mirror of https://github.com/tracel-ai/burn.git synced 2026-05-31 19:49:48 +09:00

Files

Guillaume Lagrange 27964e0160 Refactor tensor kind traits (#4957 )

* Remove generics from tensor kind traits and Elem associated type

* Move backend extension, re-export no types from burn_backend and fix into_scalar missing types

* Remove into/from primitive usage in burn-optim

* Fix backtrace

* Fix float_sort_with_indices int dtype

* Fix feature gated

* Remove dead code

* Add tensor primitive note

* Another note

* High level kind

* Working but still pub traits

* Fix into_scalar doc example

* Fix argsort out dtype

* Restrict ops traits to pub(crate)

* Add other backends for burn-vision

2026-05-14 14:02:44 -04:00

src

Refactor tensor kind traits (#4957 )

2026-05-14 14:02:44 -04:00

weights

Support importing safetensors format (#2721 )

2025-05-06 12:51:08 -04:00

Cargo.toml

Centralize internal burn-* deps in [workspace.dependencies] (#4876 )

2026-04-27 08:13:24 -04:00

README.md

Move ONNX crates to burn-onnx repository (#4393 )

2026-01-28 14:38:11 -06:00

README.md

Import Model Weights

This crate provides examples for importing model weights from different formats to Burn.

Examples

PyTorch

Imports weights from a PyTorch .pt file using burn-store.

cargo run --bin pytorch -- <image_index>

Example:

cargo run --bin pytorch -- 15

Loading PyTorch model weights from file: weights/mnist.pt
Image index: 15
Success!
Predicted: 5
Actual: 5
See the image online, click the link below:
https://huggingface.co/datasets/ylecun/mnist/viewer/mnist/test?row=15

Safetensors

Imports weights from a Safetensors file using burn-store.

cargo run --bin safetensors -- <image_index>

Example:

cargo run --bin safetensors -- 42

Loading Safetensors model weights from file: weights/mnist.safetensors
Image index: 42
Success!
Predicted: 4
Actual: 4
See the image online, click the link below:
https://huggingface.co/datasets/ylecun/mnist/viewer/mnist/test?row=42

Convert

Converts between different weight formats (PyTorch or Safetensors) to Burn's native Burnpack format.

cargo run --bin convert -- <format> <output_directory>

Where:

<format>: Either pytorch or safetensors
<output_directory>: Path to save the converted model file

Example with PyTorch:

cargo run --bin convert -- pytorch /tmp/burn-convert

Loading PyTorch weights from 'weights/mnist.pt'...
Saving model to '/tmp/burn-convert/mnist.bpk'...
Model successfully saved to '/tmp/burn-convert/mnist.bpk'.

Example with Safetensors:

cargo run --bin convert -- safetensors /tmp/burn-convert

Loading Safetensors weights from 'weights/mnist.safetensors'...
Saving model to '/tmp/burn-convert/mnist.bpk'...
Model successfully saved to '/tmp/burn-convert/mnist.bpk'.

Burnpack

Demonstrates loading and using a model from Burn's native Burnpack format.

cargo run --bin burnpack -- <image_index> <model_path>

Where:

<image_index>: Index of the MNIST test image to classify
<model_path>: Path to the model file (without extension)

Example:

cargo run --bin burnpack -- 35 /tmp/burn-convert/mnist

Loading model weights from file: /tmp/burn-convert/mnist.bpk
Image index: 35
Success!
Predicted: 2
Actual: 2
See the image online, click the link below:
https://huggingface.co/datasets/ylecun/mnist/viewer/mnist/test?row=35

Workflow

A typical workflow using these examples:

Start with pre-trained weights in either PyTorch or Safetensors format
Use the convert example to convert to Burn's native Burnpack format
Load and use the converted model with the burnpack example