Hugging Face's logo

Hub Python Library documentation

TensorBoard logger

You are viewing v0.16.3 version. A newer version v0.26.2 is available.
Hugging Face's logo
Join the Hugging Face community

and get access to the augmented documentation experience

to get started

TensorBoard logger

TensorBoard is a visualization toolkit for machine learning experimentation. TensorBoard allows tracking and visualizing metrics such as loss and accuracy, visualizing the model graph, viewing histograms, displaying images and much more. TensorBoard is well integrated with the Hugging Face Hub. The Hub automatically detects TensorBoard traces (such as tfevents) when pushed to the Hub which starts an instance to visualize them. To get more information about TensorBoard integration on the Hub, check out this guide.

To benefit from this integration, huggingface_hub provides a custom logger to push logs to the Hub. It works as a drop-in replacement for SummaryWriter with no extra code needed. Traces are still saved locally and a background job push them to the Hub at regular interval.

HFSummaryWriter

class huggingface_hub.HFSummaryWriter

< >

( *args **kwargs )

Parameters

  • repo_id (str) — The id of the repo to which the logs will be pushed.
  • logdir (str, optional) — The directory where the logs will be written. If not specified, a local directory will be created by the underlying SummaryWriter object.
  • commit_every (int or float, optional) — The frequency (in minutes) at which the logs will be pushed to the Hub. Defaults to 5 minutes.
  • repo_type (str, optional) — The type of the repo to which the logs will be pushed. Defaults to “model”.
  • repo_revision (str, optional) — The revision of the repo to which the logs will be pushed. Defaults to “main”.
  • repo_private (bool, optional) — Whether to create a private repo or not. Defaults to False. This argument is ignored if the repo already exists.
  • path_in_repo (str, optional) — The path to the folder in the repo where the logs will be pushed. Defaults to “tensorboard/“.
  • repo_allow_patterns (List[str] or str, optional) — A list of patterns to include in the upload. Defaults to "*.tfevents.*". Check out the upload guide for more details.
  • repo_ignore_patterns (List[str] or str, optional) — A list of patterns to exclude in the upload. Check out the upload guide for more details.
  • token (str, optional) — Authentication token. Will default to the stored token. See https://huggingface.co/settings/token for more details kwargs — Additional keyword arguments passed to SummaryWriter.

Wrapper around the tensorboard’s SummaryWriter to push training logs to the Hub.

Data is logged locally and then pushed to the Hub asynchronously. Pushing data to the Hub is done in a separate thread to avoid blocking the training script. In particular, if the upload fails for any reason (e.g. a connection issue), the main script will not be interrupted. Data is automatically pushed to the Hub every commit_every minutes (default to every 5 minutes).

HFSummaryWriter is experimental. Its API is subject to change in the future without prior notice.

Examples:

>>> from huggingface_hub import HFSummaryWriter

# Logs are automatically pushed every 15 minutes
>>> logger = HFSummaryWriter(repo_id="test_hf_logger", commit_every=15)
>>> logger.add_scalar("a", 1)
>>> logger.add_scalar("b", 2)
...

# You can also trigger a push manually
>>> logger.scheduler.trigger()
>>> from huggingface_hub import HFSummaryWriter

# Logs are automatically pushed every 5 minutes (default) + when exiting the context manager
>>> with HFSummaryWriter(repo_id="test_hf_logger") as logger:
...     logger.add_scalar("a", 1)
...     logger.add_scalar("b", 2)
←Space runtime Webhooks server→