Spaces:
Sleeping
Sleeping
Update app.py
Browse files
app.py
CHANGED
@@ -7,6 +7,14 @@ langs = {"German": "de", "Romanian": "ro", "English": "en", "French": "fr", "Spa
|
|
7 |
options = list(langs.keys())
|
8 |
models = ["Helsinki-NLP", "t5-base", "t5-small", "t5-large", "facebook/nllb-200-distilled-600M", "facebook/nllb-200-distilled-1.3B", "facebook/mbart-large-50-many-to-many-mmt"]
|
9 |
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
10 |
@spaces.GPU
|
11 |
def translate_text(input_text, sselected_language, tselected_language, model_name):
|
12 |
sl = langs[sselected_language]
|
@@ -16,14 +24,15 @@ def translate_text(input_text, sselected_language, tselected_language, model_nam
|
|
16 |
try:
|
17 |
model_name_full = f"Helsinki-NLP/opus-mt-{sl}-{tl}"
|
18 |
tokenizer = AutoTokenizer.from_pretrained(model_name_full)
|
19 |
-
model = AutoModelForSeq2SeqLM.from_pretrained(model_name_full)
|
20 |
except EnvironmentError:
|
21 |
try :
|
22 |
model_name_full = f"Helsinki-NLP/opus-tatoeba-{sl}-{tl}"
|
23 |
tokenizer = AutoTokenizer.from_pretrained(model_name_full)
|
24 |
-
model = AutoModelForSeq2SeqLM.from_pretrained(model_name_full)
|
25 |
except EnvironmentError as error:
|
26 |
return f"Error finding model: {model_name_full}! Try other available language combination.", error
|
|
|
27 |
if model_name.startswith('facebook/nllb'):
|
28 |
from languagecodes import nllb_language_codes
|
29 |
tokenizer = AutoTokenizer.from_pretrained(model_name, src_lang=nllb_language_codes[sselected_language])
|
@@ -31,6 +40,7 @@ def translate_text(input_text, sselected_language, tselected_language, model_nam
|
|
31 |
translator = pipeline('translation', model=model, tokenizer=tokenizer, src_lang=nllb_language_codes[sselected_language], tgt_lang=nllb_language_codes[tselected_language])
|
32 |
translated_text = translator(input_text, max_length=512)
|
33 |
return translated_text[0]['translation_text'], message_text
|
|
|
34 |
if model_name.startswith('facebook/mbart-large'):
|
35 |
from languagecodes import mbart_large_languages
|
36 |
from transformers import MBartForConditionalGeneration, MBart50TokenizerFast
|
@@ -43,18 +53,12 @@ def translate_text(input_text, sselected_language, tselected_language, model_nam
|
|
43 |
**encoded,
|
44 |
forced_bos_token_id=tokenizer.lang_code_to_id[mbart_large_languages[tselected_language]]
|
45 |
)
|
46 |
-
return tokenizer.batch_decode(generated_tokens, skip_special_tokens=True), message_text
|
47 |
|
48 |
if model_name.startswith('t5'):
|
49 |
tokenizer = T5Tokenizer.from_pretrained(model_name)
|
50 |
model = T5ForConditionalGeneration.from_pretrained(model_name, device_map="auto")
|
51 |
|
52 |
-
# Move the model to GPU if available
|
53 |
-
if torch.cuda.is_available():
|
54 |
-
model = model.to('cuda')
|
55 |
-
print("CUDA is available! Using GPU.")
|
56 |
-
else:
|
57 |
-
print("CUDA not available! Using CPU.")
|
58 |
if model_name.startswith("Helsinki-NLP"):
|
59 |
prompt = input_text
|
60 |
else:
|
|
|
7 |
options = list(langs.keys())
|
8 |
models = ["Helsinki-NLP", "t5-base", "t5-small", "t5-large", "facebook/nllb-200-distilled-600M", "facebook/nllb-200-distilled-1.3B", "facebook/mbart-large-50-many-to-many-mmt"]
|
9 |
|
10 |
+
def model_to_cuda(model):
|
11 |
+
# Move the model to GPU if available
|
12 |
+
if torch.cuda.is_available():
|
13 |
+
model = model.to('cuda')
|
14 |
+
print("CUDA is available! Using GPU.")
|
15 |
+
else:
|
16 |
+
print("CUDA not available! Using CPU.")
|
17 |
+
return model
|
18 |
@spaces.GPU
|
19 |
def translate_text(input_text, sselected_language, tselected_language, model_name):
|
20 |
sl = langs[sselected_language]
|
|
|
24 |
try:
|
25 |
model_name_full = f"Helsinki-NLP/opus-mt-{sl}-{tl}"
|
26 |
tokenizer = AutoTokenizer.from_pretrained(model_name_full)
|
27 |
+
model = model_to_cuda(AutoModelForSeq2SeqLM.from_pretrained(model_name_full))
|
28 |
except EnvironmentError:
|
29 |
try :
|
30 |
model_name_full = f"Helsinki-NLP/opus-tatoeba-{sl}-{tl}"
|
31 |
tokenizer = AutoTokenizer.from_pretrained(model_name_full)
|
32 |
+
model = model_to_cuda(AutoModelForSeq2SeqLM.from_pretrained(model_name_full))
|
33 |
except EnvironmentError as error:
|
34 |
return f"Error finding model: {model_name_full}! Try other available language combination.", error
|
35 |
+
|
36 |
if model_name.startswith('facebook/nllb'):
|
37 |
from languagecodes import nllb_language_codes
|
38 |
tokenizer = AutoTokenizer.from_pretrained(model_name, src_lang=nllb_language_codes[sselected_language])
|
|
|
40 |
translator = pipeline('translation', model=model, tokenizer=tokenizer, src_lang=nllb_language_codes[sselected_language], tgt_lang=nllb_language_codes[tselected_language])
|
41 |
translated_text = translator(input_text, max_length=512)
|
42 |
return translated_text[0]['translation_text'], message_text
|
43 |
+
|
44 |
if model_name.startswith('facebook/mbart-large'):
|
45 |
from languagecodes import mbart_large_languages
|
46 |
from transformers import MBartForConditionalGeneration, MBart50TokenizerFast
|
|
|
53 |
**encoded,
|
54 |
forced_bos_token_id=tokenizer.lang_code_to_id[mbart_large_languages[tselected_language]]
|
55 |
)
|
56 |
+
return tokenizer.batch_decode(generated_tokens, skip_special_tokens=True)[0], message_text
|
57 |
|
58 |
if model_name.startswith('t5'):
|
59 |
tokenizer = T5Tokenizer.from_pretrained(model_name)
|
60 |
model = T5ForConditionalGeneration.from_pretrained(model_name, device_map="auto")
|
61 |
|
|
|
|
|
|
|
|
|
|
|
|
|
62 |
if model_name.startswith("Helsinki-NLP"):
|
63 |
prompt = input_text
|
64 |
else:
|