MEGA Hub
New
Docs/Getting started

Hugging Face Compatibility#

MEGA supports core Hugging Face Hub workflows for models, datasets, Spaces, and Storage Buckets. Existing hf, huggingface_hub, Transformers, Diffusers, and Datasets workflows can target MEGA by changing the endpoint and token.

Configure the endpoint#

Set the endpoint before starting Python so compatible libraries read it during import:

BASH
export HF_ENDPOINT=https://mega.tensorplay.cnexport HF_TOKEN=YOUR_MEGA_TOKEN hf auth whoami

Set HF_ENDPOINT to the MEGA origin, without an /api suffix. Compatible clients add their normal API and resolver paths themselves. Keep the endpoint and token in the process environment rather than in a repository card or source file.

Use repo:read for private downloads, repo:write for uploads, and repo:delete for deletion. hf auth login --token "$HF_TOKEN" is also supported when a persisted credential is appropriate.

Compatibility at a glance#

The compatibility layer covers the repository workflows most commonly used by the Hugging Face client ecosystem:

Workflow Compatible entry point MEGA-native alternative
Identity and repository metadata HfApi.whoami, model_info, dataset_info, list_models, and list_datasets mega repos, mega models, and mega datasets
File upload and download hf upload, hf download, upload_folder, and snapshot_download mega upload, mega download, and mega snapshot
Branches, tags, copy, and visibility hf repos and HfApi repository methods mega repos branch, mega repos tag, and mega repos settings
Model loading Transformers and other libraries built on huggingface_hub megatensors for .mega artifacts
Dataset loading and derived Parquet Datasets, Dataset Viewer metadata, /api/datasets/{id}/parquet, and range-capable Parquet shards Data Studio, mega snapshot, or resolver URLs
Mutable working data hf buckets and compatible Bucket APIs mega buckets and mega://buckets/...

Compatibility changes the transport and repository endpoint. It does not turn a MEGA repository into a Hugging Face product or grant permissions that the MEGA account, organization, or token does not already have.

Use the hf CLI#

The standard repository lifecycle works against the configured endpoint:

BASH
hf repos create OWNER/demo --type model --exist-okhf upload OWNER/demo ./release . --commit-message "Publish release"hf upload OWNER/demo ./release . --commit-description "Release notes in Markdown"hf upload OWNER/demo ./release . --delete 'stale/**'mega upload OWNER/demo ./release . --create-pr --commit-message "Propose release" \  --commit-description "Explain the proposed change"mega upload OWNER/demo ./watch --every 5hf download OWNER/demo config.json --local-dir ./downloadhf repos duplicate OWNER/demo OWNER/demo-copy --type model hf repos branch create OWNER/demo nexthf repos tag create OWNER/demo v1.0 --revision main hf buckets create OWNER/artifacts --privatehf buckets sync ./artifacts hf://buckets/OWNER/artifacts

Pass --type dataset or --type space to typed commands. Repository visibility, branches, tags, files, and Bucket operations use the same MEGA permission checks as native clients.

--commit-description is stored as the optional Markdown body of the immutable commit and is returned by MEGA's native commit-history API. It is separate from the Git commit subject used by --commit-message. Periodic mega upload --every intentionally rejects a fixed description because every generated commit needs its own description.

Use Python libraries#

huggingface_hub accepts either HF_ENDPOINT or an explicit endpoint:

PYTHON
from huggingface_hub import HfApi, snapshot_download api = HfApi(endpoint="https://mega.tensorplay.cn", token="YOUR_MEGA_TOKEN")info = api.model_info("OWNER/demo")root = snapshot_download(    "OWNER/demo",    revision="main",    endpoint="https://mega.tensorplay.cn",    token="YOUR_MEGA_TOKEN",)

Libraries built on huggingface_hub can use the same configured endpoint:

PYTHON
from transformers import AutoConfig config = AutoConfig.from_pretrained("OWNER/demo")

The same endpoint works for supported Datasets workflows:

PYTHON
from datasets import load_dataset dataset = load_dataset(    "OWNER/demo-dataset",    data_files="data.jsonl",    split="train",    token="YOUR_MEGA_TOKEN",)

Pin production downloads to a commit or tag instead of main. For a MEGA-native .mega release, use the Megatensors runtime after downloading the pinned files; Transformers does not load the .mega format automatically.

Supported boundary#

The compatibility layer covers repository create, inspect, list, move, duplicate, visibility, branches, tags, commits, file trees, and resolver downloads; it also covers the matching Bucket lifecycle and sync workflows. Repository-card metadata is parsed in the Hugging Face YAML format.

MEGA does not emulate every Hugging Face product. The hf compatibility write endpoint does not accept implicit pull-request commits; use native mega upload --create-pr, which creates a branch, performs the atomic upload, then opens a native pull request. Dataset revisions are automatically converted for MEGA's Data Studio, and compatible clients can read the generated Parquet map, shards, first rows, schema, size, statistics, and Croissant metadata. Server-side Dataset Viewer rows, search, and filter are not emulated; Data Studio provides interactive browsing over the indexed Parquet view instead. Bucket S3 access and managed Space volumes are available through MEGA's native public interfaces; see Bucket S3 Gateway and Space Storage. Use Hub API and the live OpenAPI Explorer to check the current public contract.

The compatibility boundary also has finite request sizes: one compatible repository snapshot can contain at most 10,000 files and one compatibility commit request can contain at most 32 MiB of commit data. Split larger changes or use the resumable large-file workflow.

For MEGA-native commands and SDKs, see the CLI guide and Python SDK.