Spaces:
Runtime error
Runtime error
File size: 9,006 Bytes
681ffab 4817bea 681ffab 53b2750 4817bea 53b2750 681ffab 45574e9 a3e22a8 681ffab 45574e9 681ffab |
1 2 3 4 5 6 7 8 9 10 11 12 13 14 15 16 17 18 19 20 21 22 23 24 25 26 27 28 29 30 31 32 33 34 35 36 37 38 39 40 41 42 43 44 45 46 47 48 49 50 51 52 53 54 55 56 57 58 59 60 61 62 63 64 65 66 67 68 69 70 71 72 73 74 75 76 77 78 79 80 81 82 83 84 85 86 87 88 89 90 91 92 93 94 95 96 97 98 99 100 101 102 103 104 105 106 107 108 109 110 111 112 113 114 115 116 117 118 119 120 121 122 123 124 125 126 127 128 129 130 131 132 133 134 135 136 137 138 139 140 141 142 143 144 145 146 147 148 149 150 151 152 153 154 155 156 157 158 159 160 161 162 163 164 165 166 167 168 169 170 171 172 173 174 175 176 177 178 179 180 181 182 183 184 185 186 187 188 189 190 191 192 193 194 195 196 197 198 199 200 201 202 203 204 205 206 207 208 209 210 211 212 213 214 215 216 217 218 219 220 221 222 223 224 225 226 227 228 229 230 231 232 233 234 235 236 |
import time
import logging
import gradio as gr
from src.llm_boilers import llm_boiler
logging.basicConfig(format="%(asctime)s - %(message)s", level=logging.INFO)
logging.warning("READY. App started...")
class Chat:
default_system_prompt = "A conversation between a user and an LLM-based AI assistant. The assistant gives helpful and honest answers."
system_format = "<|im_start|>system\n{}<|im_end|>\n"
def __init__(
self, system: str = None, user: str = None, assistant: str = None
) -> None:
if system is not None:
self.set_system_prompt(system)
else:
self.reset_system_prompt()
self.user = user if user else "<|im_start|>user\n{}<|im_end|>\n"
self.assistant = (
assistant if assistant else "<|im_start|>assistant\n{}<|im_end|>\n"
)
self.response_prefix = self.assistant.split("{}")[0]
def set_system_prompt(self, system_prompt):
# self.system = self.system_format.format(system_prompt)
return system_prompt
def reset_system_prompt(self):
return self.set_system_prompt(self.default_system_prompt)
def history_as_formatted_str(self, system, history) -> str:
system = self.system_format.format(system)
text = system + "".join(
[
"\n".join(
[
self.user.format(item[0]),
self.assistant.format(item[1]),
]
)
for item in history[:-1]
]
)
text += self.user.format(history[-1][0])
text += self.response_prefix
# stopgap solution to too long sequences
if len(text) > 4500:
# delete from the middle between <|im_start|> and <|im_end|>
# find the middle ones, then expand out
start = text.find("<|im_start|>", 139)
end = text.find("<|im_end|>", 139)
while end < len(text) and len(text) > 4500:
end = text.find("<|im_end|>", end + 1)
text = text[:start] + text[end + 1 :]
if len(text) > 4500:
# the nice way didn't work, just truncate
# deleting the beginning
text = text[-4500:]
return text
def clear_history(self, history):
return []
def turn(self, user_input: str):
self.user_turn(user_input)
return self.bot_turn()
def user_turn(self, user_input: str, history):
history.append([user_input, ""])
return user_input, history
def bot_turn(self, system, history, openai_key):
conversation = self.history_as_formatted_str(system, history)
assistant_response = call_inf_server(conversation, openai_key)
# history[-1][-1] = assistant_response
# return history
history[-1][1] = ""
for chunk in assistant_response:
try:
decoded_output = chunk["choices"][0]["delta"]["content"]
history[-1][1] += decoded_output
yield history
except KeyError:
pass
def call_inf_server(prompt, openai_key):
model_id = "gpt-3.5-turbo" # "gpt-3.5-turbo-16k",
model = llm_boiler(model_id, openai_key)
logging.warning(f'Inf via "{model_id}"" for prompt "{prompt}"')
try:
# run text generation
response = model.run(prompt, temperature=1.0)
logging.warning(f"Result of text generation: {response}")
return response
except Exception as e:
# assume it is our error
# just wait and try one more time
print(e)
time.sleep(2)
response = model.run(prompt, temperature=1.0)
logging.warning(f"Result of text generation: {response}")
return response
with gr.Blocks(
theme=gr.themes.glass(primary_hue="yellow",
secondary_hue=gr.themes.Color(secondary_100="#736c1c", secondary_200="#716823", secondary_300="#fde047", secondary_400="#facc15", secondary_50="#f2e821", secondary_500="#eab308", secondary_600="#ca8a04", secondary_700="#a16207", secondary_800="#854d0e", secondary_900="#713f12", secondary_950="#653b12"),
neutral_hue=gr.themes.Color(neutral_100="#2a1304", neutral_200="#e7e5e4", neutral_300="#d6d3d1", neutral_400="#a8a29e", neutral_50="#97ae29", neutral_500="#78716c", neutral_600="#57534e", neutral_700="#44403c", neutral_800="#292524", neutral_900="#1c1917", neutral_950="#0f0e0d"),
text_size="text_lg",
spacing_size="spacing_md",
radius_size=gr.themes.Size(radius_lg="8px", radius_md="6px", radius_sm="4px", radius_xl="12px", radius_xs="2px", radius_xxl="22px", radius_xxs="5px"),
font=[gr.themes.GoogleFont('Optima'), gr.themes.GoogleFont('Candara'), gr.themes.GoogleFont('Noto Sans'), gr.themes.GoogleFont('source-sans-pro')],
font_mono=[gr.themes.GoogleFont('IBM Plex Mono'), gr.themes.GoogleFont('ui-monospace'), gr.themes.GoogleFont('Consolas'), gr.themes.GoogleFont('monospace')],
).set(
body_background_fill='*primary_500',
body_background_fill_dark='*primary_50',
background_fill_primary='*primary_50',
background_fill_primary_dark='*primary_50'
)),
css=".disclaimer {font-variant-caps: all-small-caps;}",
) as demo:
gr.Markdown(
"""<h1><center>Chat with gpt-3.5-turbo</center></h1>
This is a lightweight demo of gpt-3.5-turbo conversation completion. It was designed as a template for in-context learning applications to be built on top of.
"""
)
conversation = Chat()
with gr.Row():
with gr.Column():
# to do: change to openaikey input for public release
openai_key = gr.Textbox(
input="OpenAI Key",
value="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",
type="password",
placeholder="Free use",
info="You have to provide your own OpenAI API key.",
)
chatbot = gr.Chatbot().style(height=400)
with gr.Row():
with gr.Column():
msg = gr.Textbox(
label="Chat Message Box",
placeholder="Chat Message Box",
show_label=False,
).style(container=False)
with gr.Column():
with gr.Row():
submit = gr.Button("Submit")
stop = gr.Button("Stop")
clear = gr.Button("Clear")
with gr.Row():
with gr.Accordion("Advanced Options:", open=False):
with gr.Row():
with gr.Column(scale=2):
system = gr.Textbox(
label="System Prompt",
value=Chat.default_system_prompt,
show_label=False,
).style(container=False)
with gr.Column():
with gr.Row():
change = gr.Button("Change System Prompt")
reset = gr.Button("Reset System Prompt")
with gr.Row():
gr.Markdown(
"Disclaimer: The gpt-3.5-turbo model can produce factually incorrect output, and should not be solely relied on to produce "
"factually accurate information. The gpt-3.5-turbo model was trained on various public datasets; while great efforts "
"have been taken to clean the pretraining data, it is possible that this model could generate lewd, "
"biased, or otherwise offensive outputs.",
elem_classes=["disclaimer"],
)
with gr.Row():
gr.Markdown(
"[Privacy policy](https://gist.github.com/samhavens/c29c68cdcd420a9aa0202d0839876dac)",
elem_classes=["disclaimer"],
)
submit_event = msg.submit(
fn=conversation.user_turn,
inputs=[msg, chatbot],
outputs=[msg, chatbot],
queue=False,
).then(
fn=conversation.bot_turn,
inputs=[system, chatbot, openai_key],
outputs=[chatbot],
queue=True,
)
submit_click_event = submit.click(
fn=conversation.user_turn,
inputs=[msg, chatbot],
outputs=[msg, chatbot],
queue=False,
).then(
fn=conversation.bot_turn,
inputs=[system, chatbot, openai_key],
outputs=[chatbot],
queue=True,
)
stop.click(
fn=None,
inputs=None,
outputs=None,
cancels=[submit_event, submit_click_event],
queue=False,
)
clear.click(lambda: None, None, chatbot, queue=False).then(
fn=conversation.clear_history,
inputs=[chatbot],
outputs=[chatbot],
queue=False,
)
change.click(
fn=conversation.set_system_prompt,
inputs=[system],
outputs=[system],
queue=False,
)
reset.click(
fn=conversation.reset_system_prompt,
inputs=[],
outputs=[system],
queue=False,
)
demo.queue(max_size=36, concurrency_count=14).launch(debug=True)
|