py_ait 0.0.40

This module provide a toolbox for ai dev.

AI-Toolbox

Command

---

ait-download

Download hf model,dataset and will retry after a failure, default 10 times.

ait-download \
--repo_info_path=/your_repo_info.json \
--retry_count=2

Demo repo_info.json

{
    "model":[
        "OpenAssistant/falcon-40b-megacode2-oasst",
        "stabilityai/stable-diffusion-xl-base-1.0"
    ],
    "dataset":[
        "OpenAssistant/oasst1",
        "PygmalionAI/PIPPA"
    ]
}

---

ait-fork

fork hf model, dataset and will retry after a failure, default 10 times.

# fork with fork_repo_info.json
ait-fork \
--repo_info_path=/your_repo_info.json \
--retry_count=2 \
--hf_username=your_hf_username \
--hf_token=your_hf_token \
--separator=__

# fork without fork_repo_info.json
ait-fork \
--retry_count=2 \
--hf_username=your_hf_username \
--hf_token=your_hf_token \
--separator=__ \
--source_repo=source_repo \
--target_repo=your_target_repo \
--repo_type=dataset|model \
--disable-multi-task

Args description

you must pass --repo_info_path or --source_repo at least one!

name	description	required	default
--hf_username	your hugging face username	True	None
--hf_token	your hugging face token	True	None
--repo_info_path	repo info file for fork multi repos at once, demo below!	False	None
--retry_count	retry count after a failure	False	10
--separator	a string used to replace / when you did not pass target_repo without --repo_info_path	False	__
--source_repo	the source repo you want to fork! eg: PygmalionAI/PIPPA,	False	10
--target_repo	the target repo you want to fork to! eg: your_hf_username/your_repo_name, if not pass, use your_hf_username/source_repo.replace_/_with_separator	False	None
--repo_type	model or dataset	False	dataset
--disable-multi-task	disable multi task for forking multi repos when you pass, only work with --repo_info_path	False	False

Demo fork_repo_info.json

{
    "model":[
        {
            "source":"OpenAssistant/falcon-40b-megacode2-oasst",
            "target":"your_name/your_repo"
        }        
    ],
    "dataset":[
        {
            "source":"PygmalionAI/PIPPA",
            "target":"your_name/your_repo"
        }
    ]
}