Commit
·
e31682d
1
Parent(s):
f145981
Upload 2 files
Browse files- hgf.py +23 -0
- requirements.txt +1 -0
hgf.py
ADDED
@@ -0,0 +1,23 @@
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
1 |
+
from typing import Dict, List, Any
|
2 |
+
|
3 |
+
from transformers import MBart50TokenizerFast, MBartForConditionalGeneration
|
4 |
+
|
5 |
+
|
6 |
+
class EndpointHandler:
|
7 |
+
def __init__(self, path=""):
|
8 |
+
self.tokenizer = MBart50TokenizerFast.from_pretrained("facebook/mbart-large-50-many-to-many-mmt")
|
9 |
+
self.model = MBartForConditionalGeneration.from_pretrained(path)
|
10 |
+
|
11 |
+
def __call__(self, data: Dict[str, Any]) -> List[Dict[str, Any]]:
|
12 |
+
inputs = data.pop("inputs", data)
|
13 |
+
source_lang = data.pop("src_lang", None)
|
14 |
+
target_lang = data.pop("target_lang", None)
|
15 |
+
|
16 |
+
self.tokenizer.src_lang = source_lang
|
17 |
+
|
18 |
+
encoded = self.tokenizer(inputs, return_tensors='pt')
|
19 |
+
generated_tokens = self.model.generate(
|
20 |
+
**encoded, forced_bos_token_id=self.tokenizer.lang_code_to_id[target_lang]
|
21 |
+
)
|
22 |
+
d = self.tokenizer.batch_decode(generated_tokens, skip_special_tokens=True)
|
23 |
+
return d
|
requirements.txt
ADDED
@@ -0,0 +1 @@
|
|
|
|
|
1 |
+
transformers
|