Hugging Face 🤗 x OpenDILab/DI-engine

English | 简体中文

Huggingface_ding is a library for OpenDILab/DI-engine user to push and pull models from the Huggingface Hub.

• Installation
• Examples
  • Download Model
  • Upload Model
• API

Installation

With pip

pip install -e .

Examples

Download Model

I want to download a model from the Hub

from ding.bonus import TD3Agent
from huggingface_ding import pull_model_from_hub

# Pull model from Hugggingface hub
policy_state_dict, cfg = pull_model_from_hub(repo_id="OpenDILabCommunity/LunarLander-v2-TD3")
# Instantiate the agent
agent = TD3Agent(
    env_id="LunarLanderContinuous-v2",
    exp_name="LunarLander-v2-TD3",
    cfg=cfg.exp_config,
    policy_state_dict=policy_state_dict
)
# Continue training
agent.train(step=5000)
# Render the new agent performance
agent.deploy(enable_save_replay=True)

Upload Model

I trained an agent and want to upload it to the Hub

from ding.bonus import TD3Agent
from huggingface_ding import push_model_to_hub

# Instantiate the agent
agent = TD3Agent(env_id="LunarLanderContinuous-v2", exp_name="LunarLander-v2-TD3")
# Train the agent
return_ = agent.train(step=int(4000000), collector_env_num=4, evaluator_env_num=4)
# Push model to huggingface hub
push_model_to_hub(
    agent=agent.best,
    env_name="OpenAI/Gym/Box2d",
    task_name="LunarLander-v2",
    algo_name="TD3",
    wandb_url=return_.wandb_url,
    github_repo_url="https://github.com/opendilab/DI-engine",
    github_doc_model_url="https://di-engine-docs.readthedocs.io/en/latest/12_policies/td3.html",
    github_doc_env_url="https://di-engine-docs.readthedocs.io/en/latest/13_envs/lunarlander.html",
    installation_guide="pip3 install DI-engine[common_env,video]",
    usage_file_by_git_clone="./td3/lunarlander_td3_deploy.py",
    usage_file_by_huggingface_ding="./td3/lunarlander_td3_download.py",
    train_file="./td3/lunarlander_td3.py",
    repo_id="OpenDILabCommunity/LunarLander-v2-TD3",
    create_repo=True,
)

API

pull_model_from_hub

Arguments:

• agent (:obj:object): the agent instance to be uploaded.
• env_name (:obj:str): the name of environment in which the task is contained.
• task_name (:obj:str): the name of task for which the agent is designed.
• algo_name (:obj:str): the policy name of the agent.
• wandb_url (:obj:str): the wandb url of the trainning process.
• repo_id (:obj:str): the repository id of Huggingface Hub where the model is stored.
• usage_file_by_git_clone (:obj:str): the path of a python file which describes ways to use the OpenDILab/DI-engine model that git cloned from Huggingface Hub.
• usage_file_by_huggingface_ding (:obj:str): the path of a python file which describes ways to use the OpenDILab/DI-engine model that downloaded by huggingface ding.
• train_file (:obj:str): the path of a python file which describes how this model is trained.
• github_repo_url (:obj:str): the github url of the DI-engine repository which the model is used.
• github_doc_model_url (:obj:str): the github or document url of the model used.
• github_doc_env_url (:obj:str): the github or document url of the environment.
• model_description (:obj:str): a paragraph of description to the model.
• installation_guide (:obj:str): the guide for installation.
• create_repo (:obj:bool): whether to create a new repository in huggingface hub.

push_model_to_hub

Arguments:

• repo_id (:obj:str): the repository id of Huggingface Hub where the model is stored.