speechbrain.pretrained.fetching module

Downloads or otherwise fetches pretrained models

  • Aku Rouhe 2021

  • Samuele Cornell 2021




Ensures you have a local copy of the file, returns its path


speechbrain.pretrained.fetching.fetch(filename, source, savedir='./pretrained_model_checkpoints', overwrite=False, save_filename=None, use_auth_token=False)[source]

Ensures you have a local copy of the file, returns its path

In case the source is an external location, downloads the file. In case the source is already accessible on the filesystem, creates a symlink in the savedir. Thus, the side effects of this function always look similar: savedir/save_filename can be used to access the file. And save_filename defaults to the filename arg.

  • filename (str) – Name of the file including extensions.

  • source (str) – Where to look for the file. This is interpreted in special ways: First, if the source begins with “http://” or “https://”, it is interpreted as a web address and the file is downloaded. Second, if the source is a valid directory path, a symlink is created to the file. Otherwise, the source is interpreted as a Huggingface model hub ID, and the file is downloaded from there.

  • savedir (str) – Path where to save downloads/symlinks.

  • overwrite (bool) – If True, always overwrite existing savedir/filename file and download or recreate the link. If False (as by default), if savedir/filename exists, assume it is correct and don’t download/relink. Note that Huggingface local cache is always used - with overwrite=True we just relink from the local cache.

  • save_filename (str) – The filename to use for saving this file. Defaults to filename if not given.

  • use_auth_token (bool (default: False)) – If true Hugginface’s auth_token will be used to load private models from the HuggingFace Hub, default is False because majority of models are public.


Path to file on local file system.

Return type



ValueError – If file is not found