Create app.py
Browse files
app.py
ADDED
@@ -0,0 +1,216 @@
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
1 |
+
import os
|
2 |
+
import time
|
3 |
+
from typing import List, Tuple, Optional, Dict
|
4 |
+
|
5 |
+
import google.generativeai as genai
|
6 |
+
import gradio as gr
|
7 |
+
from PIL import Image
|
8 |
+
|
9 |
+
print("google-generativeai:", genai.__version__)
|
10 |
+
|
11 |
+
GG_API_KEY = os.environ.get("GG_API_KEY")
|
12 |
+
oaiusr = os.environ.get("OAI_USR")
|
13 |
+
oaipwd = os.environ.get("OAI_PWD")
|
14 |
+
|
15 |
+
TITLE = """<h2 align="center">Tomoniai's Gemini Pro Chat</h2>"""
|
16 |
+
|
17 |
+
AVATAR_IMAGES = ("./user.png", "./botg.png")
|
18 |
+
|
19 |
+
IMAGE_WIDTH = 512
|
20 |
+
|
21 |
+
|
22 |
+
def preprocess_stop_sequences(stop_sequences: str) -> Optional[List[str]]:
|
23 |
+
if not stop_sequences:
|
24 |
+
return None
|
25 |
+
return [sequence.strip() for sequence in stop_sequences.split(",")]
|
26 |
+
|
27 |
+
|
28 |
+
def preprocess_image(image: Image.Image) -> Optional[Image.Image]:
|
29 |
+
image_height = int(image.height * IMAGE_WIDTH / image.width)
|
30 |
+
return image.resize((IMAGE_WIDTH, image_height))
|
31 |
+
|
32 |
+
|
33 |
+
def preprocess_chat_history(
|
34 |
+
history: List[Tuple[Optional[str], Optional[str]]]
|
35 |
+
) -> List[Dict[str, List[str]]]:
|
36 |
+
messages = []
|
37 |
+
for user_message, model_message in history:
|
38 |
+
if user_message is not None:
|
39 |
+
messages.append({'role': 'user', 'parts': [user_message]})
|
40 |
+
if model_message is not None:
|
41 |
+
messages.append({'role': 'model', 'parts': [model_message]})
|
42 |
+
return messages
|
43 |
+
|
44 |
+
|
45 |
+
def user(text_prompt: str, chatbot: List[Tuple[str, str]]):
|
46 |
+
return "", chatbot + [[text_prompt, None]]
|
47 |
+
|
48 |
+
|
49 |
+
def bot(
|
50 |
+
google_key: str,
|
51 |
+
image_prompt: Optional[Image.Image],
|
52 |
+
temperature: float,
|
53 |
+
max_output_tokens: int,
|
54 |
+
stop_sequences: str,
|
55 |
+
top_k: int,
|
56 |
+
top_p: float,
|
57 |
+
chatbot: List[Tuple[str, str]]
|
58 |
+
):
|
59 |
+
google_key = google_key if google_key else GG_API_KEY
|
60 |
+
|
61 |
+
text_prompt = chatbot[-1][0]
|
62 |
+
genai.configure(api_key=google_key)
|
63 |
+
generation_config = genai.types.GenerationConfig(
|
64 |
+
temperature=temperature,
|
65 |
+
max_output_tokens=max_output_tokens,
|
66 |
+
stop_sequences=preprocess_stop_sequences(stop_sequences=stop_sequences),
|
67 |
+
top_k=top_k,
|
68 |
+
top_p=top_p)
|
69 |
+
|
70 |
+
if image_prompt is None:
|
71 |
+
model = genai.GenerativeModel('gemini-pro')
|
72 |
+
response = model.generate_content(
|
73 |
+
preprocess_chat_history(chatbot),
|
74 |
+
stream=True,
|
75 |
+
generation_config=generation_config)
|
76 |
+
response.resolve()
|
77 |
+
else:
|
78 |
+
image_prompt = preprocess_image(image_prompt)
|
79 |
+
model = genai.GenerativeModel('gemini-pro-vision')
|
80 |
+
response = model.generate_content(
|
81 |
+
contents=[text_prompt, image_prompt],
|
82 |
+
stream=True,
|
83 |
+
generation_config=generation_config)
|
84 |
+
response.resolve()
|
85 |
+
|
86 |
+
# streaming effect
|
87 |
+
chatbot[-1][1] = ""
|
88 |
+
for chunk in response:
|
89 |
+
for i in range(0, len(chunk.text), 10):
|
90 |
+
section = chunk.text[i:i + 10]
|
91 |
+
chatbot[-1][1] += section
|
92 |
+
time.sleep(0.01)
|
93 |
+
yield chatbot
|
94 |
+
|
95 |
+
|
96 |
+
image_prompt_component = gr.Image(type="pil", label="Image", scale=1, height=400)
|
97 |
+
chatbot_component = gr.Chatbot(
|
98 |
+
label='Gemini',
|
99 |
+
bubble_full_width=False,
|
100 |
+
avatar_images=AVATAR_IMAGES,
|
101 |
+
scale=2,
|
102 |
+
height=400
|
103 |
+
)
|
104 |
+
text_prompt_component = gr.Textbox(
|
105 |
+
placeholder="Hi there!",
|
106 |
+
label="Ask me anything and press Enter"
|
107 |
+
)
|
108 |
+
run_button_component = gr.Button()
|
109 |
+
temperature_component = gr.Slider(
|
110 |
+
minimum=0,
|
111 |
+
maximum=1.0,
|
112 |
+
value=0.4,
|
113 |
+
step=0.05,
|
114 |
+
label="Temperature",
|
115 |
+
info=(
|
116 |
+
"Temperature controls the degree of randomness in token selection. Lower "
|
117 |
+
"temperatures are good for prompts that expect a true or correct response, "
|
118 |
+
"while higher temperatures can lead to more diverse or unexpected results. "
|
119 |
+
))
|
120 |
+
max_output_tokens_component = gr.Slider(
|
121 |
+
minimum=1,
|
122 |
+
maximum=2048,
|
123 |
+
value=1024,
|
124 |
+
step=1,
|
125 |
+
label="Token limit",
|
126 |
+
info=(
|
127 |
+
"Token limit determines the maximum amount of text output from one prompt. A "
|
128 |
+
"token is approximately four characters. The default value is 2048."
|
129 |
+
))
|
130 |
+
stop_sequences_component = gr.Textbox(
|
131 |
+
label="Add stop sequence",
|
132 |
+
value="",
|
133 |
+
type="text",
|
134 |
+
placeholder="STOP, END",
|
135 |
+
info=(
|
136 |
+
"A stop sequence is a series of characters (including spaces) that stops "
|
137 |
+
"response generation if the model encounters it. The sequence is not included "
|
138 |
+
"as part of the response. You can add up to five stop sequences."
|
139 |
+
))
|
140 |
+
top_k_component = gr.Slider(
|
141 |
+
minimum=1,
|
142 |
+
maximum=40,
|
143 |
+
value=32,
|
144 |
+
step=1,
|
145 |
+
label="Top-K",
|
146 |
+
info=(
|
147 |
+
"Top-k changes how the model selects tokens for output. A top-k of 1 means the "
|
148 |
+
"selected token is the most probable among all tokens in the model’s "
|
149 |
+
"vocabulary (also called greedy decoding), while a top-k of 3 means that the "
|
150 |
+
"next token is selected from among the 3 most probable tokens (using "
|
151 |
+
"temperature)."
|
152 |
+
))
|
153 |
+
top_p_component = gr.Slider(
|
154 |
+
minimum=0,
|
155 |
+
maximum=1,
|
156 |
+
value=1,
|
157 |
+
step=0.01,
|
158 |
+
label="Top-P",
|
159 |
+
info=(
|
160 |
+
"Top-p changes how the model selects tokens for output. Tokens are selected "
|
161 |
+
"from most probable to least until the sum of their probabilities equals the "
|
162 |
+
"top-p value. For example, if tokens A, B, and C have a probability of .3, .2, "
|
163 |
+
"and .1 and the top-p value is .5, then the model will select either A or B as "
|
164 |
+
"the next token (using temperature). "
|
165 |
+
))
|
166 |
+
|
167 |
+
user_inputs = [
|
168 |
+
text_prompt_component,
|
169 |
+
chatbot_component
|
170 |
+
]
|
171 |
+
|
172 |
+
bot_inputs = [
|
173 |
+
image_prompt_component,
|
174 |
+
temperature_component,
|
175 |
+
max_output_tokens_component,
|
176 |
+
stop_sequences_component,
|
177 |
+
top_k_component,
|
178 |
+
top_p_component,
|
179 |
+
chatbot_component
|
180 |
+
]
|
181 |
+
|
182 |
+
with gr.Blocks() as demo:
|
183 |
+
gr.HTML(TITLE)
|
184 |
+
with gr.Column():
|
185 |
+
with gr.Row():
|
186 |
+
image_prompt_component.render()
|
187 |
+
chatbot_component.render()
|
188 |
+
text_prompt_component.render()
|
189 |
+
run_button_component.render()
|
190 |
+
with gr.Accordion("Parameters", open=False):
|
191 |
+
temperature_component.render()
|
192 |
+
max_output_tokens_component.render()
|
193 |
+
stop_sequences_component.render()
|
194 |
+
with gr.Accordion("Advanced", open=False):
|
195 |
+
top_k_component.render()
|
196 |
+
top_p_component.render()
|
197 |
+
|
198 |
+
run_button_component.click(
|
199 |
+
fn=user,
|
200 |
+
inputs=user_inputs,
|
201 |
+
outputs=[text_prompt_component, chatbot_component],
|
202 |
+
queue=False
|
203 |
+
).then(
|
204 |
+
fn=bot, inputs=bot_inputs, outputs=[chatbot_component],
|
205 |
+
)
|
206 |
+
|
207 |
+
text_prompt_component.submit(
|
208 |
+
fn=user,
|
209 |
+
inputs=user_inputs,
|
210 |
+
outputs=[text_prompt_component, chatbot_component],
|
211 |
+
queue=False
|
212 |
+
).then(
|
213 |
+
fn=bot, inputs=bot_inputs, outputs=[chatbot_component],
|
214 |
+
)
|
215 |
+
|
216 |
+
demo.queue(max_size=99).launch(auth=(oaiusr, oaipwd),show_api=False, debug=False, show_error=True)
|