Spaces:
Runtime error
Runtime error
add app.py
Browse files
app.py
ADDED
@@ -0,0 +1,81 @@
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
1 |
+
import gradio as gr
|
2 |
+
import requests
|
3 |
+
|
4 |
+
import requests
|
5 |
+
import json
|
6 |
+
import os
|
7 |
+
|
8 |
+
APIKEY = os.environ.get("APIKEY")
|
9 |
+
APISECRET = os.environ.get("APISECRET")
|
10 |
+
|
11 |
+
def predict(text, seed, out_seq_length, min_gen_length, sampling_strategy,
|
12 |
+
num_beams, length_penalty, no_repeat_ngram_size,
|
13 |
+
temperature, topk, topp):
|
14 |
+
global APIKEY
|
15 |
+
global APISECRET
|
16 |
+
url = 'https://wudao.aminer.cn/os/api/api/v2/completions_130B'
|
17 |
+
|
18 |
+
payload = json.dumps({
|
19 |
+
"apikey": APIKEY,
|
20 |
+
"apisecret": APISECRET,
|
21 |
+
"language": "zh-CN",
|
22 |
+
"prompt": text,
|
23 |
+
"length_penalty": length_penalty,
|
24 |
+
"temperature": temperature,
|
25 |
+
"top_k": topk,
|
26 |
+
"top_p": topp,
|
27 |
+
"min_gen_length": min_gen_length,
|
28 |
+
"sampling_strategy": sampling_strategy,
|
29 |
+
"num_beams": num_beams,
|
30 |
+
"max_tokens": out_seq_length
|
31 |
+
})
|
32 |
+
|
33 |
+
headers = {
|
34 |
+
'Content-Type': 'application/json'
|
35 |
+
}
|
36 |
+
|
37 |
+
response = requests.request("POST", url, headers=headers, data=payload)
|
38 |
+
|
39 |
+
print(response.text)
|
40 |
+
return ret.text
|
41 |
+
|
42 |
+
|
43 |
+
if __name__ == "__main__":
|
44 |
+
with gr.Blocks() as demo:
|
45 |
+
gr.Markdown(
|
46 |
+
"""
|
47 |
+
# GLM-130B
|
48 |
+
An Open Bilingual Pre-Trained Model
|
49 |
+
""")
|
50 |
+
|
51 |
+
with gr.Row():
|
52 |
+
with gr.Column():
|
53 |
+
model_input = gr.Textbox(lines=7, placeholder='Input something in English or Chinese', label='Input')
|
54 |
+
with gr.Row():
|
55 |
+
gen = gr.Button("Generate")
|
56 |
+
clr = gr.Button("Clear")
|
57 |
+
outputs = gr.Textbox(lines=7, label='Output')
|
58 |
+
|
59 |
+
|
60 |
+
seed = gr.Slider(maximum=100000, value=1234, label='Seed')
|
61 |
+
out_seq_length = gr.Slider(maximum=256, value=128, minimum=8, label='Output Sequence Length')
|
62 |
+
min_gen_length = gr.Slider(maximum=64, value=0, label='Min Generate Length')
|
63 |
+
sampling_strategy = gr.Radio(choices=['BeamSearchStrategy', 'BaseStrategy'], value='BeamSearchStrategy', label='Search Strategy')
|
64 |
+
|
65 |
+
with gr.Tabs():
|
66 |
+
with gr.TabItem("Beam Search Parameter"):
|
67 |
+
# beam search
|
68 |
+
num_beams = gr.Slider(maximum=4, value=1, minimum=1, step=1, label='Number of Beams')
|
69 |
+
length_penalty = gr.Slider(maximum=1, value=0.8, minimum=0, label='Length Penalty')
|
70 |
+
no_repeat_ngram_size = gr.Slider(maximum=5, value=3, minimum=1, step=1, label='No Repeat Ngram Size')
|
71 |
+
with gr.TabItem("Base Search Parameter"):
|
72 |
+
# base search
|
73 |
+
temperature = gr.Slider(maximum=1, value=1, minimum=0, label='Temperature')
|
74 |
+
topk = gr.Slider(maximum=8, value=1, minimum=1, step=1, label='Top K')
|
75 |
+
topp = gr.Slider(maximum=8, value=0, minimum=0, step=1, label='Top P')
|
76 |
+
|
77 |
+
inputs = [model_input, seed, out_seq_length, min_gen_length, sampling_strategy, num_beams, length_penalty, no_repeat_ngram_size, temperature, topk, topp]
|
78 |
+
gen.click(fn=predict, inputs=inputs, outputs=outputs)
|
79 |
+
clr.click(fn=lambda value: gr.update(value=""), inputs=clr, outputs=model_input)
|
80 |
+
|
81 |
+
demo.launch()
|