Spaces:
Runtime error
Runtime error
Commit ·
b849606
1
Parent(s): 401530d
Update app.py
Browse files
app.py
CHANGED
|
@@ -1,10 +1,17 @@
|
|
| 1 |
-
from transformers import MT5ForConditionalGeneration, AutoTokenizer, Text2TextGenerationPipeline
|
| 2 |
import gradio as gr
|
|
|
|
| 3 |
|
|
|
|
| 4 |
trans_mdl = MT5ForConditionalGeneration.from_pretrained("K024/mt5-zh-ja-en-trimmed")
|
| 5 |
trans_tokenizer = AutoTokenizer.from_pretrained("K024/mt5-zh-ja-en-trimmed")
|
| 6 |
trans_pipe = Text2TextGenerationPipeline(model=trans_mdl, tokenizer=trans_tokenizer)
|
| 7 |
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
| 8 |
def translation_job(job, text):
|
| 9 |
# 设置翻译任务和提示语的映射
|
| 10 |
job_key = ["中译日", "中译英", "日译中", "英译中", "日译英", "英译日"]
|
|
@@ -15,7 +22,34 @@ def translation_job(job, text):
|
|
| 15 |
print(input)
|
| 16 |
response = trans_pipe(input, max_length=100, num_beams=4)
|
| 17 |
return response[0]['generated_text']
|
| 18 |
-
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
| 19 |
|
| 20 |
|
| 21 |
with gr.Blocks() as app:
|
|
@@ -29,8 +63,16 @@ with gr.Blocks() as app:
|
|
| 29 |
source_text = gr.Textbox(lines=1, label="翻译文本", placeholder="请输入要翻译的文本")
|
| 30 |
trans_result = gr.Textbox(lines=1, label="翻译结果")
|
| 31 |
trans_btn = gr.Button("翻译")
|
| 32 |
-
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
| 33 |
trans_btn.click(translation_job, inputs=[job_name, source_text], outputs=trans_result)
|
|
|
|
|
|
|
| 34 |
app.launch()
|
| 35 |
|
| 36 |
|
|
|
|
| 1 |
+
from transformers import MT5ForConditionalGeneration, AutoTokenizer, Text2TextGenerationPipeline, AutoModelForSeq2SeqLM
|
| 2 |
import gradio as gr
|
| 3 |
+
import re
|
| 4 |
|
| 5 |
+
# 翻译任务设置
|
| 6 |
trans_mdl = MT5ForConditionalGeneration.from_pretrained("K024/mt5-zh-ja-en-trimmed")
|
| 7 |
trans_tokenizer = AutoTokenizer.from_pretrained("K024/mt5-zh-ja-en-trimmed")
|
| 8 |
trans_pipe = Text2TextGenerationPipeline(model=trans_mdl, tokenizer=trans_tokenizer)
|
| 9 |
|
| 10 |
+
# 摘要任务设置
|
| 11 |
+
sum_mdl = AutoModelForSeq2SeqLM.from_pretrained("csebuetnlp/mT5_multilingual_XLSum")
|
| 12 |
+
sum_tokenizer = AutoTokenizer.from_pretrained("csebuetnlp/mT5_multilingual_XLSum")
|
| 13 |
+
|
| 14 |
+
|
| 15 |
def translation_job(job, text):
|
| 16 |
# 设置翻译任务和提示语的映射
|
| 17 |
job_key = ["中译日", "中译英", "日译中", "英译中", "日译英", "英译日"]
|
|
|
|
| 22 |
print(input)
|
| 23 |
response = trans_pipe(input, max_length=100, num_beams=4)
|
| 24 |
return response[0]['generated_text']
|
| 25 |
+
|
| 26 |
+
|
| 27 |
+
def sum_job(text):
|
| 28 |
+
# 去除源文本中的空格
|
| 29 |
+
WHITESPACE_HANDLER = lambda k: re.sub('\s+', ' ', re.sub('\n+', ' ', k.strip()))
|
| 30 |
+
|
| 31 |
+
input_ids = sum_tokenizer(
|
| 32 |
+
[WHITESPACE_HANDLER(text)],
|
| 33 |
+
return_tensors="pt",
|
| 34 |
+
padding="max_length",
|
| 35 |
+
truncation=True,
|
| 36 |
+
max_length=512
|
| 37 |
+
)["input_ids"]
|
| 38 |
+
|
| 39 |
+
output_ids = sum_mdl.generate(
|
| 40 |
+
input_ids=input_ids,
|
| 41 |
+
max_length=84,
|
| 42 |
+
no_repeat_ngram_size=2,
|
| 43 |
+
num_beams=4
|
| 44 |
+
)[0]
|
| 45 |
+
|
| 46 |
+
response = sum_tokenizer.decode(
|
| 47 |
+
output_ids,
|
| 48 |
+
skip_special_tokens=True,
|
| 49 |
+
clean_up_tokenization_spaces=False
|
| 50 |
+
)
|
| 51 |
+
|
| 52 |
+
return response
|
| 53 |
|
| 54 |
|
| 55 |
with gr.Blocks() as app:
|
|
|
|
| 63 |
source_text = gr.Textbox(lines=1, label="翻译文本", placeholder="请输入要翻译的文本")
|
| 64 |
trans_result = gr.Textbox(lines=1, label="翻译结果")
|
| 65 |
trans_btn = gr.Button("翻译")
|
| 66 |
+
|
| 67 |
+
# 多语言自动摘要任务
|
| 68 |
+
with gr.Tab("多语言自动摘要"):
|
| 69 |
+
article_text = gr.Textbox(lines=8, label="待总结文本", placeholder="请输入要进行摘要的文本")
|
| 70 |
+
sum_result = gr.Textbox(lines=2, label="摘要结果")
|
| 71 |
+
sum_btn = gr.Button("摘要")
|
| 72 |
+
|
| 73 |
trans_btn.click(translation_job, inputs=[job_name, source_text], outputs=trans_result)
|
| 74 |
+
sum_btn.click(sum_job, inputs=article_text, outputs=sum_result)
|
| 75 |
+
|
| 76 |
app.launch()
|
| 77 |
|
| 78 |
|