File size: 1,588 Bytes
2a261b1
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
1
2
3
4
5
6
7
8
9
10
11
12
13
14
15
16
17
18
19
20
21
22
23
24
25
26
27
28
29
30
31
32
33
34
35
36
37
38
39
40
41
42
43
44
45
46
47
from typing import Dict, List, Any
from transformers import DonutProcessor, VisionEncoderDecoderModel
import torch


# check for GPU
device = 0 if torch.cuda.is_available() else -1


class EndpointHandler:
    def __init__(self, path=""):
        # load the model
        self.processor = DonutProcessor.from_pretrained(path)
        self.model = VisionEncoderDecoderModel.from_pretrained(path)
        # move model to device
        self.model.to(device)
        self.decoder_input_ids = self.processor.tokenizer(
            "<s_cord-v2>", add_special_tokens=False, return_tensors="pt"
        ).input_ids

    def __call__(self, data: Any) -> List[List[Dict[str, float]]]:

        inputs = data.pop("inputs", data)


        # preprocess the input
        pixel_values = self.processor(inputs, return_tensors="pt").pixel_values

        # forward pass
        outputs = self.model.generate(
            pixel_values.to(device),
            decoder_input_ids=self.decoder_input_ids.to(device),
            max_length=self.model.decoder.config.max_position_embeddings,
            early_stopping=True,
            pad_token_id=self.processor.tokenizer.pad_token_id,
            eos_token_id=self.processor.tokenizer.eos_token_id,
            use_cache=True,
            num_beams=1,
            bad_words_ids=[[self.processor.tokenizer.unk_token_id]],
            return_dict_in_generate=True,
        )
        # process output
        prediction = self.processor.batch_decode(outputs.sequences)[0]
        prediction = self.processor.token2json(prediction)

        return prediction