ballontranslator.install


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

post_install() {
cat << EOF
	Run "pip install spacy_pkuseg deepl sentencepiece ctranslate2 docxtxt opencc"
	because those packages are not in the AUR (yet)
	
	Install the other optional dependecies with whatever like 
	"yay -S python-docx python-jaconv python-keyboard python-fugashi python-einops python-transformers python-pytorch python-torchvision"
	and change "python-pytorch python-torchvision" that matches your hardware or just don't install them
	because "nccl" and friends (dependecies) takes a lot of space, so go to the pytorch website if you don't have an NVIDIA GPU

	OCRs 
	mit32px: "supports recognizing English, Japanese, Chinese, and Korean characters"
	mit48px: "same as above"
	manga_ocr: "only recognizes Japanese characters"

	Inpainters
	lama_mpe: "a deep-learning model: Lama with Masking Positional Encoding"
	aot: "a deep-learning model"
	patchmatch: "a randomized correspondence algorithm, Adobe uses this algorithm"
	opencv-telea: "based on the Fast Marching Method algorithm"
	
	Offline Translator
	sugoi: "offline Japanese to English"
	
	Download them and put them into /opt/ballontranslator/data/models
	or see and run "/opt/ballontranslator/scripts/download_models.sh" (may change in the future)
EOF
}

post_upgrade() {
    post_install
}