Update handler.py
Browse files- handler.py +45 -1
handler.py
CHANGED
@@ -1 +1,45 @@
|
|
1 |
-
#Handler.py file needed
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
1 |
+
#Handler.py file needed
|
2 |
+
|
3 |
+
from PIL import Image
|
4 |
+
import torch
|
5 |
+
from transformers import AutoProcessor, AutoModelForVision2Seq
|
6 |
+
|
7 |
+
class ModelHandler:
|
8 |
+
def __init__(self):
|
9 |
+
self.model = None
|
10 |
+
self.processor = None
|
11 |
+
|
12 |
+
def initialize(self, model_dir):
|
13 |
+
# Load the processor and model
|
14 |
+
self.processor = AutoProcessor.from_pretrained(model_dir)
|
15 |
+
self.model = AutoModelForVision2Seq.from_pretrained(model_dir)
|
16 |
+
|
17 |
+
def preprocess(self, inputs):
|
18 |
+
# Process the input image
|
19 |
+
image = Image.open(inputs["image"].file)
|
20 |
+
pixel_values = self.processor(images=image, return_tensors="pt").pixel_values
|
21 |
+
|
22 |
+
# Process the text context (if provided)
|
23 |
+
text_context = inputs.get("text_context", "")
|
24 |
+
if text_context:
|
25 |
+
context_inputs = self.processor(text=text_context, return_tensors="pt").input_ids
|
26 |
+
else:
|
27 |
+
context_inputs = None
|
28 |
+
|
29 |
+
return pixel_values, context_inputs
|
30 |
+
|
31 |
+
def inference(self, pixel_values, context_inputs=None):
|
32 |
+
# Run inference on the image with or without text context
|
33 |
+
with torch.no_grad():
|
34 |
+
if context_inputs is not None:
|
35 |
+
outputs = self.model.generate(pixel_values, input_ids=context_inputs)
|
36 |
+
else:
|
37 |
+
outputs = self.model.generate(pixel_values)
|
38 |
+
return outputs
|
39 |
+
|
40 |
+
def postprocess(self, outputs):
|
41 |
+
# Decode the output to text
|
42 |
+
decoded_text = self.processor.batch_decode(outputs, skip_special_tokens=True)
|
43 |
+
return {"digitized_text": decoded_text[0]}
|
44 |
+
|
45 |
+
service = ModelHandler()
|