runyard (Python client)
See the main README.
Optional extras: images (numpy, Pillow), system (GPU, CPU and RAM usage), video (imageio-ffmpeg: frames
become an H.264 mp4 instead of a GIF) and tensorboard (for runyard watch DIR, which sends TensorBoard event
files: one run per directory, --once to import and exit). runyard --help lists the commands.
With RUNYARD_SERVER and RUNYARD_API_KEY set and no wandb package installed, import wandb is runyard's
bridge, so tools that log to Weights & Biases (ai-toolkit, kohya sd-scripts, report_to="wandb") write to
runyard unchanged; RUNYARD_WANDB=1 makes the bridge win over an installed wandb, RUNYARD_WANDB=0 turns it off,
and import runyard.wandb as wandb is the same module by name.
Framework adapters
runyard.integrations.<module> imports its framework, which you install yourself. Keyword arguments other than
project and name go to runyard.init (tags, notes, system, ...).
| Module | One line | Run, config, metrics |
|---|---|---|
hf |
Trainer(..., callbacks=[RunyardCallback()]) |
per training; TrainingArguments and model config; train/, eval/ by global step |
accelerate |
Accelerator(log_with=[RunyardTracker(project=...)]) |
per init_trackers; its config; accelerator.log() |
lightning |
Trainer(logger=RunyardLogger(project=...)) |
per Trainer call; hyperparameters; logged metrics by global step |
keras |
model.fit(..., callbacks=[RunyardCallback(project=...)]) |
per fit; fit params and parameter count; train/, val/ by epoch, batch/ with log_every_n_batches |
xgboost |
xgb.train(..., callbacks=[RunyardCallback(project=...)]) |
per training; booster training params; <eval name>/<metric> by iteration |
lightgbm |
lgb.train(..., callbacks=[runyard_callback(project=...)]) |
per training; params; <valid name>/<metric> by iteration |
sklearn |
autolog(project=...); autolog(disable=True) undoes it |
per outer fit(), named after the class; get_params(deep=False); fit_seconds, train/score |
torchtune |
metric_logger: {_component_: runyard.integrations.torchtune.RunyardLogger, project: ...} |
per recipe; resolved recipe config; the recipe's metrics by step |
On macOS, xgboost and lightgbm need brew install libomp, and they crash if torch was imported first: import them
before torch.
1 maintainer
Metadata
1 maintainer
Release files for runyard 0.7.1
For a detailed explanation of source distributions (sdists) and built distributions (wheels), please see the package formats documentation.
Source distribution (sdist)
| File | Size | Uploaded | Provenance | Actions |
|---|---|---|---|---|
| runyard-0.7.1.tar.gz | 47.3 kB | With provenance | Details for runyard-0.7.1.tar.gz |
Built distribution (wheel)
| File | Interpreter | ABI | Platform | Actions Reset |
|---|---|---|---|---|
| runyard-0.7.1-py3-none-any.whl | Python 3 | none | any | Details for runyard-0.7.1-py3-none-any.whl |
Total release size: 89.2 kB
Release files / runyard-0.7.1.tar.gz
| Download URL | runyard-0.7.1.tar.gz |
|---|---|
| Size | 47.3 kB |
| Tags | Source |
|
SHA-256 checksum How to use checksums |
bac6b8a3b87636574ac165ada836df5955b5533971a7dd088d489b881cc4078e
|
|
BLAKE2b-256 checksum How to use checksums |
908920f6c35b51f58340966d1118f7f38b705e529a92fd411cd86b3f09e258c0
|
| Upload date | |
|
Uploaded using Trusted Publishing? What is trusted publishing? |
Yes |
| Uploaded via |
twine/7.0.0 CPython/3.13.16
|
Provenance
Provenance describes where a file came from. On PyPI, provenance is shared via attestations, which provide a verifiable record of the build or publishing details. View details, limitations and caveats.
PyPI Publish Attestation
PyPI verified that this artifact, at this checksum, originated from the publisher listed below.
Signed by GitHub Actions, verified by PyPI on Nov 9, 2026.
Transparency logRelease files / runyard-0.7.1-py3-none-any.whl
| Download URL | runyard-0.7.1-py3-none-any.whl |
|---|---|
| Size | 41.9 kB |
| Tags | Python 3 |
|
SHA-256 checksum How to use checksums |
d168f87c3784362a75ddfd7fe6479e5c7d85a77c65f432bb50c859700067f96f
|
|
BLAKE2b-256 checksum How to use checksums |
1b28dc39184a4daa34594b05c911eeb924205d8274340c3a1d0cdb47ad1ddeec
|
| Upload date | |
|
Uploaded using Trusted Publishing? What is trusted publishing? |
Yes |
| Uploaded via |
twine/7.0.0 CPython/3.13.16
|
Provenance
Provenance describes where a file came from. On PyPI, provenance is shared via attestations, which provide a verifiable record of the build or publishing details. View details, limitations and caveats.
PyPI Publish Attestation
PyPI verified that this artifact, at this checksum, originated from the publisher listed below.
Signed by GitHub Actions, verified by PyPI on Nov 9, 2026.
Transparency logSecurity for runyard 0.7.1
Provenance
Provenance describes where a file came from. On PyPI, provenance is shared via attestations, which provide a verifiable record of the build or publishing details. View details, limitations and caveats.
Full provenance
Both files in this release have been uploaded with provenance metadata.
Attestation for runyard-0.7.1.tar.gz
PyPI Publish Attestation
PyPI verified that this artifact, at this checksum, originated from the publisher listed below.
Signed by GitHub Actions, verified by PyPI on Nov 9, 2026.
Transparency logA note on package security
PyPI is a repository for distributing packages, not a service for auditing them. While third-party security researchers and our staff do their best to detect and remove malicious packages, PyPI cannot guarantee the quality, security or suitability of any given package.
To learn more about protecting yourself from malicious actors, visit our package security documentation.
