|
--- |
|
license: mit |
|
--- |
|
### Usage |
|
Inference Code for this model |
|
|
|
``` |
|
import re |
|
import transformers |
|
from transformers import DonutProcessor, VisionEncoderDecoderModel |
|
import torch |
|
|
|
fine_tuned_model = VisionEncoderDecoderModel.from_pretrained("aravind-selvam/donut_finetuned_chart") |
|
processor = DonutProcessor.from_pretrained("aravind-selvam/donut_finetuned_chart") |
|
|
|
# Move model to GPU |
|
device = "cuda" if torch.cuda.is_available() else "cpu" |
|
fine_tuned_model.to(device) |
|
|
|
# Load random document image from the test set |
|
dataset = load_dataset("hf-internal-testing/example-documents", split="test") |
|
sample_image = dataset[1] |
|
|
|
def run_prediction(sample, model=fine_tuned_model, processor=processor): |
|
# pixel values |
|
pixel_values = processor(image, return_tensors="pt").pixel_values |
|
# prepare inputs |
|
task_prompt = "<s>" |
|
decoder_input_ids = processor.tokenizer(task_prompt, add_special_tokens=False, return_tensors="pt").input_ids |
|
|
|
# run inference |
|
outputs = model.generate( |
|
pixel_values.to(device), |
|
decoder_input_ids=decoder_input_ids.to(device), |
|
max_length=model.decoder.config.max_position_embeddings, |
|
early_stopping=True, |
|
pad_token_id=processor.tokenizer.pad_token_id, |
|
eos_token_id=processor.tokenizer.eos_token_id, |
|
use_cache=True, |
|
num_beams=2, |
|
# bad_words_ids=[[processor.tokenizer.unk_token_id]], |
|
return_dict_in_generate=True, |
|
) |
|
|
|
# process output |
|
prediction = processor.batch_decode(outputs.sequences)[0] |
|
prediction = re.sub(r"<one>", "1", prediction) |
|
prediction = processor.token2json(prediction) |
|
|
|
|
|
# load reference target |
|
target = processor.token2json(test_sample["target_sequence"]) |
|
return prediction, target |
|
|
|
prediction, target = run_prediction(sample_image) |
|
print(f"Reference:\n {target}") |
|
print(f"Prediction:\n {prediction}") |
|
``` |
