burn

Using Burn means having your models optimized on any backend. When possible, we provide a way to
automatically and dynamically create custom kernels that minimize data relocation between different
memory spaces, extremely useful when moving memory is the bottleneck.

As an example, you could write your own GELU activation function with the high level tensor api (see
Rust code snippet below).

fn gelu_custom<B: Backend, const D: usize>(x: Tensor<B, D>) -> Tensor<B, D> {
    let x = x.clone() * ((x / SQRT_2).erf() + 1);
    x / 2
}

Then, at runtime, a custom low-level kernel will be automatically created for your specific
implementation and will rival a handcrafted GPU implementation. The kernel consists of about 60
lines of WGSL WebGPU Shading Language,
an extremely verbose lower level shader language you probably don\’t want to program your deep
learning models in!

For first-party backends, an asynchronous execution style
is used, which allows to perform various optimizations, such as the previously mentioned automatic
kernel fusion.

Asynchronous execution also ensures that the normal execution of the framework does not block the
model computations, which implies that the framework overhead won\’t impact the speed of execution
significantly. Conversely, the intense computations in the model do not interfere with the
responsiveness of the framework. For more information about our asynchronous backends, see
this blog post.

Burn emphasizes thread safety by leveraging the
ownership system of Rust.
With Burn, each module is the owner of its weights. It is therefore possible to send a module to
another thread for computing the gradients, then send the gradients to the main thread that can
aggregate them, and voilà, you get multi-device training.

This is a very different approach from what PyTorch does, where backpropagation actually mutates the
grad attribute of each tensor parameter. This is not a thread-safe operation and therefore
requires lower level synchronization primitives, see
distributed training for reference. Note that
this is still very fast, but not compatible across different backends and quite hard to implement.

One of the main roles of a deep learning framework is to reduce the amount of memory necessary to
run models. The naive way of handling memory is that each tensor has its own memory space, which is
allocated when the tensor is created then deallocated as the tensor gets out of scope. However,
allocating and deallocating data is very costly, so a memory pool is often required to achieve good
throughput. Burn offers an infrastructure that allows for easily creating and selecting memory
management strategies for backends. For more details on memory management in Burn, see
this blog post.

Another very important memory optimization of Burn is that we keep track of when a tensor can be
mutated in-place just by using the ownership system well. Even though it is a rather small memory
optimization on its own, it adds up considerably when training or running inference with larger
models and contributes to reduce the memory usage even more. For more information, see
this blog post about tensor handling.

A good deep learning framework should ensure that models run smoothly on all hardware. However, not
all hardware share the same behavior in terms of execution speed. For instance, a matrix
multiplication kernel can be launched with many different parameters, which are highly sensitive to
the size of the matrices and the hardware. Using the wrong configuration could reduce the speed of
execution by a large factor (10 times or even more in extreme cases), so choosing the right kernels
becomes a priority.

With our home-made backends, we run benchmarks automatically and choose the best configuration for
the current hardware and matrix sizes with a reasonable caching strategy.

This adds a small overhead by increasing the warmup execution time, but stabilizes quickly after a
few forward and backward passes, saving lots of time in the long run. Note that this feature isn\’t
mandatory, and can be disabled when cold starts are a priority over optimized throughput.

It is no secret that deep learning is mostly relying on matrix multiplication as its core operation,
since this is how fully-connected neural networks are modeled.

More and more, hardware manufacturers optimize their chips specifically for matrix multiplication
workloads. For instance, Nvidia has its Tensor Cores and today most cellphones have AI specialized
chips. As of this moment, we support Tensor Cores with our LibTorch, Candle, CUDA, Metal and WGPU/SPIR-V
backends, but not other accelerators yet. We hope
this issue gets resolved at some point to bring
support to our WGPU backend.

Burn aims to be the most flexible deep learning framework. While it\’s crucial to maintain
compatibility with a wide variety of backends, Burn also provides the ability to extend the
functionalities of a backend implementation to suit your personal modeling requirements.

This versatility is advantageous in numerous ways, such as supporting custom operations like flash
attention or manually writing your own kernel for a specific backend to enhance performance. See
this section in the Burn Book
for more details.

Contrary to the aforementioned backends, Autodiff is actually a backend decorator. This means that
it cannot exist by itself; it must encapsulate another backend.

The simple act of wrapping a base backend with Autodiff transparently equips it with
autodifferentiation support, making it possible to call backward on your model.

use burn::backend::{Autodiff, Wgpu};
use burn::tensor::{Distribution, Tensor};

fn main() {
    type Backend = Autodiff<Wgpu>;

    let x: Tensor<Backend, 2> = Tensor::random([32, 32], Distribution::Default);
    let y: Tensor<Backend, 2> = Tensor::random([32, 32], Distribution::Default).require_grad();

    let tmp = x.clone() + y.clone();
    let tmp = tmp.matmul(x);
    let tmp = tmp.exp();

    let grads = tmp.backward();
    let y_grad = y.grad(&grads).unwrap();
    println!(\"{y_grad}\");
}

Of note, it is impossible to make the mistake of calling backward on a model that runs on a backend
that does not support autodiff (for inference), as this method is only offered by an Autodiff
backend.

See the Autodiff Backend README for more details.

This backend decorator enhances a backend with kernel fusion, provided that the inner backend
supports it. Note that you can compose this backend with other backend decorators such as Autodiff.
For now, only the WGPU and CUDA backends have support for fused kernels.

use burn::backend::{Autodiff, Fusion, Wgpu};
use burn::tensor::{Distribution, Tensor};

fn main() {
    type Backend = Autodiff<Fusion<Wgpu>>;

    let x: Tensor<Backend, 2> = Tensor::random([32, 32], Distribution::Default);
    let y: Tensor<Backend, 2> = Tensor::random([32, 32], Distribution::Default).require_grad();

    let tmp = x.clone() + y.clone();
    let tmp = tmp.matmul(x);
    let tmp = tmp.exp();

    let grads = tmp.backward();
    let y_grad = y.grad(&grads).unwrap();
    println!(\"{y_grad}\");
}

Of note, we plan to implement automatic gradient checkpointing based on compute bound and memory
bound operations, which will work gracefully with the fusion backend to make your code run even
faster during training, see this issue.

See the Fusion Backend README for more details.

That backend simplifies hardware operability, if for instance you want to execute some operations on the CPU and other operations on the GPU.

use burn::tensor::{Distribution, Tensor};
use burn::backend::{
    NdArray, Router, Wgpu, ndarray::NdArrayDevice, router::duo::MultiDevice, wgpu::WgpuDevice,
};

fn main() {
    type Backend = Router<(Wgpu, NdArray)>;

    let device_0 = MultiDevice::B1(WgpuDevice::DiscreteGpu(0));
    let device_1 = MultiDevice::B2(NdArrayDevice::Cpu);

    let tensor_gpu =
        Tensor::<Backend, 2>::random([3, 3], burn::tensor::Distribution::Default, &device_0);
    let tensor_cpu =
        Tensor::<Backend, 2>::random([3, 3], burn::tensor::Distribution::Default, &device_1);
}

That backend has two parts, one client and one server.
The client sends tensor operations over the network to a remote compute backend.
You can use any first-party backend as server in a single line of code:

fn main_server() {
    // Start a server on port 3000.
    burn::server::start::<burn::backend::Cuda>(Default::default(), 3000);
}

fn main_client() {
    // Create a client that communicate with the server on port 3000.
    use burn::backend::{Autodiff, RemoteBackend};

    type Backend = Autodiff<RemoteDevice>;

    let device = RemoteDevice::new(\"ws://localhost:3000\");
    let tensor_gpu =
        Tensor::<Backend, 2>::random([3, 3], Distribution::Default, &device);
}

As you can see in the previous video (click on the picture!), a new terminal UI dashboard based on
the Ratatui crate allows users to follow their training
with ease without having to connect to any external application.

You can visualize your training and validation metrics updating in real-time and analyze the
lifelong progression or recent history of any registered metrics using only the arrow keys. Break
from the training loop without crashing, allowing potential checkpoints to be fully written or
important pieces of code to complete without interruption ?

ONNX (Open Neural Network Exchange) is an open-standard format that exports both the architecture
and the weights of a deep learning model.

Burn supports the importation of models that follow the ONNX standard so you can easily port a model
you have written in another framework like TensorFlow or PyTorch to Burn to benefit from all the
advantages our framework offers.

Our ONNX support is further described in
this section of the Burn Book .

Note: This crate is in active development and currently supports a
limited set of ONNX operators.

You can load weights from PyTorch or Safetensors formats directly into your Burn-defined models. This makes it easy to reuse existing models while benefiting from Burn\’s performance and deployment features.

Learn more:

• Import pre-trained PyTorch models into Burn
• Load models from Safetensors format

Several of our backends can compile to Web Assembly: Candle and NdArray for CPU, and WGPU for GPU.
This means that you can run inference directly within a browser. We provide several examples of
this:

• MNIST where you can draw digits and a small convnet tries to
  find which one it is! 2️⃣ 7️⃣ ?
• Image Classification where you can upload images and
  classify them! ?

Burn\’s core components support no_std. This
means it can run in bare metal environment such as embedded devices without an operating system.

As of now, only the NdArray backend can be used in a no_std environment.

To begin working effectively with Burn, it is crucial to understand its key components and
philosophy. This is why we highly recommend new users to read the first sections of
The Burn Book . It provides detailed examples and explanations
covering every facet of the framework, including building blocks like tensors, modules, and
optimizers, all the way to advanced usage, like coding your own GPU kernels.

The project is constantly evolving, and we try as much as possible to keep the book up to date
with new additions. However, we might miss some details sometimes, so if you see something weird,
let us know! We also gladly accept Pull Requests ?

Let\’s start with a code snippet that shows how intuitive the framework is to use! In the following,
we declare a neural network module with some parameters along with its forward pass.

use burn::nn;
use burn::module::Module;
use burn::tensor::backend::Backend;

#[derive(Module, Debug)]
pub struct PositionWiseFeedForward<B: Backend> {
    linear_inner: nn::Linear<B>,
    linear_outer: nn::Linear<B>,
    dropout: nn::Dropout,
    gelu: nn::Gelu,
}

impl<B: Backend> PositionWiseFeedForward<B> {
    pub fn forward<const D: usize>(&self, input: Tensor<B, D>) -> Tensor<B, D> {
        let x = self.linear_inner.forward(input);
        let x = self.gelu.forward(x);
        let x = self.dropout.forward(x);

        self.linear_outer.forward(x)
    }
}

We have a somewhat large amount of examples in the repository that shows how to use
the framework in different scenarios.

Following the book:

• Basic Workflow : Creates a custom CNN Module to train on the MNIST dataset
  and use for inference.
• Custom Training Loop : Implements a basic training loop instead
  of using the Learner.
• Custom WGPU Kernel : Learn how to create your own custom
  operation with the WGPU backend.

Additional examples:

• Custom CSV Dataset : Implements a dataset to parse CSV data for a
  regression task.
• Regression : Trains a simple MLP on the California Housing dataset
  to predict the median house value for a district.
• Custom Image Dataset : Trains a simple CNN on custom image
  dataset following a simple folder structure.
• Custom Renderer : Implements a custom renderer to display the
  Learner progress.
• Image Classification Web : Image classification web browser
  demo using Burn, WGPU and WebAssembly.
• MNIST Inference on Web : An interactive MNIST inference demo in
  the browser. The demo is available online.
• MNIST Training : Demonstrates how to train a custom Module (MLP) with the
  Learner configured to log metrics and keep training checkpoints.
• Named Tensor : Performs operations with the experimental NamedTensor
  feature.
• ONNX Import Inference : Imports an ONNX model pre-trained on MNIST to
  perform inference on a sample image with Burn.
• PyTorch Import Inference : Imports a PyTorch model pre-trained on
  MNIST to perform inference on a sample image with Burn.
• Text Classification : Trains a text classification transformer
  model on the AG News or DbPedia dataset. The trained model can then be used to classify a text
  sample.
• Text Generation : Trains a text generation transformer model on the
  DbPedia dataset.
• Wasserstein GAN MNIST : Trains a WGAN model to generate new handwritten digits
  based on MNIST.

For more practical insights, you can clone the repository and run any of them directly on your
computer!

We keep an updated and curated list of models and examples built with Burn, see the
tracel-ai/models repository for more details.

Don\’t see the model you want? Don\’t hesitate to open an issue, and we may prioritize it. Built a
model using Burn and want to share it? You can also open a Pull Request and add your model under the
community section!

Deep Learning is a special form of software where you need very high level abstractions as well as
extremely fast execution time. Rust is the perfect candidate for that use case since it provides
zero-cost abstractions to easily create neural network modules, and fine-grained control over memory
to optimize every detail.

It\’s important that a framework be easy to use at a high level so that its users can focus on
innovating in the AI field. However, since running models relies so heavily on computations,
performance can\’t be neglected.

To this day, the mainstream solution to this problem has been to offer APIs in Python, but rely on
bindings to low-level languages such as C/C++. This reduces portability, increases complexity and
creates frictions between researchers and engineers. We feel like Rust\’s approach to abstractions
makes it versatile enough to tackle this two languages dichotomy.

Rust also comes with the Cargo package manager, which makes it incredibly easy to build, test, and
deploy from any environment, which is usually a pain in Python.

Although Rust has the reputation of being a difficult language at first, we strongly believe it
leads to more reliable, bug-free solutions built faster (after some practice ?)!

In the event that you are trying to load a model record saved in a version older than 0.14.0, make
sure to use a compatible version (0.14, 0.15 or 0.16) with the record-backward-compat
feature flag.

features = [..., \"record-backward-compat\"]

Otherwise, the record won\’t be deserialized correctly and you will get an error message. This error
will also point you to the backward compatible feature flag.

The backward compatibility was maintained for deserialization when loading records. Therefore, as
soon as you have saved the record again it will be saved according to the new structure and you can
upgrade back to the current version

Please note that binary formats are not backward compatible. Thus, you will need to load your record
in a previous version and save it in any of the other self-describing record format (e.g., using the
NamedMpkFileRecorder) before using a compatible version (as described) with the
record-backward-compat feature flag.