Posit AI Weblog: safetensors 0.1.0
safetensors is a brand new, easy, quick, and secure file format for storing tensors. The design of the file format and its authentic implementation are being led
by Hugging Face, and it’s getting largely adopted of their widespread ‘transformers’ framework. The safetensors R bundle is a pure-R implementation, permitting to each learn and write safetensor recordsdata.
The preliminary model (0.1.0) of safetensors is now on CRAN.
Motivation
The primary motivation for safetensors within the Python group is safety. As famous
within the official documentation:
The primary rationale for this crate is to take away the necessity to use pickle on PyTorch which is utilized by default.
Pickle is taken into account an unsafe format, because the motion of loading a Pickle file can
set off the execution of arbitrary code. This has by no means been a priority for torch
for R customers, because the Pickle parser that’s included in LibTorch solely helps a subset
of the Pickle format, which doesn’t embrace executing code.
Nonetheless, the file format has further benefits over different generally used codecs, together with:
-
Help for lazy loading: You possibly can select to learn a subset of the tensors saved within the file.
-
Zero copy: Studying the file doesn’t require extra reminiscence than the file itself.
(Technically the present R implementation does makes a single copy, however that may
be optimized out if we actually want it sooner or later). -
Easy: Implementing the file format is easy, and doesn’t require complicated dependencies.
Which means that it’s a very good format for exchanging tensors between ML frameworks and
between completely different programming languages. As an illustration, you’ll be able to write a safetensors file
in R and cargo it in Python, and vice-versa.
There are further benefits in comparison with different file codecs widespread on this house, and
you’ll be able to see a comparability desk here.
Format
The safetensors format is described within the determine beneath. It’s principally a header file
containing some metadata, adopted by uncooked tensor buffers.
Primary utilization
safetensors might be put in from CRAN utilizing:
install.packages("safetensors")We can then write any named list of torch tensors:
library(torch)
library(safetensors)
tensors <- list(
x = torch_randn(10, 10),
y = torch_ones(10, 10)
)
str(tensors)
#> List of 2
#> $ x:Float [1:10, 1:10]
#> $ y:Float [1:10, 1:10]
tmp <- tempfile()
safe_save_file(tensors, tmp)It’s possible to pass additional metadata to the saved file by providing a metadata
parameter containing a named list.
Reading safetensors files is handled by safe_load_file, and it returns the named
list of tensors along with the metadata attribute containing the parsed file header.
tensors <- safe_load_file(tmp)
str(tensors)
#> List of 2
#> $ x:Float [1:10, 1:10]
#> $ y:Float [1:10, 1:10]
#> - attr(*, "metadata")=List of 2
#> ..$ x:List of 3
#> .. ..$ shape : int [1:2] 10 10
#> .. ..$ dtype : chr "F32"
#> .. ..$ data_offsets: int [1:2] 0 400
#> ..$ y:List of 3
#> .. ..$ shape : int [1:2] 10 10
#> .. ..$ dtype : chr "F32"
#> .. ..$ data_offsets: int [1:2] 400 800
#> - attr(*, "max_offset")= int 929Currently, safetensors only supports writing torch tensors, but we plan to add
support for writing plain R arrays and tensorflow tensors in the future.
Future directions
The next version of torch will use safetensors as its serialization format,
meaning that when calling torch_save() on a model, list of tensors, or other
types of objects supported by torch_save, you will get a valid safetensors file.
This is an improvement over the previous implementation because:
-
It’s much faster. More than 10x for medium sized models. Could be even more for large files.
This also improves the performance of parallel dataloaders by ~30%. -
It enhances cross-language and cross-framework compatibility. You can train your model
in R and use it in Python (and vice-versa), or train your model in tensorflow and run it
with torch.
If you want to try it out, you can install the development version of torch with:
remotes::install_github("mlverse/torch")Photo by Nick Fewings on Unsplash
Reuse
Textual content and figures are licensed below Inventive Commons Attribution CC BY 4.0. The figures which were reused from different sources do not fall below this license and might be acknowledged by a observe of their caption: “Determine from …”.
Quotation
For attribution, please cite this work as
Falbel (2023, June 15). Posit AI Weblog: safetensors 0.1.0. Retrieved from https://blogs.rstudio.com/tensorflow/posts/2023-06-15-safetensors/
BibTeX quotation
@misc{safetensors,
creator = {Falbel, Daniel},
title = {Posit AI Weblog: safetensors 0.1.0},
url = {https://blogs.rstudio.com/tensorflow/posts/2023-06-15-safetensors/},
12 months = {2023}
}
