Commit
•
baa2ff5
1
Parent(s):
993e825
Updating docstring, loading local model weights and adding parameters
Browse files- pipeline.py +21 -11
pipeline.py
CHANGED
@@ -5,7 +5,7 @@ import torch
|
|
5 |
import base64
|
6 |
import os
|
7 |
from io import BytesIO
|
8 |
-
from
|
9 |
from torchvision import transforms
|
10 |
from torchvision.transforms.functional import InterpolationMode
|
11 |
|
@@ -13,10 +13,15 @@ device = torch.device('cuda' if torch.cuda.is_available() else 'cpu')
|
|
13 |
print(device)
|
14 |
|
15 |
class PreTrainedPipeline():
|
16 |
-
def __init__(self
|
17 |
# load the optimized model
|
18 |
-
self.
|
19 |
-
self.model = blip_decoder(
|
|
|
|
|
|
|
|
|
|
|
20 |
self.model.eval()
|
21 |
self.model = self.model.to(device)
|
22 |
|
@@ -29,23 +34,28 @@ class PreTrainedPipeline():
|
|
29 |
|
30 |
|
31 |
|
32 |
-
def __call__(self, data: Any) ->
|
33 |
"""
|
34 |
Args:
|
35 |
data (:obj:):
|
36 |
includes the input data and the parameters for the inference.
|
37 |
Return:
|
38 |
-
A :obj:`
|
39 |
-
- "
|
40 |
-
- "score": A score between 0 and 1 describing how confident the model is for this label/class.
|
41 |
"""
|
42 |
inputs = data.pop("inputs", data)
|
43 |
-
parameters = data.pop("parameters",
|
44 |
|
45 |
# decode base64 image to PIL
|
46 |
image = Image.open(BytesIO(base64.b64decode(inputs['image'])))
|
47 |
image = self.transform(image).unsqueeze(0).to(device)
|
48 |
with torch.no_grad():
|
49 |
-
caption = self.model.generate(
|
|
|
|
|
|
|
|
|
|
|
|
|
50 |
# postprocess the prediction
|
51 |
-
return caption
|
|
|
5 |
import base64
|
6 |
import os
|
7 |
from io import BytesIO
|
8 |
+
from models.blip_decoder import blip_decoder
|
9 |
from torchvision import transforms
|
10 |
from torchvision.transforms.functional import InterpolationMode
|
11 |
|
|
|
13 |
print(device)
|
14 |
|
15 |
class PreTrainedPipeline():
|
16 |
+
def __init__(self):
|
17 |
# load the optimized model
|
18 |
+
self.model_path = 'model_base_capfilt_large.pth'
|
19 |
+
self.model = blip_decoder(
|
20 |
+
pretrained=self.model_path,
|
21 |
+
image_size=384,
|
22 |
+
vit='large',
|
23 |
+
med_config=os.path.join(path, 'configs/med_config.json')
|
24 |
+
)
|
25 |
self.model.eval()
|
26 |
self.model = self.model.to(device)
|
27 |
|
|
|
34 |
|
35 |
|
36 |
|
37 |
+
def __call__(self, data: Any) -> Dict[str]:
|
38 |
"""
|
39 |
Args:
|
40 |
data (:obj:):
|
41 |
includes the input data and the parameters for the inference.
|
42 |
Return:
|
43 |
+
A :obj:`dict`:. The object returned should be a dict of one list like [[{"label": 0.9939950108528137}]] containing :
|
44 |
+
- "caption": A string corresponding to the generated caption.
|
|
|
45 |
"""
|
46 |
inputs = data.pop("inputs", data)
|
47 |
+
parameters = data.pop("parameters", {})
|
48 |
|
49 |
# decode base64 image to PIL
|
50 |
image = Image.open(BytesIO(base64.b64decode(inputs['image'])))
|
51 |
image = self.transform(image).unsqueeze(0).to(device)
|
52 |
with torch.no_grad():
|
53 |
+
caption = self.model.generate(
|
54 |
+
image,
|
55 |
+
sample=parameters.get('sample',True),
|
56 |
+
top_p=parameters.get('top_p',0.9),
|
57 |
+
max_length=parameters.get('max_length',20),
|
58 |
+
min_length=parameters.get('min_length',5)
|
59 |
+
)
|
60 |
# postprocess the prediction
|
61 |
+
return {"caption": caption}
|