File size: 8,183 Bytes
7b361da 0f877eb 7b361da 0f877eb 7b361da |
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 99 100 101 102 103 104 105 106 107 108 109 110 111 112 113 114 115 116 117 118 119 120 121 122 123 124 125 126 127 128 129 130 131 132 133 134 135 136 137 138 139 140 141 142 143 144 145 146 147 148 149 150 151 152 153 154 155 156 157 158 159 160 161 162 163 164 165 166 167 168 169 170 171 172 173 174 175 176 177 178 179 180 181 182 183 184 185 186 187 188 189 190 191 192 193 194 195 196 197 198 199 200 201 202 203 204 205 206 207 208 209 210 211 212 213 214 215 216 217 218 219 220 221 222 223 224 225 226 227 228 229 230 231 232 233 234 235 236 237 238 239 240 241 242 243 244 245 246 247 248 249 250 251 252 253 254 255 256 257 258 259 260 261 262 263 264 265 266 267 268 269 270 |
# -*- coding:utf-8 -*-
import os
import logging
import gradio as gr
import gc
from interface.hddr_llama_onnx_interface import LlamaOnnxInterface
from interface.empty_stub_interface import EmptyStubInterface
from ChatApp.app_modules.utils import (
reset_textbox,
transfer_input,
reset_state,
delete_last_conversation,
cancel_outputing,
)
from ChatApp.app_modules.presets import (
small_and_beautiful_theme,
title,
description_top,
description,
)
from ChatApp.app_modules.overwrites import postprocess
logging.basicConfig(
level=logging.DEBUG,
format="%(asctime)s [%(levelname)s] [%(filename)s:%(lineno)d] %(message)s",
)
# we can filter this dictionary at the start according to the actual available files on disk
empty_stub_model_name = "_Empty Stub_"
top_directory = os.path.dirname(os.path.dirname(os.path.realpath(__file__)))
tokenizer_path = os.path.join(top_directory, "tokenizer.model")
available_models = {
"Llama-2 Chat 13B Float16": {
"onnx_file": os.path.join(
top_directory, "FP16-Chat", "LlamaV2_13B_FT_float32.onnx"
),
"tokenizer_path": tokenizer_path,
"embedding_file": os.path.join(top_directory, "embeddings-chat.pth"),
},
"Llama-2 Chat 13B Float32": {
"onnx_file": os.path.join(
top_directory, "FP32-Chat", "LlamaV2_13B_FT_float32.onnx"
),
"tokenizer_path": tokenizer_path,
"embedding_file": os.path.join(
top_directory, "embeddings-chat.pth"
),
},
"Llama-2 13B Float16": {
"onnx_file": os.path.join(
top_directory, "FP16", "LlamaV2_13B_float16.onnx"
),
"tokenizer_path": tokenizer_path,
"embedding_file": os.path.join(top_directory, "embeddings.pth"),
},
"Llama-2 13B Float32": {
"onnx_file": os.path.join(
top_directory, "FP32", "LlamaV2_13B_float16.onnx"
),
"tokenizer_path": tokenizer_path,
"embedding_file": os.path.join(
top_directory, "embeddings.pth"
),
},
}
interface = EmptyStubInterface()
interface.initialize()
# interface = None
gr.Chatbot.postprocess = postprocess
with open("ChatApp/assets/custom.css", "r", encoding="utf-8") as f:
custom_css = f.read()
def change_model_listener(new_model_name):
if new_model_name is None:
new_model_name = empty_stub_model_name
global interface
# if a model exists - shut it down before trying to create the new one
if interface is not None:
interface.shutdown()
del interface
gc.collect()
logging.info(f"Creating a new model [{new_model_name}]")
if new_model_name == empty_stub_model_name:
interface = EmptyStubInterface()
interface.initialize()
else:
d = available_models[new_model_name]
interface = LlamaOnnxInterface(
onnx_file=d["onnx_file"],
tokenizer_path=d["tokenizer_path"],
embedding_file=d["embedding_file"],
)
interface.initialize()
return new_model_name
def interface_predict(*args):
global interface
res = interface.predict(*args)
for x in res:
yield x
def interface_retry(*args):
global interface
res = interface.retry(*args)
for x in res:
yield x
with gr.Blocks(css=custom_css, theme=small_and_beautiful_theme) as demo:
history = gr.State([])
user_question = gr.State("")
with gr.Row():
gr.HTML(title)
status_display = gr.Markdown("Success", elem_id="status_display")
gr.Markdown(description_top)
with gr.Row():
with gr.Column(scale=5):
with gr.Row():
chatbot = gr.Chatbot(elem_id="chuanhu_chatbot", height=900)
with gr.Row():
with gr.Column(scale=12):
user_input = gr.Textbox(show_label=False, placeholder="Enter text")
with gr.Column(min_width=70, scale=1):
submit_button = gr.Button("Send")
with gr.Column(min_width=70, scale=1):
cancel_button = gr.Button("Stop")
with gr.Row():
empty_button = gr.Button(
"🧹 New Conversation",
)
retry_button = gr.Button("🔄 Regenerate")
delete_last_button = gr.Button("🗑️ Remove Last Turn")
with gr.Column():
with gr.Column(min_width=50, scale=1):
with gr.Tab(label="Parameter Setting"):
gr.Markdown("# Model")
model_name = gr.Dropdown(
choices=[empty_stub_model_name] + list(available_models.keys()),
label="Model",
show_label=False, # default="Empty STUB",
)
model_name.change(
change_model_listener, inputs=[model_name], outputs=[model_name]
)
gr.Markdown("# Parameters")
top_p = gr.Slider(
minimum=-0,
maximum=1.0,
value=0.9,
step=0.05,
interactive=True,
label="Top-p",
)
temperature = gr.Slider(
minimum=0.1,
maximum=2.0,
value=0.75,
step=0.1,
interactive=True,
label="Temperature",
)
max_length_tokens = gr.Slider(
minimum=0,
maximum=512,
value=256,
step=8,
interactive=True,
label="Max Generation Tokens",
)
max_context_length_tokens = gr.Slider(
minimum=0,
maximum=4096,
value=2048,
step=128,
interactive=True,
label="Max History Tokens",
)
gr.Markdown(description)
predict_args = dict(
# fn=interface.predict,
fn=interface_predict,
inputs=[
user_question,
chatbot,
history,
top_p,
temperature,
max_length_tokens,
max_context_length_tokens,
],
outputs=[chatbot, history, status_display],
show_progress=True,
)
retry_args = dict(
fn=interface_retry,
inputs=[
user_input,
chatbot,
history,
top_p,
temperature,
max_length_tokens,
max_context_length_tokens,
],
outputs=[chatbot, history, status_display],
show_progress=True,
)
reset_args = dict(fn=reset_textbox, inputs=[], outputs=[user_input, status_display])
# Chatbot
transfer_input_args = dict(
fn=transfer_input,
inputs=[user_input],
outputs=[user_question, user_input, submit_button],
show_progress=True,
)
predict_event1 = user_input.submit(**transfer_input_args).then(**predict_args)
predict_event2 = submit_button.click(**transfer_input_args).then(**predict_args)
empty_button.click(
reset_state,
outputs=[chatbot, history, status_display],
show_progress=True,
)
empty_button.click(**reset_args)
predict_event3 = retry_button.click(**retry_args)
delete_last_button.click(
delete_last_conversation,
[chatbot, history],
[chatbot, history, status_display],
show_progress=True,
)
cancel_button.click(
cancel_outputing,
[],
[status_display],
cancels=[predict_event1, predict_event2, predict_event3],
)
demo.load(change_model_listener, inputs=None, outputs=model_name)
demo.title = "Llama-2 Chat UI"
demo.queue(concurrency_count=1).launch()
|