Spaces:
Sleeping
Sleeping
Update app.py
Browse fileschange model to indic languages
app.py
CHANGED
@@ -17,10 +17,10 @@ app.add_middleware(
|
|
17 |
)
|
18 |
|
19 |
model = AutoModelForSeq2SeqLM.from_pretrained(
|
20 |
-
"ai4bharat/indictrans2-
|
21 |
)
|
22 |
tokenizer = AutoTokenizer.from_pretrained(
|
23 |
-
"ai4bharat/indictrans2-
|
24 |
)
|
25 |
|
26 |
ip = IndicProcessor(inference=True)
|
@@ -28,9 +28,8 @@ DEVICE = "cuda" if torch.cuda.is_available() else "cpu"
|
|
28 |
model = model.to(DEVICE)
|
29 |
|
30 |
|
31 |
-
def translate_text(sentences: List[str], target_lang: str):
|
32 |
try:
|
33 |
-
src_lang = "eng_Latn"
|
34 |
batch = ip.preprocess_batch(sentences, src_lang=src_lang, tgt_lang=target_lang)
|
35 |
inputs = tokenizer(
|
36 |
batch,
|
@@ -68,6 +67,7 @@ def read_root():
|
|
68 |
|
69 |
class TranslateRequest(BaseModel):
|
70 |
sentences: List[str]
|
|
|
71 |
target_lang: str
|
72 |
|
73 |
|
|
|
17 |
)
|
18 |
|
19 |
model = AutoModelForSeq2SeqLM.from_pretrained(
|
20 |
+
"ai4bharat/indictrans2-indic-indic-1B", trust_remote_code=True
|
21 |
)
|
22 |
tokenizer = AutoTokenizer.from_pretrained(
|
23 |
+
"ai4bharat/indictrans2-indic-indic-1B", trust_remote_code=True
|
24 |
)
|
25 |
|
26 |
ip = IndicProcessor(inference=True)
|
|
|
28 |
model = model.to(DEVICE)
|
29 |
|
30 |
|
31 |
+
def translate_text(sentences: List[str], src_lang: str, target_lang: str):
|
32 |
try:
|
|
|
33 |
batch = ip.preprocess_batch(sentences, src_lang=src_lang, tgt_lang=target_lang)
|
34 |
inputs = tokenizer(
|
35 |
batch,
|
|
|
67 |
|
68 |
class TranslateRequest(BaseModel):
|
69 |
sentences: List[str]
|
70 |
+
src_lang
|
71 |
target_lang: str
|
72 |
|
73 |
|