-
Notifications
You must be signed in to change notification settings - Fork 16
/
Copy pathMagicPrompt.py
190 lines (161 loc) · 7.38 KB
/
MagicPrompt.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
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
#!/usr/bin/env python
# coding: utf-8
import os
import html
import torch
import gradio as gr
from transformers import GPT2LMHeadModel, GPT2Tokenizer, AutoTokenizer, AutoModelForCausalLM
from omegaconf import OmegaConf
conf = OmegaConf.load('models.yaml')
base_dir = os.path.dirname(os.path.realpath(__file__))
models_dir = os.path.join(base_dir, conf.models_dir)
class Model:
name = None
model = None
tokenizer = None
models = {}
current = Model()
def model_selection_changed(model_name):
if model_name == "None":
current.tokenizer = None
current.model = None
current.name = None
else:
current.tokenizer = models[model_name].tokenizer
current.model = models[model_name].model
current.name = models[model_name].name
def loading_models():
for item in conf.models:
_model = Model()
_model.name = item["Name"]
_model.model = item["Model"]
_model.tokenizer = item["Tokenizer"]
models[_model.name] = _model
model_selection_changed(list(models.keys())[0])
def get_model_path(name):
dirname = os.path.join(models_dir, name)
if not os.path.isdir(dirname):
return name
return dirname
def check_model():
model_name = current.name
path = get_model_path(model_name)
model_chk = os.path.exists(models_dir+model_name+'/pytorch_model.bin')
if model_chk is False:
print('model not found, start cloning from Hugging Face')
dir_chk = os.path.exists(models_dir)
if dir_chk is False:
os.makedirs(models_dir)
if model_name != 'None':
path = os.path.join(models_dir, model_name)
dir_chk = os.path.exists(path)
if dir_chk is False:
os.makedirs(path)
try:
tokenizer_dl = GPT2Tokenizer.from_pretrained(current.tokenizer)
model_dl = GPT2LMHeadModel.from_pretrained(current.model)
tokenizer_dl.save_pretrained(path)
model_dl.save_pretrained(path)
print('model cloned from Hugging Face')
except Exception as e:
print(
f"Exception encountered while attempting to install tokenizer: {e}")
return gr.update(), f"Error: {e}"
def generate(model, prompt, temperature, top_k, min_length, max_length, repetition_penalty, num_return_sequences):
try:
if current.name != model:
current.tokenizer = None
current.model = None
current.name = None
if model != 'None':
check_model()
path = get_model_path(model)
current.tokenizer = AutoTokenizer.from_pretrained(
models[model].tokenizer)
current.model = AutoModelForCausalLM.from_pretrained(
models[model].model)
current.name = current.name
assert current.model, 'No model available'
assert current.tokenizer, 'No tokenizer available'
except Exception as e:
print(f"Exception: {e}")
try:
print(f"Generate new prompt from: \"{prompt}\" with {current.name}")
input_ids = current.tokenizer(prompt, return_tensors='pt').input_ids
if input_ids.shape[1] == 0:
input_ids = torch.asarray(
[[current.tokenizer.bos_token_id]], dtype=torch.long)
output = current.model.generate(input_ids,
do_sample=True,
temperature=max(
float(temperature), 1e-6),
top_k=round(top_k),
max_length=max_length,
num_return_sequences=num_return_sequences,
repetition_penalty=float(
repetition_penalty),
pad_token_id=current.tokenizer.pad_token_id or current.tokenizer.eos_token_id
)
print("Generation complete!")
texts = current.tokenizer.batch_decode(
output, skip_special_tokens=True)
index = 0
markup = ''
for generated_text in texts:
index += 1
markup += f"""
<div class="box" style="margin-bottom: var(--size-3);border: 1px solid var(--color-border-primary);border-radius: var(--radius-lg);background: var(--color-background-tertiary);color: var(--color-text-body);">
<p id='prompt_res_{index}' style="font-size:var(--scale-0);padding:var(--size-2-5) var(--size-3)">{html.escape(generated_text)}</p>
</div>
"""
return markup
except Exception as e:
print(
f"Exception encountered while attempting to generate prompt: {e}")
return gr.update(), f"Error: {e}"
with gr.Blocks(analytics_enabled=0, title="MagicPrompt Generator") as magicprompt:
gr.HTML("<h1 style='text-align:center;'>MagicPrompt Generator</h1>")
with gr.Tab("Generator"):
with gr.Row():
with gr.Column(scale=80):
text_input = gr.Textbox(
lines=2, show_label=False, value="", placeholder="Enter your prompt...")
with gr.Column(scale=10):
submit = gr.Button('Generate', variant='primary')
with gr.Row():
with gr.Column():
with gr.Row():
temp_slider = gr.Slider(
elem_id="temp_slider", label="Temperature", interactive=True, minimum=0, maximum=4, value=1)
top_k_slider = gr.Slider(
elem_id="top_k_slider", label="Top K", value=12, minimum=1, maximum=50, step=1, interactive=True)
repetition_penalty = gr.Slider(
label="Repetition penalty", elem_id="repetition_penalty", value=1, minimum=1, maximum=4, step=0.01)
with gr.Row():
min_length_slider = gr.Slider(
elem_id="min_length_slider", label="Min Length", interactive=True, minimum=1, maximum=400, step=1, value=20)
max_length_slider = gr.Slider(
elem_id="max_length_slider", label="Max Length", interactive=True, minimum=1, maximum=400, step=1, value=90)
num_return_sequences_slider = gr.Slider(
elem_id="num_return_sequences_slider", label="How Many To Generate", value=5, minimum=1, maximum=20, interactive=True, step=1)
with gr.Row():
loading_models()
models_list = list(models.keys())
model_selection = gr.Dropdown(
label="Model", elem_id="prompt_model", interactive=True, value=models_list[0], choices=["None"] + models_list)
with gr.Column():
with gr.Row():
res = gr.HTML()
submit.click(
fn=generate,
inputs=[model_selection, text_input, temp_slider, top_k_slider, min_length_slider,
max_length_slider, repetition_penalty, num_return_sequences_slider],
outputs=[res]
)
model_selection.change(
fn=model_selection_changed,
inputs=[model_selection],
outputs=[],
)
if __name__ == "__main__":
magicprompt.queue(concurrency_count=20).launch(server_name="0.0.0.0", server_port=8090, show_api=False, debug=True)