|
|
|
57 |
),
|
|
|
|
|
|
|
58 |
],
|
|
|
|
|
|
|
|
|
|
|
59 |
)
|
60 |
|
|
|
61 |
|
62 |
-
if __name__ == "__main__":
|
63 |
demo.launch()
|
|
|
1 |
import gradio as gr
|
2 |
+
|
3 |
from huggingface_hub import InferenceClient
|
4 |
|
5 |
+
from transformers import AutoTokenizer, AutoModelForCausalLM
|
6 |
+
|
7 |
+
import torch
|
8 |
+
|
9 |
+
# Initialize the InferenceClient
|
10 |
+
|
11 |
+
client = InferenceClient("01-ai/Yi-Coder-9B-Chat")
|
12 |
|
13 |
+
# Initialize tokenizer and model
|
14 |
+
|
15 |
+
model_path = "01-ai/Yi-Coder-9B-Chat" # Make sure this is correct
|
16 |
+
|
17 |
+
tokenizer = AutoTokenizer.from_pretrained(model_path)
|
18 |
+
|
19 |
+
model = AutoModelForCausalLM.from_pretrained(model_path, device_map="auto").eval()
|
20 |
|
21 |
def respond(
|
22 |
+
|
23 |
message,
|
24 |
+
|
25 |
history: list[tuple[str, str]],
|
26 |
+
|
27 |
system_message,
|
28 |
+
|
29 |
max_tokens,
|
30 |
+
|
31 |
temperature,
|
32 |
+
|
33 |
top_p,
|
34 |
+
|
35 |
+
use_local_model: bool,
|
36 |
+
|
37 |
):
|
38 |
+
|
39 |
messages = [{"role": "system", "content": system_message}]
|
40 |
|
41 |
+
for user, assistant in history:
|
42 |
+
|
43 |
+
if user:
|
44 |
+
|
45 |
+
messages.append({"role": "user", "content": user})
|
46 |
+
|
47 |
+
if assistant:
|
48 |
+
|
49 |
+
messages.append({"role": "assistant", "content": assistant})
|
50 |
|
51 |
messages.append({"role": "user", "content": message})
|
52 |
|
53 |
+
if use_local_model:
|
54 |
+
|
55 |
+
# Use local model
|
56 |
+
|
57 |
+
input_ids = tokenizer.encode("".join([m["content"] for m in messages]), return_tensors="pt")
|
58 |
+
|
59 |
+
input_ids = input_ids.to(model.device)
|
60 |
+
|
61 |
+
|
62 |
+
|
63 |
+
with torch.no_grad():
|
64 |
+
|
65 |
+
output = model.generate(
|
66 |
+
|
67 |
+
input_ids,
|
68 |
+
|
69 |
+
max_new_tokens=max_tokens,
|
70 |
+
|
71 |
+
temperature=temperature,
|
72 |
+
|
73 |
+
top_p=top_p,
|
74 |
+
|
75 |
+
do_sample=True,
|
76 |
+
|
77 |
+
pad_token_id=tokenizer.eos_token_id,
|
78 |
|
79 |
+
)
|
80 |
+
|
81 |
+
|
82 |
+
|
83 |
+
response = tokenizer.decode(output[0], skip_special_tokens=True)
|
|
|
|
|
|
|
84 |
|
|
|
85 |
yield response
|
86 |
|
87 |
+
else:
|
88 |
+
|
89 |
+
# Use Hugging Face Inference API
|
90 |
+
|
91 |
+
response = ""
|
92 |
+
|
93 |
+
for message in client.text_generation(
|
94 |
+
|
95 |
+
"".join([m["content"] for m in messages]),
|
96 |
+
|
97 |
+
max_new_tokens=max_tokens,
|
98 |
+
|
99 |
+
stream=True,
|
100 |
+
|
101 |
+
temperature=temperature,
|
102 |
+
|
103 |
+
top_p=top_p,
|
104 |
+
|
105 |
+
):
|
106 |
+
|
107 |
+
response += message
|
108 |
+
|
109 |
+
yield response
|
110 |
+
|
111 |
+
# Create Gradio interface
|
112 |
+
|
113 |
demo = gr.ChatInterface(
|
114 |
+
|
115 |
respond,
|
116 |
+
|
117 |
additional_inputs=[
|
118 |
+
|
119 |
+
gr.Textbox(value="Odpowiadasz w Jezyku Polskim jesteś Coder/Developer/Programista tworzysz pełny kod..", label="System message"),
|
120 |
+
|
121 |
+
gr.Slider(minimum=1, maximum=2048, value=2048, step=1, label="Max new tokens"),
|
122 |
+
|
123 |
gr.Slider(minimum=0.1, maximum=4.0, value=0.7, step=0.1, label="Temperature"),
|
124 |
+
|
125 |
gr.Slider(
|
126 |
+
|
127 |
minimum=0.1,
|
128 |
+
|
129 |
maximum=1.0,
|
130 |
+
|
131 |
value=0.95,
|
132 |
+
|
133 |
step=0.05,
|
134 |
+
|
135 |
label="Top-p (nucleus sampling)",
|
136 |
+
|
137 |
),
|
138 |
+
|
139 |
+
gr.Checkbox(label="Use Local Model", value=False),
|
140 |
+
|
141 |
],
|
142 |
+
|
143 |
+
title="Advanced Chat Interface",
|
144 |
+
|
145 |
+
description="Chat with an AI model using either the Hugging Face Inference API or a local model.",
|
146 |
+
|
147 |
)
|
148 |
|
149 |
+
if name == "__main__":
|
150 |
|
|
|
151 |
demo.launch()
|