Skip to main content

Starwhale Evaluation

Starwhale Evaluation helps you pick out the best-performed model.

It is easy to find the difference, identify the best possible model, and then deploy it.

Starwhale Dataset

Starwhale Dataset supports you in managing your machine learning datasets.

Conveniently get insights from datasets and label visualization.
 # create a new dataset named mnist, and add a row into the dataset
ds = dataset("mnist")
ds.exists() # return False, "mnist" dataset is not existing.
ds.append({"img": Image(), "label": 1})

Starwhale Model

Starwhale Model is the standard model format used in model delivery.

Starwhale model can be stripped of redundant information to get a smaller package, supporting collaboration with the production team without showing python inference code to avoid jeopardizing.
root@susan:~/starwhale# swcli dataset info mnist/version/latest
╭─ Starwhale Instance ────────────────────────────────────────────────────────────────────────────────────────────────────────╮
│ ⭐ local (local) 🐳 🤡root@normal │
─────────────────────────────────────────────────────── Inspect Details ───────────────────────────────────────────────────────
'uri': 'local/project/self/dataset/mnist/version/latest',
'project': 'self',
'name': 'mnist',
'snapshot_workdir': '/root/.starwhale/self/dataset/mnist/33/33tycn2mku346m6gttwcucznbsr4vltlbwh3sxjo.swds',
'bundle_path': '/root/.starwhale/self/dataset/mnist/33/33tycn2mku346m6gttwcucznbsr4vltlbwh3sxjo.swds',
'version': '33tycn2mku346m6gttwcucznbsr4vltlbwh3sxjo',
'config': {
'build': {
'os': 'Linux',
'sw_version': '0.4.0'
'created_at': '2023-02-17 03:52:22 CST',
'dataset_attr': {
'alignment_size': 1024,
'data_mime_type': 'x/grayscale',
'volume_size': 4194304
'dataset_byte_size': 9920000,
'dataset_summary': {
'data_byte_size': 9920000,
'increased_rows': 10000,
'rows': 10000,
'unchanged_rows': 0
'handler': 'DatasetProcessExecutor',
'signature': [
'version': '33tycn2mku346m6gttwcucznbsr4vltlbwh3sxjo'
'tags': [

Starwhale Runtime

Starwhale Runtime supports mounts of software runtime solutions, such as Conda, Pip, and docker images.

Runtime describes software dependencies to "run" a model, and can be synchronized on different machines and switch between runtimes.

Experiment coming soon

Track and log your experiments to see how different parameters affect.

Try today

If you want more details about Starwhale, click the "GET STARTED" button.


If you want more information or leave a message to us, click the "CONTACT US" button.CONTACT US