#tensor-flow #neural-networks


Tiny, no-nonsense, self contained, TensorFlow and ONNX inference

32 releases

0.17.0 Jun 13, 2022
0.16.7 May 16, 2022
0.16.3 Mar 30, 2022
0.15.8 Nov 18, 2021
0.11.2 Oct 26, 2020

#6 in #tensor-flow

Download history 559/week @ 2022-03-11 686/week @ 2022-03-18 794/week @ 2022-03-25 827/week @ 2022-04-01 799/week @ 2022-04-08 897/week @ 2022-04-15 1729/week @ 2022-04-22 1344/week @ 2022-04-29 1190/week @ 2022-05-06 1428/week @ 2022-05-13 898/week @ 2022-05-20 1161/week @ 2022-05-27 978/week @ 2022-06-03 1351/week @ 2022-06-10 1130/week @ 2022-06-17 1038/week @ 2022-06-24

4,595 downloads per month
Used in 6 crates (2 directly)




rustc >= 1.59.0 MIT/Apache 2 Native Linux test status Embedded targets status Doc

Sonos' Neural Network inference engine.

This project used to be called tfdeploy, or Tensorflow-deploy-rust.

What ?

tract is a Neural Network inference toolkit. It can read Tensorflow 1, ONNX or NNEF, optimize them and run data through them.

Quick start

There is also some technical documentation and blog posts.

Tract in the landscape


As of today (October 2020), tract passes successfully about 85% of ONNX backends tests. All "real life" integration tests in Onnx test suite are passing: bvlc_alexnet, densenet121, inception_v1, inception_v2, resnet50, shufflenet, squeezenet, vgg19, zfnet512.

The following operators are implemented and tested.

Abs, Acos, Acosh, Add, And, ArgMax, ArgMin, Asin, Asinh, Atan, Atanh, AveragePool, BatchNormalization, BitShift, Cast, CategoryMapper, Ceil, Clip, Compress, Concat, Constant, ConstantLike, ConstantOfShape, Conv, ConvInteger, ConvTranspose, Cos, Cosh, CumSum, DepthToSpace, DequantizeLinear, Div, Dropout, DynamicQuantizeLinear, Elu, Equal, Erf, Exp, Expand, EyeLike, Flatten, Floor, GRU, Gather, GatherElements, GatherND, Gemm, GlobalAveragePool, GlobalLpPool, GlobalMaxPool, Greater, GreaterOrEqual, HardSigmoid, Hardmax, Identity, If, InstanceNormalization, IsInf, IsNaN, LRN, LSTM, LeakyRelu, Less, LessOrEqual, Log, LogSoftmax, MatMul, MatMulInteger, Max, MaxPool, Mean, Min, Mod, Mul, Neg, NonZero, Not, OneHot, Or, PRelu, Pad, ParametricSoftplus, Pow, QLinearConv, QLinearMatMul, QuantizeLinear, RNN, Range, Reciprocal, ReduceL1, ReduceL2, ReduceLogSum, ReduceLogSumExp, ReduceMax, ReduceMean, ReduceMin, ReduceProd, ReduceSum, ReduceSumSquare, Relu, Reshape, Resize, Round, Rsqrt, ScaledTanh, Scan, Scatter, ScatterElements, ScatterND, Selu, Shape, Shrink, Sigmoid, Sign, Sin, Sinh, Size, Slice, Softmax, Softplus, Softsign, SpaceToDepth, Split, Sqrt, Squeeze, Sub, Sum, Tan, Tanh, ThresholdedRelu, Tile, Transpose, TreeEnsembleClassifier, Unsqueeze, Where, Xor

We test these operators against Onnx 1.4.1 (operator set 9), Onnx 1.5.0 (operator set 10), Onnx 1.6.0 (operator set 11), Onnx 1.7.0 (operator set 12), Onnx 1.8.1 (operator set 13), Onnx 1.9.0 (operator set 14), and Onnx 1.10.1 (operator set 15). Many networks in operator set 8 are also working.

TensorFlow 1.x

Even if tract is very far from supporting any arbitrary model, it can run Google Inception v3 and Snips wake word models. Missing operators are relatively easy to add. The lack of easy to reuse test suite, and the wide diversity of operators in Tensorflow make it difficult to target a full support.

The following operators are implemented and tested:

Abs, Add, AddN, AddV2, Assign, AvgPool, BatchToSpaceND, BiasAdd, BlockLSTM, Cast, Ceil, ConcatV2, Const, Conv2D, DepthwiseConv2dNative, Div, Enter, Equal, Exit, ExpandDims, FakeQuantWithMinMaxVars, Fill, FloorMod, FusedBatchNorm, GatherNd, GatherV2, Greater, GreaterEqual, Identity, Less, LessEqual, Log, LogicalAnd, LogicalOr, LoopCond, MatMul, Max, MaxPool, Maximum, Mean, Merge, Min, Minimum, Mul, Neg, NoOp, Pack, Pad, Placeholder, Pow, Prod, RandomUniform, RandomUniformInt, Range, RealDiv, Relu, Relu6, Reshape, Rsqrt, Shape, Sigmoid, Slice, Softmax, SpaceToBatchND, Squeeze, StridedSlice, Sub, Sum, Switch, Tanh, Tile, Transpose, VariableV2

Addiotionaly, the complexity of TensorFlow 2 make it very unlikely that a direct support will ever exist in tract. Many TensorFlow 2 nets can be converted to ONNX and loaded in tract.


Long story short, TensorFlow and Onnx formats are good for designing and training networks. They need to move fast to follow the research field, tend to integrate new features and operators greedily. They also exhibit a high level of expressivity to facilitate network design.

On the other hand, only a subset of operators and network features actually reach production, so systems running production network do not have to deal with so many operators. Furthermore, some information required for training can be stripped from the network before going to production for prediction.

NNEF tries to bridge the gap between training frameworks and inference by proposing a format dedicated to production and prediction.

Tract supports NNEF:

  • tract_nnef can load and execute NNEF networks
  • tract supports most of the NNEF specification, the most notable exception being the ROI operators and deconvolution
  • tract introduces tract-OPL, a series of NNEF extensions to support other operators (or extend some operators semantics) in order to represent the full range of tract-core neural network support: any network understood by tract should be serializable to tract-OPL. This is a work in progress.
  • tract command line can translate networks from TensorFlow or ONNX to NNEF/OPL.

Example of supported networks

These models among others, are used to track tract performance evolution as part of the Continuous Integration jobs. See .travis/README.md and .travis/bundle-entrypoint.sh for more information.

Keyword spotting on Arm Cortex-M Microcontrollers


ARM demonstrated the capabilited of the Cortex-M family by providing tutorials and pre-trained models for keyword spotting. While the exercise is ultimately meant for micro-controllers, tract can run the intermediate TensorFlow models.

For instance, on a Rasperry Pi Zero, the "CNN M" model runs in about 70 micro-seconds, and 11 micro-seconds on a Raspberry Pi 3.

Snips wake word models


Snips uses tract to run the wake word detectors. While earlier models were class-based and did not require any special treatment, tract pulsing capabilities made it possible to run WaveNet models efficiently enough for a Raspberry Pi Zero.

Inception v3

Device Family TensorFlow-lite tract
Raspberry Pi Zero Armv6 VFP 113s 39s
Raspberry Pi 2 Armv7 NEON 25s 7s
Raspberry Pi 3 aarch32 NEON 5s 5s


  • while the Raspberry Pi 3 is an Armv8 device, this bench is running on Raspbian, an armv6 operating system, crippling the performance of both benches
  • there exists other benches on the internet that show better performance results for TensorFlow (not -Lite) on the Pi 3. They use all four cores of the device. Both TensorFlow-Lite and tract here have been made to run on a single-core.


Note: files in the tensorflow/protos directory are copied from the TensorFlow project and are not covered by the following licence statement.

Note: files in the onnx/protos directory are copied from the ONNX project and are not covered by the following licence statement.

Apache 2.0/MIT

All original work licensed under either of


Unless you explicitly state otherwise, any contribution intentionally submitted for inclusion in the work by you, as defined in the Apache-2.0 license, shall be dual licensed as above, without any additional terms or conditions.


~257K SLoC