modelscope-hub 0.0.5

The official Python client to connect with ModelScope Hub.

The official Python SDK & CLI for ModelScope Hub — download, upload, and manage AI assets from one unified interface.

modelscope.cn | modelscope.ai

Why modelscope-hub?

modelscope-hub connects your code to the ModelScope ecosystem — models, datasets, Studio spaces, skills, and MCP servers — through a single HubApi class or the ms CLI.

• Unified repo interface — one set of methods for models, datasets, studios, skills, and MCP servers
• OpenAPI-first — built on the ModelScope OpenAPI surface with transparent legacy fallback
• Production-grade downloads — HTTP Range resume, parallel range download for large files, per-file retry with backoff, SHA256 integrity checks, file lock for multiprocess safety, offline mode, progress callbacks, and intra-cloud acceleration
• Full lifecycle CLI — download, upload, deploy, manage secrets, inspect cache — all from the terminal
• Deep ecosystem integration — seamless access to 100K+ models and datasets on ModelScope Hub; works with the modelscope training framework, Studio deployment platform, and MCP server infrastructure

---

Installation

pip install modelscope-hub

Requires Python 3.10+. Lightweight — only requests, tqdm, filelock, urllib3.

---

Quick Start

Authenticate

ms login
# or pass a token directly
ms login --token $MODELSCOPE_API_TOKEN

Get your token at modelscope.cn/my/access/token or modelscope.ai/my/access/token.

from modelscope_hub import HubApi

api = HubApi(token="your-token")
user = api.whoami()
print(user.username)

Download

# Full snapshot
ms download Qwen/Qwen3-0.6B

# Single file
ms download Qwen/Qwen3-0.6B config.json

# With filters
ms download Qwen/Qwen3-0.6B --include "*.safetensors" --exclude "*.bin"

# Directly into a local directory (bypasses cache)
ms download Qwen/Qwen3-0.6B --local-dir ./my-model

path = api.download_file("Qwen/Qwen3-0.6B", "model", "config.json")

snapshot = api.download_repo(
    "Qwen/Qwen3-0.6B", "model",
    allow_patterns=["*.safetensors", "*.json"],
    max_workers=8,
)

# Offline mode — return cached path without network access
path = api.download_file("Qwen/Qwen3-0.6B", "model", "config.json", local_files_only=True)

Upload

ms upload my-org/my-model ./weights.safetensors
ms upload my-org/my-model ./output --repo-type model --commit-message "add weights"

api.upload_file("my-org/my-model", "model", "./weights.safetensors", "weights.safetensors")
api.upload_folder("my-org/my-model", "model", "./output", path_in_repo="")

Create a Repository

ms create my-org/my-model --repo-type model --visibility private

api.create_repo("my-org/my-model", "model", visibility="private", license="apache-2.0")

Deploy a Studio

ms deploy my-org/chat-demo --repo-type studio
ms logs my-org/chat-demo --log-type run
ms stop my-org/chat-demo --repo-type studio

api.deploy_repo("my-org/chat-demo", "studio")
api.get_repo_logs("my-org/chat-demo", log_type="run")
api.stop_repo("my-org/chat-demo", "studio")

---

CLI Reference

The CLI is available as both ms and modelscope.

Global options (placed before or after the subcommand):

Option	Description
--token TOKEN	API token (overrides env and persisted token)
--endpoint URL	API endpoint (default: https://modelscope.cn)
-v, --verbose	Enable DEBUG logging (global only)
-V, --version	Print version and exit (global only)

--token and --endpoint can be placed either before or after the subcommand: ms --token xxx download ... and ms download ... --token xxx are equivalent.

ms login

Authenticate and persist your token locally.

ms login                          # interactive prompt
ms login --token $MY_TOKEN        # non-interactive

Option	Description
--token TOKEN	API token; prompted interactively if omitted

ms whoami

Show the user associated with the current token.

ms whoami
ms whoami --token $MY_TOKEN   # check a specific token without logging in

ms download

Download a single file or a full repository snapshot.

ms download Qwen/Qwen3-0.6B                                  # full snapshot
ms download Qwen/Qwen3-0.6B config.json                      # single file
ms download Qwen/Qwen3-0.6B --include "*.safetensors"         # filter by glob
ms download Qwen/Qwen3-0.6B --local-dir ./out --max-workers 8 # direct download
ms download my-org/my-data --repo-type dataset --revision v2   # dataset at tag

Argument / Option	Required	Description
repo_id	yes	Repository identifier (owner/name)
files...	no	Specific file paths; omit for full snapshot
--repo-type {model,dataset}	no	Default: model
--revision REV	no	Branch, tag, or commit hash (default: master)
--local-dir DIR	no	Download directly here (bypasses cache layout)
--cache-dir DIR	no	Override default cache directory
--include GLOB...	no	Only download matching files; repeatable
--exclude GLOB...	no	Skip matching files; repeatable
--max-workers N	no	Parallel download threads (default: 4)
--force	no	Re-download even if cached

Advanced examples

# Download multiple specific files at once
ms download Qwen/Qwen3-0.6B config.json tokenizer.json generation_config.json

# Download only safetensors, skip GGUF and bin weights
ms download Qwen/Qwen3-0.6B --include "*.safetensors" --exclude "*.bin" "*.gguf"

# Download a dataset at a specific tag into a local directory
ms download my-org/my-data --repo-type dataset --revision v2 --local-dir ./data

# Use a custom cache directory and 8 parallel threads
ms download Qwen/Qwen3-0.6B --cache-dir /data/hub-cache --max-workers 8

# Force re-download even if already cached
ms download Qwen/Qwen3-0.6B config.json --force

# Download all skills from a collection (legacy flag)
ms download --collection my-org/skill-collection

# Enable parallel range download for large files (env var)
MODELSCOPE_DOWNLOAD_PARALLELS=4 ms download Qwen/Qwen3-0.6B

# Use the modelscope.ai endpoint (global option, before subcommand)
ms --endpoint https://modelscope.ai download Qwen/Qwen3-0.6B

ms upload

Upload a file or folder to a repository.

ms upload my-org/my-model ./weights.safetensors                       # single file
ms upload my-org/my-model ./output models/ --repo-type model          # folder → subdir
ms upload my-org/my-model . --include "*.py" --commit-message "code"  # filtered folder

Argument / Option	Required	Description
repo_id	yes	Repository identifier
local_path	no	Local file or folder (default: inferred from repo name)
path_in_repo	no	Destination path inside the repo
--repo-type {model,dataset}	no	Default: model
--revision REV	no	Target branch (default: master)
--commit-message MSG	no	Commit message
--commit-description DESC	no	Extended commit description
--include GLOB...	no	Include filter for folder mode; repeatable
--exclude GLOB...	no	Exclude filter for folder mode; repeatable
--max-workers N	no	Parallel upload threads
--use-cache / --no-cache	no	Enable/disable resumable upload cache (default: on)
--disable-tqdm	no	Disable progress bars

Advanced examples

# Upload a single file with a custom commit message
ms upload my-org/my-model ./weights.safetensors --commit-message "add fp16 weights"

# Upload a folder into a subdirectory of the repo
ms upload my-org/my-model ./output models/ --repo-type model

# Upload only Python files from the current directory
ms upload my-org/my-model . --include "*.py" --commit-message "update code"

# Upload only safetensors, skip checkpoints
ms upload my-org/my-model ./output --include "*.safetensors" --exclude "*.ckpt" "*.bin"

# Upload to a dataset repo on a specific branch
ms upload my-org/my-data ./data --repo-type dataset --revision dev

# Upload with extended commit description
ms upload my-org/my-model ./weights.safetensors \
  --commit-message "v2 weights" \
  --commit-description "Retrained with extended dataset, 3 epochs, lr=2e-5"

# Resumable upload: interrupted uploads resume automatically via cache
ms upload my-org/my-model ./large-folder
# If interrupted, just re-run the same command — already uploaded files are skipped

# Disable upload cache (no resume, fresh upload every time)
ms upload my-org/my-model ./output --no-cache

# Disable progress bars (useful for CI/CD pipelines)
ms upload my-org/my-model ./output --disable-tqdm

ms create / ms info / ms list / ms delete

Repository management.

ms create my-org/my-model --repo-type model --visibility private
ms create my-org/demo --repo-type studio --sdk-type gradio
ms info my-org/my-model --repo-type model
ms list --repo-type model --owner my-org --page-size 20
ms delete my-org/my-model --repo-type model --yes

ms create options

Argument / Option	Required	Description
repo_id	yes	Repository identifier
--repo-type	yes	model, dataset, studio, or skill
--visibility	no	public, private, or internal
--license	no	SPDX license identifier (e.g. apache-2.0)
--chinese-name	no	Display name in Chinese
--description	no	Repository description
--exist-ok	no	No error if repository already exists
--sdk-type	no	Studio SDK: gradio, streamlit, docker, static
--sdk-version	no	Studio SDK version
--base-image	no	Studio base Docker image
--cover-image	no	Studio cover image URL
--hardware	no	Studio hardware spec

ms deploy / ms stop / ms logs / ms settings

Manage Studio and MCP deployments.

ms deploy my-org/chat-demo --repo-type studio
ms logs my-org/chat-demo --log-type run --keyword ERROR --page-size 50
ms settings my-org/chat-demo cpu=4 memory=8192
ms stop my-org/chat-demo --repo-type studio

Options

Command	--repo-type	Key Options
ms deploy <repo_id>	{studio,mcp} (default: studio)	—
ms stop <repo_id>	{studio,mcp} (default: studio)	—
ms logs <repo_id>	{studio} only	--log-type {run,build}, --keyword, --page, --page-size
ms settings <repo_id> key=val...	{studio,skill} (default: studio)	Key-value pairs passed to backend

Note: ms logs only supports Studio spaces. MCP server logs are not available via this command. ms settings supports Studio and Skill repos; for MCP servers use ms mcp deploy with configuration payload.

ms secret

Manage secrets for Studio spaces (studio only, --repo-type defaults to studio).

ms secret add my-org/demo API_KEY sk-xxx
ms secret list my-org/demo
ms secret update my-org/demo API_KEY sk-new
ms secret delete my-org/demo API_KEY --yes

Subcommands

Subcommand	Arguments	Description
add	repo_id key value	Add a new secret
list	repo_id	List all secret keys
update	repo_id key value	Update a secret value
delete	repo_id key [--yes]	Delete a secret

All subcommands accept --repo-type (default: studio, currently the only supported type).

ms mcp

Manage MCP (Model Context Protocol) servers.

ms mcp list --search weather --page-size 10
ms mcp info my-org/weather-mcp
ms mcp deploy my-org/weather-mcp
ms mcp undeploy my-org/weather-mcp

Subcommands

Subcommand	Arguments	Key Options
list	—	--search, --page, --page-size
info	server_id	—
deploy	server_id	—
undeploy	server_id	—

ms cache

Inspect and clean the local download cache.

ms cache scan
ms cache scan --cache-dir /data/cache
ms cache clear --repo-type model --yes
ms cache clear --repo-id my-org/old-model --repo-type model --yes

Options

Subcommand	Key Options
scan	--cache-dir DIR
clear	--repo-type, --repo-id, --cache-dir, --yes

---

SDK API Overview

All operations go through a single entry point:

from modelscope_hub import HubApi

# Connect to modelscope.cn (default)
api = HubApi(token="...")

# Or connect to modelscope.ai
api = HubApi(token="...", endpoint="https://modelscope.ai")

Full method reference

Category	Method	Description
Auth	login(token)	Persist and verify token
	logout()	Clear stored credentials
	whoami()	Get current user info
Repo	create_repo(repo_id, repo_type, ...)	Create a repository
	get_repo(repo_id, repo_type)	Get repository metadata
	list_repos(repo_type, ...)	Paginated listing
	delete_repo(repo_id, repo_type)	Delete a repository
	repo_exists(repo_id, repo_type)	Check existence
Files	upload_file(repo_id, repo_type, local, remote)	Upload a single file
	upload_folder(repo_id, repo_type, folder, ...)	Upload a directory
	download_file(repo_id, repo_type, file, ...)	Download a single file (with retry, resume, offline mode)
	download_repo(repo_id, repo_type, ...)	Download full snapshot (parallel, file lock, progress callbacks)
	list_repo_files(repo_id, repo_type)	List files in a repo
	delete_files(repo_id, repo_type, paths)	Remove files
Version	list_repo_revisions(repo_id, repo_type)	List branches and tags
	create_repo_tag(repo_id, repo_type, tag)	Create a tag
Deploy	deploy_repo(repo_id, repo_type)	Deploy Studio or MCP
	stop_repo(repo_id, repo_type)	Stop deployment
	get_repo_logs(repo_id, ...)	Fetch logs
	update_repo_settings(repo_id, repo_type, ...)	Update settings
Secrets	add_secret(repo_id, key, value)	Add a secret
	list_secrets(repo_id)	List secrets
	update_secret(repo_id, key, value)	Update a secret
	delete_secret(repo_id, key)	Delete a secret
MCP	list_mcp_servers(...)	List available MCP servers
	get_mcp_server(server_id)	Get server details
	deploy_mcp_server(server_id)	Deploy an MCP server
	undeploy_mcp_server(server_id)	Undeploy an MCP server
Cache	scan_cache(cache_dir)	Inspect local cache
	clear_cache(cache_dir, ...)	Free disk space

---

Ecosystem Integration

modelscope-hub is the hub connectivity layer for the ModelScope ecosystem:

┌────────────────────────────────────────────────┐
│              ModelScope Platform                │
│   modelscope.cn  ·  modelscope.ai              │
│                                                │
│  Models · Datasets · Studios · Skills · MCP    │
└───────────────────┬────────────────────────────┘
                    │  OpenAPI / Legacy API
                    ▼
            ┌───────────────┐
            │ modelscope-hub│  ← this library
            │  SDK  +  CLI  │
            └───┬───────┬───┘
                │       │
        ┌───────┘       └────────┐
        ▼                        ▼
  modelscope framework    your application
  (training · eval)       (inference · deploy)

• Browse & discover — search 100K+ models and datasets via list_repos / ms repo list
• Download & cache — pull model weights, tokenizer configs, or entire datasets into a managed cache or a local directory; supports offline mode via local_files_only
• Train & fine-tune — use with the modelscope framework: train locally, then push results back
• Deploy — launch a Studio space or MCP server directly from the CLI or SDK
• Automate — integrate into CI/CD pipelines with environment-variable auth and --yes flags for non-interactive operation

---

Configuration

Environment Variable	Purpose	Default
MODELSCOPE_API_TOKEN	Default API token	—
MODELSCOPE_ENDPOINT	API endpoint	https://modelscope.cn
MODELSCOPE_CACHE	Override cache directory	~/.modelscope/hub
MODELSCOPE_DOWNLOAD_PARALLELS	Parallel range-download parts for large files	1 (disabled)
MODELSCOPE_PARALLEL_DOWNLOAD_THRESHOLD_MB	File size threshold (MB) to trigger parallel range download	500
DOWNLOAD_RETRY_TIMES	Per-file download retry count	5
DOWNLOAD_TIMEOUT	Per-request download timeout (seconds)	60
MODELSCOPE_HUB_FILE_LOCK	Enable file lock for multiprocess download safety	true
INTRA_CLOUD_ACCELERATION	Enable Alibaba cloud intra-cloud download acceleration	true

Token is persisted locally after ms login and auto-loaded in subsequent sessions.

---

Backward Compatibility

modelscope-hub provides a compatibility layer for code written against the old modelscope.hub API surface. The old SDK can delegate directly to modelscope_hub.compat:

from modelscope_hub.compat import snapshot_download, model_file_download
from modelscope_hub.compat import LegacyHubApi as HubApi

All legacy parameter names (allow_file_pattern, ignore_file_pattern, cookies, etc.) are accepted and mapped to the new implementation.

---

Development

git clone https://github.com/modelscope/modelscope_hub.git
cd modelscope_hub
make install    # pip install -e ".[dev]"
make test       # unit tests (no network)
make lint       # ruff check
make typecheck  # mypy

See make help for all available targets.

---

License

Apache 2.0 — see LICENSE.

---

Citation

@Misc{modelscope-hub,
  title = {modelscope-hub: The official Python client to connect with ModelScope Hub.},
  author = {The ModelScope Team},
  howpublished = {\url{https://github.com/modelscope/modelscope_hub}},
  year = {2026}
}