caching.md

Caching model downloads from TF Hub

Summary

The tensorflow_hub library caches models on the filesystem when they have been downloaded from tfhub.dev (or other hosting sites) and decompressed. The download location defaults to a local temporary directory but can be customized by setting the environment variable TFHUB_CACHE_DIR (recommended) or passing the command-line flag --tfhub_cache_dir. When using a persistent location, be aware that there is no automatic cleanup.

The calls to tensorflow_hub functions in the actual Python code can and should continue to use the canonical tfhub.dev URLs of models, which are portable across systems and navigable for documentation.

Specific execution environments

If and how the default TFHUB_CACHE_DIR needs changing depends on the execution environment.

Running locally on a workstation

For users running TensorFlow programs on their workstation, it should just work in most cases to keep using the default location /tmp/tfhub_modules, or whatever it is that Python returns for os.path.join(tempfile.gettempdir(), "tfhub_modules").

Users who prefer persistent caching across system reboots can instead set TFHUB_CACHE_DIR to a location in their home directory. For example, a user of the bash shell on a Linux system can add a line like the following to ~/.bashrc

export TFHUB_CACHE_DIR=$HOME/.cache/tfhub_modules

...restart the shell, and then this location will be used.

Running on TPU in Colab notebooks

For running TensorFlow on CPU and GPU from within a Colab notebook, using the default local cache location should just work.

Running on TPU delegates to another machine that does not have access to the default local cache location. Users with their own Google Cloud Storage (GCS) bucket can work around this by setting a directory in that bucket as the cache location with code like

import os
os.environ["TFHUB_CACHE_DIR"] = "gs://my-bucket/tfhub-modules-cache"

...before calling the tensorflow_hub library.