Spaces:
Running
Running
Add application file
Browse files
app.py
ADDED
@@ -0,0 +1,90 @@
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
1 |
+
import gradio as gr
|
2 |
+
from transformers import AutoTokenizer, AutoModelWithLMHead, AutoModelForSeq2SeqLM
|
3 |
+
import torch
|
4 |
+
|
5 |
+
|
6 |
+
max_input_length = 128
|
7 |
+
max_target_length = 128
|
8 |
+
|
9 |
+
device = 'cuda' if torch.cuda.is_available() else 'cpu'
|
10 |
+
print(f'Using {device} device')
|
11 |
+
|
12 |
+
# 1是中文到满语
|
13 |
+
model1_checkpoint = "Helsinki-NLP/opus-mt-zh-en"
|
14 |
+
tokenizer1 = AutoTokenizer.from_pretrained(model1_checkpoint)
|
15 |
+
model1 = AutoModelForSeq2SeqLM.from_pretrained(model1_checkpoint)
|
16 |
+
model1 = model1.to(device)
|
17 |
+
|
18 |
+
model1.load_state_dict(torch.load('../model/chineseToManju/epoch_41_valid_bleu_100.00_model_weights.bin'))
|
19 |
+
model1.eval()
|
20 |
+
# 2是满语到中文
|
21 |
+
model2_checkpoint = "Helsinki-NLP/opus-mt-en-zh"
|
22 |
+
tokenizer2 = AutoTokenizer.from_pretrained(model2_checkpoint)
|
23 |
+
model2 = AutoModelForSeq2SeqLM.from_pretrained(model2_checkpoint)
|
24 |
+
model2 = model2.to(device)
|
25 |
+
|
26 |
+
model2.load_state_dict(torch.load('../model/manjuToChinese/epoch_41_valid_bleu_0.00_model_weights.bin'))
|
27 |
+
model2.eval()
|
28 |
+
|
29 |
+
def chineseToManju(text):
|
30 |
+
batch_data = tokenizer1(
|
31 |
+
text,
|
32 |
+
padding=True,
|
33 |
+
max_length=max_input_length,
|
34 |
+
truncation=True,
|
35 |
+
return_tensors="pt"
|
36 |
+
)
|
37 |
+
generated_tokens = model1.generate(
|
38 |
+
batch_data["input_ids"],
|
39 |
+
attention_mask=batch_data["attention_mask"],
|
40 |
+
max_length=max_target_length,
|
41 |
+
).cpu().numpy()
|
42 |
+
res = tokenizer1.batch_decode(generated_tokens, skip_special_tokens=True)
|
43 |
+
return res
|
44 |
+
|
45 |
+
def manjuToChinese(text):
|
46 |
+
batch_data = tokenizer2(
|
47 |
+
text,
|
48 |
+
padding=True,
|
49 |
+
max_length=max_input_length,
|
50 |
+
truncation=True,
|
51 |
+
return_tensors="pt"
|
52 |
+
)
|
53 |
+
generated_tokens = model2.generate(
|
54 |
+
batch_data["input_ids"],
|
55 |
+
attention_mask=batch_data["attention_mask"],
|
56 |
+
max_length=max_target_length,
|
57 |
+
).cpu().numpy()
|
58 |
+
res = tokenizer2.batch_decode(generated_tokens, skip_special_tokens=True)
|
59 |
+
return res
|
60 |
+
|
61 |
+
with gr.Blocks() as demo:
|
62 |
+
#用markdown语法编辑输出一段话
|
63 |
+
gr.Markdown("## 满语翻译演示")
|
64 |
+
# 设置tab选项卡
|
65 |
+
with gr.Tab("满to中"):
|
66 |
+
#Blocks特有组件,设置所有子组件按垂直排列
|
67 |
+
#垂直排列是默认情况,不加也没关系
|
68 |
+
with gr.Column():
|
69 |
+
text_input1 = gr.Textbox(lines=2, placeholder="请输入满语",label="manju")
|
70 |
+
text_button1 = gr.Button("翻译")
|
71 |
+
text_output1 = gr.Textbox(lines=2, label="chinese")
|
72 |
+
with gr.Tab("中to满"):
|
73 |
+
#Blocks特有组件,设置所有子组件按水平排列
|
74 |
+
with gr.Column():
|
75 |
+
text_input2 = gr.Textbox(lines=2, placeholder="请输入中文",label="chinese")
|
76 |
+
text_button2 = gr.Button("翻译")
|
77 |
+
text_output2 = gr.Textbox(lines=2, label="manju")
|
78 |
+
#设置折叠内容
|
79 |
+
with gr.Accordion(""):
|
80 |
+
gr.Markdown("## 东北师范大学信息科学与技术学院 满语智能处理实验室")
|
81 |
+
gr.Markdown("#### 注意事项")
|
82 |
+
gr.Markdown("最长语句不能超过128个词!")
|
83 |
+
gr.Markdown("#### 以下是一些例子")
|
84 |
+
gr.Markdown('"manju": "sakda amji,be yabume oho.", "chinese": "大爷,我们该走了。"')
|
85 |
+
gr.Markdown('"manju": "ume ekxere,jai emu majige teki,majige muke be omiki.", "chinese": "忙什么呀,再坐一会儿,喝点水。"')
|
86 |
+
gr.Markdown('"manju": "omirakv oho,ubade emgeri hontoha inenggi tehe,suwembe ambula jobohuha.", "chinese": "不了,在这呆了有半天了,打扰你们了。"')
|
87 |
+
gr.Markdown('"manju": "ume yabure,ubade emu erin buda be jejfi jai yabuki.", "chinese": "别走了,在这吃了饭再走吧。"')
|
88 |
+
text_button1.click(manjuToChinese, inputs=text_input1, outputs=text_output1)
|
89 |
+
text_button2.click(chineseToManju, inputs=text_input2, outputs=text_output2)
|
90 |
+
demo.launch(server_name="0.0.0.0", server_port=1234, share=True)
|