Spaces:
Runtime error
Runtime error
Update app.py
Browse files
app.py
CHANGED
@@ -4,6 +4,7 @@ import gradio as gr
|
|
4 |
import time
|
5 |
from transformers import AutoTokenizer, AutoModelForSeq2SeqLM, pipeline
|
6 |
from flores200_codes import flores_codes
|
|
|
7 |
|
8 |
def load_models():
|
9 |
# build model and tokenizer
|
@@ -24,25 +25,38 @@ def load_models():
|
|
24 |
|
25 |
return model_dict
|
26 |
|
27 |
-
|
28 |
def translation(source, target, text):
|
29 |
-
if len(model_dict) == 2:
|
30 |
-
#model_name = 'nllb-distilled-600M'
|
31 |
-
model_name = 'nllb-distilled-1.3B'
|
32 |
-
|
33 |
-
start_time = time.time()
|
34 |
source = flores_codes[source]
|
35 |
target = flores_codes[target]
|
36 |
-
|
37 |
-
model = model_dict[model_name + '_model']
|
38 |
-
tokenizer = model_dict[model_name + '_tokenizer']
|
39 |
-
|
40 |
-
translator = pipeline('translation', model=model, tokenizer=tokenizer, src_lang=source, tgt_lang=target)
|
41 |
-
chunks = text.splitlines(True)
|
42 |
output = ""
|
43 |
-
|
44 |
-
|
45 |
-
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
46 |
|
47 |
end_time = time.time()
|
48 |
|
@@ -52,7 +66,6 @@ def translation(source, target, text):
|
|
52 |
'result': output}
|
53 |
return result
|
54 |
|
55 |
-
|
56 |
if __name__ == '__main__':
|
57 |
print('\tinit models')
|
58 |
|
|
|
4 |
import time
|
5 |
from transformers import AutoTokenizer, AutoModelForSeq2SeqLM, pipeline
|
6 |
from flores200_codes import flores_codes
|
7 |
+
import vinai_translator
|
8 |
|
9 |
def load_models():
|
10 |
# build model and tokenizer
|
|
|
25 |
|
26 |
return model_dict
|
27 |
|
|
|
28 |
def translation(source, target, text):
|
|
|
|
|
|
|
|
|
|
|
29 |
source = flores_codes[source]
|
30 |
target = flores_codes[target]
|
|
|
|
|
|
|
|
|
|
|
|
|
31 |
output = ""
|
32 |
+
|
33 |
+
start_time = time.time()
|
34 |
+
|
35 |
+
if source = "vie_Latn" or target ="eng_Latn":
|
36 |
+
chunks = text.splitlines(True)
|
37 |
+
for chunk in chunks:
|
38 |
+
stchunk = vinai_translator.translate_vi2en(chunk)
|
39 |
+
output += stchunk+"\n"
|
40 |
+
pass
|
41 |
+
elif source = "eng_Latn" or target ="vie_Latn":
|
42 |
+
chunks = text.splitlines(True)
|
43 |
+
for chunk in chunks:
|
44 |
+
stchunk = vinai_translator.translate_en2vi(chunk)
|
45 |
+
output += stchunk+"\n"
|
46 |
+
pass
|
47 |
+
else:
|
48 |
+
if len(model_dict) == 2:
|
49 |
+
#model_name = 'nllb-distilled-600M'
|
50 |
+
model_name = 'nllb-distilled-1.3B'
|
51 |
+
|
52 |
+
model = model_dict[model_name + '_model']
|
53 |
+
tokenizer = model_dict[model_name + '_tokenizer']
|
54 |
+
|
55 |
+
translator = pipeline('translation', model=model, tokenizer=tokenizer, src_lang=source, tgt_lang=target)
|
56 |
+
chunks = text.splitlines(True)
|
57 |
+
for chunk in chunks:
|
58 |
+
stchunk = translator(chunk+"<mask>", max_length=400, num_beams=5)
|
59 |
+
output += stchunk[0]['translation_text']+"\n"
|
60 |
|
61 |
end_time = time.time()
|
62 |
|
|
|
66 |
'result': output}
|
67 |
return result
|
68 |
|
|
|
69 |
if __name__ == '__main__':
|
70 |
print('\tinit models')
|
71 |
|