Spaces:
Runtime error
Runtime error
Update inference.py
Browse files- inference.py +16 -4
inference.py
CHANGED
@@ -2,6 +2,18 @@ import torch
|
|
2 |
from torchtext.data.utils import get_tokenizer
|
3 |
from model_arch import TextClassifierModel, load_state_dict
|
4 |
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
5 |
model_trained = torch.load('model_checkpoint.pth')
|
6 |
vocab = torch.load('vocab.pt')
|
7 |
tokenizer = get_tokenizer("spacy", language="es")
|
@@ -17,8 +29,8 @@ model = TextClassifierModel(vocab_size, embed_size, num_class)
|
|
17 |
model = load_state_dict(model, model_trained, vocab)
|
18 |
|
19 |
def predict(text, model=model, text_pipeline=text_pipeline):
|
20 |
-
with torch.no_grad()
|
21 |
-
|
22 |
-
|
23 |
-
|
24 |
|
|
|
2 |
from torchtext.data.utils import get_tokenizer
|
3 |
from model_arch import TextClassifierModel, load_state_dict
|
4 |
|
5 |
+
labels = {0: 'messaging',
|
6 |
+
1: 'calling',
|
7 |
+
2: 'event',
|
8 |
+
3: 'timer',
|
9 |
+
4: 'music',
|
10 |
+
5: 'weather',
|
11 |
+
6: 'alarm',
|
12 |
+
7: 'people',
|
13 |
+
8: 'reminder',
|
14 |
+
9: 'recipes',
|
15 |
+
10: 'news'}
|
16 |
+
|
17 |
model_trained = torch.load('model_checkpoint.pth')
|
18 |
vocab = torch.load('vocab.pt')
|
19 |
tokenizer = get_tokenizer("spacy", language="es")
|
|
|
29 |
model = load_state_dict(model, model_trained, vocab)
|
30 |
|
31 |
def predict(text, model=model, text_pipeline=text_pipeline):
|
32 |
+
with torch.no_grad():
|
33 |
+
model.eval()
|
34 |
+
text_tensor = torch.tensor(text_pipeline(text))
|
35 |
+
return labels[model(text_tensor, torch.tensor([0])).argmax(1).item()]
|
36 |
|