s444415/donut
1
0
Fork 0
Code Issues Pull Requests Projects Releases Wiki Activity

config eval and donut eval better params

Browse Source
This commit is contained in:
s444415 2023-01-04 10:00:50 +01:00
parent ac724952a1
commit 9a53f14742
2 changed files with 9 additions and 6 deletions
Show Stats Download Patch File Download Diff File Expand all files Collapse all files

3
config-eval.yaml
View File

@ -6,3 +6,6 @@ has_metadata: False
print_output: True print_output: True
output_file_dir: "../../gonito-outs" output_file_dir: "../../gonito-outs"
test_name: "fine-tuned-test" test_name: "fine-tuned-test"
image_size: [1920, 2560]
use_enc_dec_config: False
max_dec_length: 768

10
donut-eval.py
View File

@ -15,15 +15,15 @@ from sconf import Config
def main(config): def main(config):
# image_size = [1920, 2560] if config.use_enc_dec_config:
# config_vision = VisionEncoderDecoderConfig.from_pretrained(config.pretrained_model_path) config_vision = VisionEncoderDecoderConfig.from_pretrained(config.pretrained_model_path)
# config_vision.encoder.image_size = image_size # (height, width) config_vision.encoder.image_size = config.image_size # (height, width)
# config_vision.decoder.max_length = 768 config_vision.decoder.max_length = config.max_dec_length
processor = DonutProcessor.from_pretrained(config.pretrained_processor_path) processor = DonutProcessor.from_pretrained(config.pretrained_processor_path)
model = VisionEncoderDecoderModel.from_pretrained(config.pretrained_model_path) model = VisionEncoderDecoderModel.from_pretrained(config.pretrained_model_path)
# processor.image_processor.size = image_size[::-1] # should be (width, height) processor.image_processor.size = config.image_size[::-1] # should be (width, height)
processor.image_processor.do_align_long_axis = False processor.image_processor.do_align_long_axis = False
dataset = load_dataset(config.validation_dataset_path, split=config.validation_dataset_split) dataset = load_dataset(config.validation_dataset_path, split=config.validation_dataset_split)