MEGA Hub
New
Docs/Models

Downloading Models#

Download a complete model revision or the specific files your application needs. Use a tag or commit ID for reproducible automation.

BASH
mega models download OWNER/MODEL --revision v1.0 --local-dir ./modelmega download OWNER/MODEL config.json tokenizer.json --revision v1.0 --local-dir ./model

Compatible Hugging Face clients can target MEGA with HF_ENDPOINT; see Hugging Face Compatibility.

For a conventional model layout, the same endpoint can be used with huggingface_hub and Transformers:

PYTHON
from huggingface_hub import hf_hub_downloadfrom transformers import AutoConfig config_path = hf_hub_download(    repo_id="OWNER/MODEL",    filename="config.json",    revision="v1.0",    endpoint="https://mega.tensorplay.cn",    token="YOUR_MEGA_TOKEN",)config = AutoConfig.from_pretrained(    "OWNER/MODEL",    revision="v1.0",    token="YOUR_MEGA_TOKEN",)print(config_path, config.model_type)

The client downloads files; it does not convert a MEGA-native model.mega.index.json release into a Transformers checkpoint. Open that format with Megatensors instead.

For a gated or private model, authenticate with a token that has repo:read and has access to the repository. Review the model card and license before loading a model into a production workflow.