forked from 768365110/train_custom_LLM
-
Notifications
You must be signed in to change notification settings - Fork 1
/
webui.py
150 lines (127 loc) · 6.25 KB
/
webui.py
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
import os
import torch
from transformers import AutoModel, AutoTokenizer
import gradio as gr
import mdtex2html
from model import load_model
from transformers import GenerationConfig
# from utils import load_model_on_gpus
"""Override Chatbot.postprocess"""
def postprocess(self, y):
if y is None:
return []
for i, (message, response) in enumerate(y):
y[i] = (
None if message is None else mdtex2html.convert((message)),
None if response is None else mdtex2html.convert(response),
)
return y
def parse_text(text):
"""copy from https://github.com/GaiZhenbiao/ChuanhuChatGPT/"""
lines = text.split("\n")
lines = [line for line in lines if line != ""]
count = 0
for i, line in enumerate(lines):
if "```" in line:
count += 1
items = line.split('`')
if count % 2 == 1:
lines[i] = f'<pre><code class="language-{items[-1]}">'
else:
lines[i] = f'<br></code></pre>'
else:
if i > 0:
if count % 2 == 1:
line = line.replace("`", "\`")
line = line.replace("<", "<")
line = line.replace(">", ">")
line = line.replace(" ", " ")
line = line.replace("*", "*")
line = line.replace("_", "_")
line = line.replace("-", "-")
line = line.replace(".", ".")
line = line.replace("!", "!")
line = line.replace("(", "(")
line = line.replace(")", ")")
line = line.replace("$", "$")
lines[i] = "<br>"+line
text = "".join(lines)
return text
def stream_predict(input, chatbot, max_length, top_p, temperature, history, past_key_values):
chatbot.append((parse_text(input), ""))
if len(history) > 2 * MAX_TURNS:
history = history[-MAX_TURNS:]
for response, history, past_key_values in model.stream_chat(tokenizer, input, history, past_key_values=past_key_values,
return_past_key_values=True,
max_length=max_length, top_p=top_p,
temperature=temperature):
chatbot[-1] = (parse_text(input), parse_text(response))
yield chatbot, history, past_key_values
def predict(input_text, chatbot, max_length, top_p, temperature, history, past_key_values):
chatbot.append((parse_text(input_text), ""))
prompt = "Human: " + input_text + "\n\nAssistant: "
inputs = tokenizer(prompt, return_tensors="pt").to(model.device)
generation_config = GenerationConfig(
temperature=temperature,
top_p=top_p,
do_sample=True,
repetition_penalty=2.0,
max_new_tokens=max_length, # max_length=max_new_tokens+input_sequence
)
generate_ids = model.generate(**inputs, generation_config=generation_config)
output = tokenizer.decode(generate_ids[0][len(inputs.input_ids[0]):])
chatbot[-1] = (parse_text(input_text), parse_text(output))
return chatbot, None, None
def reset_user_input():
return gr.update(value='')
def reset_state():
return [], [], None
def parse_arg():
import argparse
parser = argparse.ArgumentParser()
parser.add_argument("--model_name", help="Model name optional [baichuan, chatGLM]")
parser.add_argument("--model_path", help="model checkpoint folder")
parser.add_argument("--lora_path", default=None, help="lora checkpoint folder")
parser.add_argument("--max_turns", default=20, help="max multi-rounds chat turns")
parser.add_argument("--quantize", default='4bit', help="quantization config optional [None, 4bit, 8bit]")
args = parser.parse_args()
return args
if __name__ == "__main__":
args = parse_arg()
MAX_TURNS = args.max_turns # int参数无法传入，在外部全局定义
name = args.model_name + (" lora" if args.lora_path else "") + (f" quantize {args.quantize}" if args.quantize else "")
model, tokenizer = load_model(args.model_name, args.model_path, args.quantize, torch.cuda.current_device())
if args.lora_path:
print("load lora weight")
from peft import PeftModel
model = PeftModel.from_pretrained(model, args.lora_path)
model = model.eval()
gr.Chatbot.postprocess = postprocess
with gr.Blocks() as demo:
gr.HTML(f"""<h1 align="center">{name}</h1>""")
chatbot = gr.Chatbot(scale=8)
with gr.Row():
with gr.Column(scale=4):
with gr.Column(scale=12):
user_input = gr.Textbox(show_label=False, placeholder="Input...", lines=10).style(
container=False)
with gr.Column(min_width=32, scale=1):
submitBtn = gr.Button("Submit", variant="primary")
with gr.Column(scale=1):
emptyBtn = gr.Button("Clear History")
max_length = gr.Slider(0, 32768, value=8192, step=1.0, label="Maximum length", interactive=True)
top_p = gr.Slider(0, 1, value=0.8, step=0.01, label="Top P", interactive=True)
temperature = gr.Slider(0, 1, value=0.95, step=0.01, label="Temperature", interactive=True)
history = gr.State([])
past_key_values = gr.State(None)
if "chatglm" in args.model_name:
submitBtn.click(stream_predict, [user_input, chatbot, max_length, top_p, temperature, history, past_key_values],
[chatbot, history, past_key_values], show_progress=True)
elif "baichuan" in args.model_name:
submitBtn.click(predict,
[user_input, chatbot, max_length, top_p, temperature, history, past_key_values],
[chatbot, history, past_key_values], show_progress=True)
submitBtn.click(reset_user_input, [], [user_input])
emptyBtn.click(reset_state, outputs=[chatbot, history, past_key_values], show_progress=True)
demo.queue().launch(share=True, inbrowser=True)
# CUDA_VISIBLE_DEVICES=0 python.py webui.py --model baichuan --model_ckpt --lora_ckpt --quantize 4bit