Spaces:
Running
Running
Update src/translate/Translate.py
Browse files- src/translate/Translate.py +30 -33
src/translate/Translate.py
CHANGED
@@ -3,9 +3,9 @@ from transformers import AutoTokenizer, AutoModelForSeq2SeqLM
|
|
3 |
import torch
|
4 |
import src.exception.Exception.Exception as ExceptionCustom
|
5 |
|
6 |
-
|
7 |
METHOD = "TRANSLATE"
|
8 |
|
|
|
9 |
tokenizerROMENG = AutoTokenizer.from_pretrained("BlackKakapo/opus-mt-ro-en")
|
10 |
modelROMENG = AutoModelForSeq2SeqLM.from_pretrained("BlackKakapo/opus-mt-ro-en")
|
11 |
|
@@ -16,40 +16,37 @@ device = torch.device("cuda" if torch.cuda.is_available() else "cpu")
|
|
16 |
modelROMENG.to(device)
|
17 |
modelENGROM.to(device)
|
18 |
|
|
|
|
|
|
|
|
|
19 |
|
20 |
-
|
21 |
-
|
22 |
-
exception = ""
|
23 |
-
result_value = ""
|
24 |
-
|
25 |
-
exception = ExceptionCustom.checkForException(requestValue, METHOD)
|
26 |
-
if exception != "":
|
27 |
-
return "", exception
|
28 |
-
|
29 |
-
tokenized_sent_list = sent_tokenize(requestValue)
|
30 |
|
31 |
-
|
32 |
if model == 'roen':
|
33 |
input_ids = tokenizerROMENG(SENTENCE, return_tensors='pt').to(device)
|
34 |
-
|
35 |
-
|
36 |
-
|
37 |
-
|
38 |
-
|
39 |
-
|
40 |
-
|
41 |
-
|
42 |
-
|
43 |
else:
|
44 |
-
|
45 |
-
|
46 |
-
|
47 |
-
|
48 |
-
|
49 |
-
|
50 |
-
|
51 |
-
|
52 |
-
|
53 |
-
|
54 |
-
|
55 |
-
|
|
|
|
3 |
import torch
|
4 |
import src.exception.Exception.Exception as ExceptionCustom
|
5 |
|
|
|
6 |
METHOD = "TRANSLATE"
|
7 |
|
8 |
+
# Load models and tokenizers
|
9 |
tokenizerROMENG = AutoTokenizer.from_pretrained("BlackKakapo/opus-mt-ro-en")
|
10 |
modelROMENG = AutoModelForSeq2SeqLM.from_pretrained("BlackKakapo/opus-mt-ro-en")
|
11 |
|
|
|
16 |
modelROMENG.to(device)
|
17 |
modelENGROM.to(device)
|
18 |
|
19 |
+
def paraphraseTranslateMethod(requestValue: str, model: str):
|
20 |
+
exception = ExceptionCustom.checkForException(requestValue, METHOD)
|
21 |
+
if exception:
|
22 |
+
return "", exception
|
23 |
|
24 |
+
tokenized_sent_list = sent_tokenize(requestValue)
|
25 |
+
result_value = []
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
26 |
|
27 |
+
for SENTENCE in tokenized_sent_list:
|
28 |
if model == 'roen':
|
29 |
input_ids = tokenizerROMENG(SENTENCE, return_tensors='pt').to(device)
|
30 |
+
output = modelROMENG.generate(
|
31 |
+
input_ids=input_ids.input_ids,
|
32 |
+
do_sample=True,
|
33 |
+
max_length=512,
|
34 |
+
top_k=90,
|
35 |
+
top_p=0.97,
|
36 |
+
early_stopping=False
|
37 |
+
)
|
38 |
+
result = tokenizerROMENG.batch_decode(output, skip_special_tokens=True)[0]
|
39 |
else:
|
40 |
+
input_ids = tokenizerENGROM(SENTENCE, return_tensors='pt').to(device)
|
41 |
+
output = modelENGROM.generate(
|
42 |
+
input_ids=input_ids.input_ids,
|
43 |
+
do_sample=True,
|
44 |
+
max_length=512,
|
45 |
+
top_k=90,
|
46 |
+
top_p=0.97,
|
47 |
+
early_stopping=False
|
48 |
+
)
|
49 |
+
result = tokenizerENGROM.batch_decode(output, skip_special_tokens=True)[0]
|
50 |
+
result_value.append(result)
|
51 |
+
|
52 |
+
return " ".join(result_value).strip(), model
|