OFA-OCR / run_scripts /image_gen /generate_for_vqgan_code.sh
JustinLin610's picture
first commit
ee21b96
raw
history blame
No virus
1.39 kB
#!/usr/bin/env
# for text-image paired data, each line of the given input file should contain these information (separated by tabs):
# input format
# uniq-id, image-id, image base64 string and text
# input example
# 162365 12455 /9j/4AAQSkZJ....UCP/2Q== two people in an ocean playing with a yellow frisbee.
#
# output format
# uniq-id, image-id, text and code
# output example
# 162364 12455 two people in an ocean playing with a yellow frisbee. 6288 4495 4139...4691 4844 6464
CUDA_VISIBLE_DEVICES=0 python generate_code.py \
--file ./custom_data.txt \
--outputs ./custom_data_code.txt \
--selected_cols 0,1,2,3 \
--code_image_size 256 \
--vq_model vqgan \
--vqgan_model_path ../../checkpoints/vqgan/last.ckpt \
--vqgan_config_path ../../checkpoints/vqgan/model.yaml
# for image-only data each line of the given input file should contain these information (separated by tabs):
# input format
# image-id and image base64 string
# input example:
# 12455 /9j/4AAQSkZJ....UCP/2Q==
#
# output format
# image-id and code
# 12455 6288 4495 4139...4691 4844 6464
CUDA_VISIBLE_DEVICES=0 python generate_code.py \
--file ./custom_data.txt \
--outputs ./custom_data_code.txt \
--selected_cols 0,1 \
--code_image_size 256 \
--vq_model vqgan \
--vqgan_model_path ../../checkpoints/vqgan/last.ckpt \
--vqgan_config_path ../../checkpoints/vqgan/model.yaml