Posit AI Weblog: Hugging Face Integrations

We’re pleased to announce the primary releases of hfhub and tok are actually on CRAN.
hfhub is an R interface to Hugging Face Hub, permitting customers to obtain and cache information
from Hugging Face Hub whereas tok implements R bindings for the Hugging Face tokenizers
library.

Hugging Face quickly turned the platform to construct, share and collaborate on
deep studying purposes and we hope these integrations will assist R customers to
get began utilizing Hugging Face instruments in addition to constructing novel purposes.

We even have beforehand introduced the safetensors
bundle permitting to learn and write information within the safetensors format.

hfhub

hfhub is an R interface to the Hugging Face Hub. hfhub at the moment implements a single
performance: downloading information from Hub repositories. Mannequin Hub repositories are
primarily used to retailer pre-trained mannequin weights along with some other metadata
essential to load the mannequin, such because the hyperparameters configurations and the
tokenizer vocabulary.

Downloaded information are ached utilizing the identical structure because the Python library, thus cached
information may be shared between the R and Python implementation, for simpler and faster
switching between languages.

We already use hfhub within the minhub bundle and
within the ‘GPT-2 from scratch with torch’ weblog put up to
obtain pre-trained weights from Hugging Face Hub.

You should use hub_download() to obtain any file from a Hugging Face Hub repository
by specifying the repository id and the trail to file that you just need to obtain.
If the file is already within the cache, then the perform returns the file path imediately,
in any other case the file is downloaded, cached after which the entry path is returned.

path <- hfhub::hub_download("gpt2", "mannequin.safetensors")
path
#> /Customers/dfalbel/.cache/huggingface/hub/models--gpt2/snapshots/11c5a3d5811f50298f278a704980280950aedb10/mannequin.safetensors

Tokenizers are accountable for changing uncooked textual content into the sequence of integers that
is usually used because the enter for NLP fashions, making them an essential part of the
NLP pipelines. If you’d like a better degree overview of NLP pipelines, you would possibly need to learn
our earlier weblog put up ‘What are Massive Language Fashions? What are they not?’.

When utilizing a pre-trained mannequin (each for inference or for wonderful tuning) it’s very
necessary that you just use the very same tokenization course of that has been used throughout
coaching, and the Hugging Face crew has achieved an incredible job ensuring that its algorithms
match the tokenization methods used most LLM’s.

tok supplies R bindings to the 🤗 tokenizers library. The tokenizers library is itself
carried out in Rust for efficiency and our bindings use the extendr mission
to assist interfacing with R. Utilizing tok we are able to tokenize textual content the very same manner most
NLP fashions do, making it simpler to load pre-trained fashions in R in addition to sharing
our fashions with the broader NLP neighborhood.

tok may be put in from CRAN, and at the moment it’s utilization is restricted to loading
tokenizers vocabularies from information. For instance, you may load the tokenizer for the GPT2
mannequin with:

tokenizer <- tok::tokenizer$from_pretrained("gpt2")
ids <- tokenizer$encode("Good day world! You should use tokenizers from R")$ids
ids
#> [1] 15496   995     0   921   460   779 11241 11341   422   371
tokenizer$decode(ids)
#> [1] "Good day world! You should use tokenizers from R"

Keep in mind which you can already host
Shiny (for R and Python) on Hugging Face Areas. For example, we’ve got constructed a Shiny
app that makes use of:

• torch to implement GPT-NeoX (the neural community structure of StableLM – the mannequin used for chatting)
• hfhub to obtain and cache pre-trained weights from the StableLM repository
• tok to tokenize and pre-process textual content as enter for the torch mannequin. tok additionally makes use of hfhub to obtain the tokenizer’s vocabulary.

The app is hosted at on this Area.
It at the moment runs on CPU, however you may simply swap the the Docker picture in order for you
to run it on a GPU for sooner inference.

The app supply code can also be open-source and may be discovered within the Areas file tab.

Trying ahead

It’s the very early days of hfhub and tok and there’s nonetheless quite a lot of work to do
and performance to implement. We hope to get neighborhood assist to prioritize work,
thus, if there’s a function that you’re lacking, please open a problem within the
GitHub repositories.

Reuse

Textual content and figures are licensed underneath Artistic Commons Attribution CC BY 4.0. The figures which have been reused from different sources do not fall underneath this license and may be acknowledged by a be aware of their caption: “Determine from …”.

Quotation

For attribution, please cite this work as

Falbel (2023, July 12). Posit AI Weblog: Hugging Face Integrations. Retrieved from https://blogs.rstudio.com/tensorflow/posts/2023-07-12-hugging-face-integrations/

BibTeX quotation

@misc{hugging-face-integrations,
  writer = {Falbel, Daniel},
  title = {Posit AI Weblog: Hugging Face Integrations},
  url = {https://blogs.rstudio.com/tensorflow/posts/2023-07-12-hugging-face-integrations/},
  12 months = {2023}
}