Spaces:
Sleeping
Sleeping
Update app.py
Browse files
app.py
CHANGED
@@ -7,6 +7,10 @@ from transformers import AutoTokenizer, pipeline
|
|
7 |
# Modelo de Meta capaz de traducir a más de 200 idiomas
|
8 |
model = 'facebook/nllb-200-distilled-600M'
|
9 |
tokenizer = model
|
|
|
|
|
|
|
|
|
10 |
|
11 |
flores_codes = {}
|
12 |
flores_codes["Asturianu"] = "ast_Latn"
|
@@ -21,11 +25,16 @@ def translation(source, target, text):
|
|
21 |
#start_time = time.time()
|
22 |
source = flores_codes[source]
|
23 |
target = flores_codes[target]
|
24 |
-
|
25 |
-
|
26 |
-
|
27 |
-
|
28 |
-
|
|
|
|
|
|
|
|
|
|
|
29 |
|
30 |
#end_time = time.time()
|
31 |
|
|
|
7 |
# Modelo de Meta capaz de traducir a más de 200 idiomas
|
8 |
model = 'facebook/nllb-200-distilled-600M'
|
9 |
tokenizer = model
|
10 |
+
# Pero como traduce mal al asturiano, añadimos este otro que traduce del español
|
11 |
+
# Es del proyecto AINA: https://huggingface.co/projecte-aina
|
12 |
+
model_ast = "projecte-aina/aina-translator-es-ast"
|
13 |
+
tokenizer_ast = model_ast
|
14 |
|
15 |
flores_codes = {}
|
16 |
flores_codes["Asturianu"] = "ast_Latn"
|
|
|
25 |
#start_time = time.time()
|
26 |
source = flores_codes[source]
|
27 |
target = flores_codes[target]
|
28 |
+
if target == flores_codes["Asturianu"]:
|
29 |
+
texto_castellano = source
|
30 |
+
if source != flores_codes["Castellano"]:
|
31 |
+
translator = pipeline('translation', model=model, tokenizer=tokenizer, src_lang=source, tgt_lang=flores_codes["Castellano"])
|
32 |
+
texto_castellano = translator(text, max_length=400)
|
33 |
+
translator_ast = pipeline('translation', model=model_ast, tokenizer=tokenizer_ast)
|
34 |
+
output = translator_ast(text, max_length=400)
|
35 |
+
else
|
36 |
+
translator = pipeline('translation', model=model, tokenizer=tokenizer, src_lang=source, tgt_lang=target)
|
37 |
+
output = translator(text, max_length=400)
|
38 |
|
39 |
#end_time = time.time()
|
40 |
|