amaye15
commited on
Commit
·
c26d617
1
Parent(s):
8d41aec
Docstring V2
Browse files- .gitignore +2 -1
- handler.py +25 -73
.gitignore
CHANGED
@@ -1 +1,2 @@
|
|
1 |
-
*.DS*
|
|
|
|
1 |
+
*.DS*
|
2 |
+
*__pycache__*
|
handler.py
CHANGED
@@ -1,63 +1,3 @@
|
|
1 |
-
# import torch
|
2 |
-
# from typing import Dict, Any
|
3 |
-
# from PIL import Image
|
4 |
-
# import base64
|
5 |
-
# from io import BytesIO
|
6 |
-
|
7 |
-
|
8 |
-
# class EndpointHandler:
|
9 |
-
# def __init__(self, path: str = ""):
|
10 |
-
# # Import your model and processor inside the class
|
11 |
-
# from colpali_engine.models import ColQwen2, ColQwen2Processor
|
12 |
-
|
13 |
-
# # Load the model and processor
|
14 |
-
# self.model = ColQwen2.from_pretrained(
|
15 |
-
# path,
|
16 |
-
# torch_dtype=torch.bfloat16,
|
17 |
-
# ).eval()
|
18 |
-
# self.processor = ColQwen2Processor.from_pretrained(path)
|
19 |
-
|
20 |
-
# # Determine the device
|
21 |
-
# self.device = torch.device("cuda" if torch.cuda.is_available() else "cpu")
|
22 |
-
# self.model.to(self.device)
|
23 |
-
|
24 |
-
# def __call__(self, data: Dict[str, Any]) -> Dict[str, Any]:
|
25 |
-
# # Extract images from the input data
|
26 |
-
# images_data = data.get("inputs", [])
|
27 |
-
|
28 |
-
# if not images_data:
|
29 |
-
# return {"error": "No images provided in 'inputs'."}
|
30 |
-
|
31 |
-
# # Process images
|
32 |
-
# images = []
|
33 |
-
# for img_data in images_data:
|
34 |
-
# if isinstance(img_data, str):
|
35 |
-
# try:
|
36 |
-
# # Assume base64-encoded image
|
37 |
-
# image_bytes = base64.b64decode(img_data)
|
38 |
-
# image = Image.open(BytesIO(image_bytes)).convert("RGB")
|
39 |
-
# images.append(image)
|
40 |
-
# except Exception as e:
|
41 |
-
# return {"error": f"Invalid image data: {e}"}
|
42 |
-
# else:
|
43 |
-
# return {"error": "Images should be base64-encoded strings."}
|
44 |
-
|
45 |
-
# # Prepare inputs
|
46 |
-
# batch_images = self.processor.process_images(images)
|
47 |
-
|
48 |
-
# # Move tensors to the device
|
49 |
-
# batch_images = {k: v.to(self.device) for k, v in batch_images.items()}
|
50 |
-
|
51 |
-
# # Generate embeddings
|
52 |
-
# with torch.no_grad():
|
53 |
-
# image_embeddings = self.model(**batch_images)
|
54 |
-
|
55 |
-
# # Convert embeddings to a list
|
56 |
-
# embeddings_list = image_embeddings.cpu().tolist()
|
57 |
-
|
58 |
-
# return {"embeddings": embeddings_list}
|
59 |
-
|
60 |
-
|
61 |
import torch
|
62 |
from typing import Dict, Any, List
|
63 |
from PIL import Image
|
@@ -67,13 +7,17 @@ from io import BytesIO
|
|
67 |
|
68 |
class EndpointHandler:
|
69 |
"""
|
70 |
-
A handler class for processing image data
|
71 |
|
72 |
Attributes:
|
73 |
-
model:
|
74 |
-
|
75 |
-
|
76 |
-
|
|
|
|
|
|
|
|
|
77 |
"""
|
78 |
|
79 |
def __init__(self, path: str = "", default_batch_size: int = 4):
|
@@ -81,8 +25,13 @@ class EndpointHandler:
|
|
81 |
Initializes the EndpointHandler with a specified model path and default batch size.
|
82 |
|
83 |
Args:
|
84 |
-
path (str):
|
85 |
-
|
|
|
|
|
|
|
|
|
|
|
86 |
"""
|
87 |
from colpali_engine.models import ColQwen2, ColQwen2Processor
|
88 |
|
@@ -101,10 +50,11 @@ class EndpointHandler:
|
|
101 |
Processes a batch of images and generates embeddings.
|
102 |
|
103 |
Args:
|
104 |
-
images (List[Image.Image]):
|
|
|
105 |
|
106 |
-
|
107 |
-
List[List[float]]
|
108 |
"""
|
109 |
batch_images = self.processor.process_images(images)
|
110 |
batch_images = {k: v.to(self.device) for k, v in batch_images.items()}
|
@@ -119,10 +69,12 @@ class EndpointHandler:
|
|
119 |
Processes input data containing base64-encoded images, decodes them, and generates embeddings.
|
120 |
|
121 |
Args:
|
122 |
-
data (Dict[str, Any]):
|
|
|
123 |
|
124 |
-
|
125 |
-
|
|
|
126 |
"""
|
127 |
images_data = data.get("inputs", [])
|
128 |
batch_size = data.get("batch_size", self.default_batch_size)
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
1 |
import torch
|
2 |
from typing import Dict, Any, List
|
3 |
from PIL import Image
|
|
|
7 |
|
8 |
class EndpointHandler:
|
9 |
"""
|
10 |
+
A handler class for processing image data and generating embeddings using a specified model and processor.
|
11 |
|
12 |
Attributes:
|
13 |
+
model (:obj:):
|
14 |
+
The pre-trained model used for generating embeddings.
|
15 |
+
processor (:obj:):
|
16 |
+
The pre-trained processor used to process images before model inference.
|
17 |
+
device (:obj:):
|
18 |
+
The device (CPU or CUDA) used to run model inference.
|
19 |
+
default_batch_size (:obj:int:):
|
20 |
+
The default batch size for processing images in batches.
|
21 |
"""
|
22 |
|
23 |
def __init__(self, path: str = "", default_batch_size: int = 4):
|
|
|
25 |
Initializes the EndpointHandler with a specified model path and default batch size.
|
26 |
|
27 |
Args:
|
28 |
+
path (:obj:`str`, optional):
|
29 |
+
Path to the pre-trained model and processor.
|
30 |
+
default_batch_size (:obj:`int`, optional):
|
31 |
+
Default batch size for image processing.
|
32 |
+
|
33 |
+
Return:
|
34 |
+
None
|
35 |
"""
|
36 |
from colpali_engine.models import ColQwen2, ColQwen2Processor
|
37 |
|
|
|
50 |
Processes a batch of images and generates embeddings.
|
51 |
|
52 |
Args:
|
53 |
+
images (:obj:`List[Image.Image]`):
|
54 |
+
List of images to process.
|
55 |
|
56 |
+
Return:
|
57 |
+
A :obj:`List[List[float]]`. A list of embeddings for each image, where each embedding is a list of floats.
|
58 |
"""
|
59 |
batch_images = self.processor.process_images(images)
|
60 |
batch_images = {k: v.to(self.device) for k, v in batch_images.items()}
|
|
|
69 |
Processes input data containing base64-encoded images, decodes them, and generates embeddings.
|
70 |
|
71 |
Args:
|
72 |
+
data (:obj:`Dict[str, Any]`):
|
73 |
+
Includes the input data and the parameters for the inference, such as "inputs" containing a list of base64-encoded images and an optional "batch_size".
|
74 |
|
75 |
+
Return:
|
76 |
+
A :obj:`dict`. The object returned should be a dict like {"embeddings": [[0.6331314444541931, 0.8802216053009033, ..., -0.7866355180740356]]} containing:
|
77 |
+
- "embeddings": A list of lists, where each inner list is a set of floats corresponding to the embeddings of each image.
|
78 |
"""
|
79 |
images_data = data.get("inputs", [])
|
80 |
batch_size = data.get("batch_size", self.default_batch_size)
|